This was a tough call, because we continue to believe in our thesis that domain experts will own AI applications, and better tooling is needed to enable this. However, our team hasn’t been strong enough on GTM, particularly top-of-funnel, and we haven’t been able to fix that at the founder level. For a more detailed explanation, we’ve written a full post-mortem here.

Over the next couple of months, we’ll be offboarding customers, speaking with potential acquirers, and preparing for a clean wind-down. We’ll also be thinking about what’s next in our careers!

What’s next for us?

We (Olly, Yousef and Sam) are broadly interested in applied AI roles. We each combine an AI PhD, full-stack engineering and commercial experience - a mix that’s well suited to delivering real-world value from AI. We’re open to a range of roles across company sizes and industries, so please get in touch if you come across anything that may be a good fit.

Lastly, we want to say a big thanks to you all for reading these updates. Every response has been a boost along our (often tumultuous) journey. We hope you’ll keep in touch in the future!

En Taro Tassadar,
Artanis Team

🙋 Ways you can help - read and share our new AI blogs 🙋
No matter how good AI models get, they can’t read your mind. We’ve published several longer blogs recently on how to clearly communicate what you want your AI application to do, or your “Policy”, in scenarios where the correct output is subjective. We recommend starting with this post - please subscribe / share if you like the content!

📉 Progress in April - mixed signals 📈
We missed our April target, which was to grow from 4 to 5 customers. Our metrics were:

Customers: 3 (-1)
Monthly revenue: $350 (-$50)

Of the four customers we signed in March, three were highly engaged in April. They invested significant amounts of their own team’s time into using Artanis and/or bought more seats, which is why our revenue only fell by $50. We felt we were delivering clear value to them.

One customer churned. The churn reasons were more about lack of intent to write evals / alter prompts in the foreseeable future. This was likely a consequence of being fairly relaxed in taking on customers outside our ICP in April, rather than problems in our product.

We didn’t sign up any new customers. Top-of-funnel particularly has consistently been a problem for us as a technical founding team, and we’re struggling to cut through the noise in the evals space. We’re also finding it hard to communicate how Artanis is superior to building in-house tooling for domain experts to review AI outputs, which is our main competition.

🏹 Goal for May - tread water… 🏹
We normally set growth targets then, if we miss them, reflect on what went wrong. However, Sam is having his first child in May and it’s not realistic to aim for growth when ⅓ of our team will be incapacitated. So for May we’ll be focused on delivering for current customers, and treating any growth as a bonus!

🙏 Shout-outs 🙏
Special thanks for April go to:

Finn W - for the detailed look at evaluating Claude skills
Ainhoa A - for the call about CS and prompt management
Michael T - for getting us into Sifted
Peter L - for helping find solutions to data sensitivity
Paul B - for getting us into Lenny’s Newsletter
Abbas & Qasim - for insights on onboarding to their products
Eve T - for making our day with your PhD topic!

En Taro Tassadar,
Artanis Team

🙋 Ways you can help - watch and share our demo video 🙋
The response to our new product has been solid, so we’re looking to get more eyes on it. People who review AI output are already doing all the work needed to create evals - they’re just not capturing it. Our product turns their work into evals, which lets teams manage prompts more systematically and feel less like playing whack-a-mole.

Please watch our 2-minute demo video, and share it with people who are managing prompts!

📉 Progress in March - exceeding targets! 📈
Our metrics for March were:

Customers: 4 (+4)
Monthly revenue: $400
NB: we’re not reporting customers on old products that we’re retiring.

Our goal was to sign up the first customer for our new product, so we were thrilled to get four instead! They’re all building AI products and have shifted prompt management away from engineering towards domain experts, such as:

1/ Nurses, at a healthtech
2/ Teachers, at an edtech
3/ Product managers

The unifying theme among customers who responded well was that their engineers didn’t have the domain expertise to write evals or manage prompts. They want a product that enables domain experts to do this instead, as they were aware it would free up engineering time to focus on technical issues, rather than trying to evaluate output in domains they don’t understand.

💡Challenges - a multiplayer problem💡
Writing evals and managing prompts usually involves multiple stakeholders. 

1/ The domain experts are responsible for deciding what “right” and “wrong” look like (i.e. writing the evals)
2/ Engineers are responsible for building the system that performs well against those evals 

Our solution needs to please both stakeholders, which slows down sales cycles. This is because it’s often unclear who the main decision-maker is; it could be either stakeholder. But it’s also because we need to build something that works for both, and they need different interfaces. Domain experts prefer a visual UI, whereas engineers want to work programmatically. These multiplayer dynamics are common in enterprise sales, but we’re finding them even when selling to seed-stage startups.

🏹 Goal for April: grow from four to five 🏹
Our main goal for April is to grow from four to five customers. This will require retaining all of our initial customers and signing up at least one new customer.

We’d also like to see at least one of our first four customers clearly getting value from Artanis for prompt management. It’s harder to define a metric for this at our stage - it’s more of a “know it when you see it” - but we’ll report on this too.

🙏 Shout-outs 🙏
Special thanks for March go to:

Peter L - for being such an avid reader of our updates
Sergey C - for doing a call with us in the middle of a move
David R - for putting in the yards on email review
Gunisha V - for mobilising your team to make intros
Yiming Y - for detailed value prop feedback
Dom B - for helping get the band back together
Chidi W - for being a good sport
Michael T - for both taking the leap and making intros!
Adam R - for going the extra mile in thinking about leads
Lorenzo S - for the office tour

En Taro Tassadar,
Artanis Team

1) They’re at a startup building an AI product
2) They’re a non-technical domain expert, like a Product Manager or Customer Service rep
3) They’re responsible for managing prompts or writing evals

Please get in touch if you know anyone who fits the bill!

🔬 A seamless way to write evals 🔬

We’ve come up with a new product! Now you can write systematic evals by just giving a thumbs up/down with a short reason. This hugely reduces the time and effort required compared to defining “correct” output from scratch. We’re very excited about this, because we think it gets to the core problem in AI development today - efficiently aligning output to human preferences.

📉 Progress in February 📈
Our metrics for February were:

Monthly revenue: $300 (unchanged)
Customers: 1 (unchanged)

Despite our efforts to bring down the engineering effort required to integrate Artanis, we missed our growth targets for the second month running. The main themes were:

1) Domain experts, such as PMs and CS teams, are becoming the frontline for dealing with AI output quality issues. They do this by changing the prompts, which doesn’t usually require code changes. The burden of dealing with AI quality issues is therefore shifting from engineers to domain experts. This shift is most obvious in the products that have the most traction.

2) Actioning human feedback isn’t a big timesink for most teams. Once an example has been flagged as a problem, it’s often relatively quick to find the root cause. The bigger issue was knowing which examples should be looked into, because AI teams often aren’t collecting human feedback to flag them.

3) We met several teams who resonated with the problem, but had recently vibe-coded an internal tool that analysed their traces. It seemed to be working well enough for them, so they’d pretty much solved their own problem. This shift in “buy vs build” is a broader trend in SaaS that’s being talked about a lot - we got on the wrong end of it.

Ultimately, we felt we were on the wrong side of history, particularly about the first theme. Ownership of AI prompts will shift from engineers to domain experts as LLMs get better, because the main problem is becoming alignment rather than intelligence. This is already happening and it’s most obvious in the companies with the most traction. While currently a niche, we believe that domain experts owning prompts will become the dominant paradigm. This is the market we want to build for.

💡 New persona, new problems💡
Domain experts who own prompts currently suffer from the “whack-a-mole” problem. They make changes to prompts to fix problems, but this often causes regressions in previous behaviour or new problems to appear. This leads to a feeling of going in circles, and a lack of confidence that prompt changes will make the AI better overall.

The best way to stay on top of this is to have systemic evals. However, there’s currently a cultural issue where responsibility for evals is falling between domain experts and engineering. Engineers don’t have the expertise to write the evals themselves. And domain experts can’t convert their intuitive notion of right/wrong into automated tests. 

Until recently, we’d been quite pessimistic about solving this with software. We felt the main problem with evals was cultural, not technical. It’s currently laborious for domain experts to write good evals. While they can quickly spot when something looks wrong, it’s hard to articulate upfront what “good” output looks like, particularly as AI products often do many things at once. We felt the only route through was to more heavily incentivise domain experts to write evals. 

We can now make it seamless for domain experts to write evals, which may reduce the incentives needed. Curious about how? Watch our demo above!

🏹 Goal for March: zero-to-one on the new product 🏹
We’ve got a new product, so we need to pick up our first customer for it! This will either be an existing customer on a prior product we’ve built, or someone new.

🙏 Shout-outs 🙏
Special thanks for February go to:

Cait C - for going the extra mile
Henry M - for the intro to Danai
George T - for being very generous about our work with Convergence
Michael F - for connecting us to Beam
Polina M - for direct and clear feedback
Asita R - for the kind words and speaking opportunity
Ibrahim J - for being super responsive on intros
Kieran G - for being proactive with thinking of Manos
Simon and Dylan - for insights into the growing role of CS in AI development
Namid S - for bringing your whole team to a demo!

En Taro Tassadar,
Artanis Team

🙋 Ways you can help - intros to AI teams 🙋
We’d like to speak with people with the following profile:

1/ Technical leader e.g. CTO, Head of AI/Engineering/Product
2/ They are building an LLM-based product and have users/revenue
3/ They have fewer than 100 employees

Please get in touch if you know anyone who fits the bill!

📉 Progress in January 📈
Our metrics for January were:

Monthly revenue: $300 (unchanged)
Customers: 1 (unchanged)

We failed to meet our target to grow from 1 to 2 customers. We retained our first customer, as they saw the value in even a very early version of our new product. However, we didn’t sign up a new customer. Reflecting on our main hypotheses going into January:

1/ Our “Stacktrace for AI mistakes” makes it faster for AI teams to action human feedback
This was mostly true. The concept was clear and resonated with AI teams as a time-saver. We booked 12 demos across January and customers clearly understood the potential value. Top-of-funnel was also a strong point. We held another ~35 discovery calls and we don’t take this for granted, given that getting customers to talk to you is the hill that many startups die on!

2/ Actioning human feedback on AI output quality is a significant resource commitment.
The evidence on this was mixed. This was likely the main cause of lost deals, so we write in more detail below about how we’re planning to respond.

🔬 Learnings about actioning human feedback on AI output 🔬
We held 12 demos in January but didn’t convert a new customer. The main objections were:

1/ Artanis takes too much engineering effort to integrate.
Human feedback about AI comes at irregular “bursty” intervals. Engineers may spend several days in a row looking at traces, followed by weeks of inactivity. However, the first version of our product wasn’t self-serve; we needed 2 weeks to onboard new customers. This meant there was a mismatch between the “bursty” nature of the problem and how quickly we could solve it. We’re now addressing this by making our product self-serve, so they can get up and running in minutes rather than days.

2/ We’re not collecting enough human feedback on AI output.
This was usually due to their product being pre-launch or not having enough usage. In these cases, our product doesn’t save much engineering time. This isn’t an objection we can address; it’s a qualification criterion that may impact our commercial viability. We’re taking a bet that this is a “small-but-growing” problem: AI products will pick up more usage over time, and that companies will collect more human feedback to improve their AI.

3/ We don’t want to pay upfront, can we do a free trial instead?
We lost deals because we require 1 month payment upfront and offer a refund if they’re unhappy, instead of offering a free trial like most software products. We don’t want to address this just yet, as it’s an intentional decision to select early adopters who most strongly need our product. We’d rather have 2-3 highly engaged customers than a larger number who’re less involved.

🏹 Goal for February - grow from 1 to 3 customers 🏹
We’re aiming to grow from 1 to 3 customers, with the following product changes:

1/ Let customers use Artanis on top of their existing tracing, rather than replacing it.
2/ Enable self-serve installation within an hour, rather than requiring a 2-week onboarding.
3/ Allow manual annotation of human feedback on traces, rather than requiring a fully automated connection.

These should shorten time-to-value, so we can deliver during the short “bursty” periods where they’re looking at traces. It should also reduce the engineering effort required from their team to integrate Artanis. 

🙏 Shout-outs 🙏
Special thanks for January go to:

Paul F - for being January’s intro MVP!
Zoe M - for the intro to Clement
Conrad - for continued support
Sogo - for the several-degrees-removed effort
Matt R - for the CS idea and DX suggestions
Marissa R - for thinking of Paraglide
JJ, Dominic, Sergey & Yiming - for very clear feedback after a demo
Csongor - for several good intros
Sivesh - for being a Value-Add-Investor
Hanna L - for proactive help with other startups
Ismail S - for the intro to Sanj
Bailey & Rod - for inviting us to speak at a great AI eng. event
Henry I - for connecting us to your CS team
Gunisha & Archana - for architectural advice

En Taro Tassadar,
Artanis Team

🙋 Ways you can help - intros to AI teams 🙋
We’d like to speak with people with the following profile:

1/ Technical leader e.g. CTO, Head of AI, Head of Engineering
2/ They have an LLM-based product in the market
3/ Their company is post-revenue, but has fewer than 100 employees

Please get in touch if you know anyone who fits the bill!

💡 New product - Stacktrace for AI mistakes💡
When code fails to run, the Stacktrace shows exactly where in the codebase the problem originated. This makes it much faster for engineers to fix software mistakes, as they don’t need to dig through every line of their codebase. Stacktracing is a critical part of the software development workflow.

Fixing AI mistakes is much harder, as they’re not caught by the compiler. Instead, engineers find out about AI mistakes because of human feedback. This means that people - either users or internal team members - complain about the output. It’s time-consuming to find the precise cause of the mistake, as it could be:

1/ In an intermediate step of a complex pipeline. Engineers need to inspect the input/output in every step of their pipeline to figure out which step caused the problem.

2/ The mistake was in the data being given to the AI, such as out-of-date docs. In this case, engineers need to dig through the raw input data rather than their pipeline’s inputs/outputs.

3/ The person complaining was wrong and the AI was right. In this case, engineers often chase their tail looking for an AI problem that doesn’t exist.

We’re solving this problem by building a “Stack trace for AI mistakes”. When people complain about AI output, we’ll automatically determine the root cause of the mistake from the buckets above. If it’s an AI problem, rather than a human mistake, we’ll also identify the specific part of the pipeline that caused it. This will free up engineering time to focus on fixing the AI problem, rather than digging through data.

How did we get here? See the next section if you care for the details!

🔬 Customer discovery - the state of teams building AI products 🔬
We started December with three hypotheses about companies building AI products. Here’s what we learned over the course of 35 discovery calls.

1/ There’s a strong link between their AI accuracy and revenue.
This was correct, but framed incorrectly. AI accuracy is a priority for most teams, but none could quantify a direct link to revenue. Instead, the focus on accuracy is because of concerns such as churn, winning pilots, brand and legal compliance. Most of these are linked to revenue, so we’re happy to move forward with helping teams improve accuracy.

2/ The main bottleneck to better accuracy is a lack of robust measurement (or “evals”).
Not usually - instead, the most commonly cited bottleneck was lack of engineering resource. While most teams think evals are important, they aren’t usually a current priority. The main challenge in writing better evals is the relationship with the domain expert, who often doesn’t have an incentive to put time into this or awareness of its importance. The solution to this challenge is cultural, rather than software.

3/ Existing monitoring solutions can measure inputs and outputs, but not accuracy.
This was correct. There’s no dominant observability solution: a plurality (but not a majority) of teams build in-house. There’s also a long tail of 3rd party solutions (e.g. Langsmith, Langfuse), which teams often buy in the hope they’ll fix their problems with evals. However, we didn’t meet any teams where this worked in practice - they just use the I/O monitoring features and reverted to managing their evals separately.

📉 Progress in December - zero-to-one! 📈
Our metrics for December were:

Monthly revenue: $300
Customers: 1

Signing the first customer for a new product within a month was a big win! This was particularly fast, given the run-up to Xmas is usually poor timing for closing deals. It also feels repeatable: we had good results with our outbound motion, booking 39 discovery calls in 3 weeks through a mix of warm intros and cold calling. This gives us confidence to commit further to our current route.

🏹 Goal for January - grow from 1 to 2 customers 🏹
We’re aiming to grow from 1 to 2 customers. To do so, we’ll need to both i) retain our first customer, and ii) acquire a new customer. Our main hypotheses are now changing to:

1/ Actioning human feedback on AI output quality is a significant resource commitment.
If this is true, teams will pay for solutions that save them time.

2/ Our “Stacktrace for AI mistakes” makes it faster for AI teams to action human feedback.
There are two parts to this: i) we’ll sign new deals if prospects believe it, and ii) we’ll retain existing customers when we’ve deployed this in practice.

🙏 Shout-outs 🙏
Special thanks for December go to:

Chidi W - for being customer #1
Lorenzo S - for being an intro MVP
Ollie T - for a good chat on buttons vs text input!
Aleks M - for the quote about gradients
Zahid M - for the intro
Sunny Z - for the chat on compliance
Jeff K - for the industry insights
Henry M - for the deep dive into your website time machine
David S - for the legal tech call
Laura R - for the extra hustle
Sergey C - for being such an avid update reader!
Nick E - for providing some pre-LLM perspective
Ben L - for the intro to Pravin
Jeylani J - for the look into where CS agents get tricky
Tom V - for the AI grading catch-up
David R - for holding us accountable
Nisarg M - for the Raft intro

En Taro Tassadar,
Artanis Team

🙋 Ways you can help - intros to AI teams 🙋
We’d like to meet people with the following profile:

1/ Startup with an LLM-based product in the market
2/ Technical leader e.g. CTO, Head of AI
3/ The startup is post-revenue, but has fewer than 100 employees

Please get in touch if you know anyone who fits the bill!

📉 Progress in November - ruling out FMCG operations 📈
Our main goal for November was to commit to a narrower product, because initial discovery calls suggested that a horizontal solution for process automation was too broad. We weren’t able to find a compelling opportunity for a narrower product after considering the following:

1/ Order processing - there was some market pull for this, but it wasn’t a slam-dunk. Only some FMCG ops teams do this manually, and only for a fraction of their orders. There are also other AI order processing solutions. Overall, we felt there was a viable business opportunity, but it would’ve been an uphill struggle and not a strong fit for our team.

2/ Demand planning - this came up repeatedly on customer calls as a major pain point. However, it’s a longstanding problem that mostly stems from poor data availability. We therefore felt like there was no compelling “why now” for how we could solve it as a team.

A bright spot was great results from cold calling, as we improved our lead selection and how we asked questions on the call. This got a lot of interest in the last update, so we decided to share the metrics again:

- 264 dials (across ~100 prospects)
- converted to 49 answered
- converted to 15 discovery calls

Ruling out a strategy after speaking to customers isn’t a bad thing - it’s arguably the name of the startup game. But it does lead to a difficult question: what’s next?

🤔 How well do you need to understand your customers? 🤔
We didn’t feel in a strong position when speaking to FMCG operational leaders. While we felt credible when talking about AI, we didn’t have expertise when broader operational topics came up. This is tolerable if you’re willing to put in the time to build expertise. We’ve written before on how well you need to understand your customer, and realised we weren’t following our own advice!

We were also concerned that we were becoming “AI in search of a problem”. We wanted to deal with customers for whom AI is the problem, rather than (potentially) the solution. AI is most likely to be the problem for AI teams.

We decided our next move needs to take us back to a customer segment we felt stronger with - teams building AI products. Our expertise is AI, and that’s where we want to keep building.

🐌 The hard thing about going back to square one 🐌
Pivoting customer segment is a major decision. It takes us back to being not sure what we need to build, or even exactly what problems we’re looking to solve. The team we’d hired in March, initially to scale a Palantir-type approach, was no longer the right fit for the new direction. As a result, our team has shrunk from five to three.

If you’re looking for an AI engineer, then get in touch or reach out to Andrew directly. He's excelled at our customer-facing work, getting rapidly up to speed with existing customers and building solutions for new customers. He brings a rare combination of technical depth and stakeholder-engagement/communication skills.

🏹 Goal for December - validate accuracy monitoring 🏹
Monitoring accuracy in production is a problem for AI teams. It’s why observability and tracing tools get a lot of attention: AI teams hope that it will allow them to monitor accuracy after deployment. However, they’re missing a big piece. While they can monitor inputs and outputs, they can’t verify whether the output was correct. This would require knowledge of precisely what the model should have done, i.e. the “ground truth”.

We have a hunch that there’s a gap in the market for a monitoring platform that enables robust accuracy measurement. Our main goal is to validate three key hypotheses by speaking with AI teams:

1/ There’s a strong link between their AI accuracy and revenue
2/ The main bottleneck to better accuracy is a lack of robust measurement (or “evals”)
3/ Existing solutions can measure inputs and outputs, but not accuracy

🙏 Shout-outs 🙏
Special thanks for November go to:

Srecko D - for being a great “sparring partner” (again)
Conrad L - for always looking to help
Lorenzo S - for good advice, over Vietnamese
Henry M - for being the most helpful £1k angel in the UK
Guillaume B - for not one but two informative discovery calls
Paul F - for the intro to Guillaume!
Mo N - for repeating our advice back to us
Oliver W - for the detailed look at where accuracy matters
Amil K - for the shout on exploring compliance use cases
Zahid M - for LLM-as-judge in prod insights
Stefano G - for insights on how accuracy affects buyer trust
Maurice B - for the intro to JJ, and plugging our Substack!

En Taro Tassadar,
Artanis Team

🙋 Ways you can help - intros to operational leaders 🙋
We’d like to speak with people with the following profile:

1/ Operational role in their business e.g. COO, Director of Operations, Chief of Staff, Head of Operations
2/ The business has >£10m annual revenue
3/ They’re in operationally heavy sectors, such as F&B or FMCG

Please get in touch if you know anyone who fits the bill!

📉 Progress in October - working on top-of-funnel 📈
Our main goal for October was to book 25 discovery calls, and we ultimately booked 20. Of the 20 booked calls, 15 were from cold calling, which was our team’s first foray into this. Once we got over the “icky” feeling and found an authentic way to speak, it was super effective! Our metrics from ~3 weeks of cold calling were:

- 1505 dials (across ~250 leads), of which…
- 156 answered the phone, of which…
- 15booked discovery calls, of which…
- 0 became new customers (yet!)

From the first ~5 discovery calls, the following themes are starting to emerge:

1/ Demand planning is a very widespread issue. Companies often struggle to predict swings in demand, which can lead to both overbuying stock and lost revenue from supply shortages. The root cause of this problem is complex, and it’s unclear whether AI is part of the solution.

2/ Manual order processing is a common pain point for F&B companies that sell B2B. This tracks with our experience from a recent consulting project, which is a good sign. However, we’re still unsure how widespread and how costly it is.

3/ Mid-market companies may not be operating at the scale to make automating internal operations, such as order processing, worthwhile. It’s consistently described as “on the roadmap” but not an immediate priority.

We haven’t yet been able to draw strong conclusions, as most of our discovery calls are scheduled for November. We’ll be digging further into the themes above.

🏹 Goal for November - commit to a narrower product 🏹
We’re learning that operational leaders aren’t looking for a horizontal platform to help with AI transformation. Instead, they’re more inclined to seek solutions for specific problems. It’s also been hard to clearly communicate how our platform can address multiple AI use cases, and puts too much work onto the customer to figure out its value.

Our main goal in November is therefore to commit to a narrower proposition, using the learnings from our discovery calls. This may focus on automating just one defined process. For example: if 15 out of 25 companies report that order processing is an immediate priority, then we’d focus on that. Our long-term vision remains to enable more widespread process automation, but our initial focus would be on automating a single process.

🙏 Shout-outs 🙏
Special thanks for September go to:

David R - for getting us over the cold calling hump
Paul F - for intros to Guillaume & Nikesh
Rod R - for inviting us onto your AMA cast
Callum NF - for an interesting F&B opportunity
Paul B - for sharing F&B insights
Benji F - for intro to Paul B
Phil P - for the invitation to speak
Lily F - for being a great co-host

En Taro Tassadar,
Artanis Team

🙋 Ways you can help #1 - watch our new demo video 🙋‍♀️
We’ve put together a 2-minute demo video explaining the new product we’re launching. We’d love it if you could provide feedback (particularly about anything you find unclear). You can keep any feedback on the quality of my voice acting to yourself!!

🙋 Ways you can help #2 - survey on using AI in your business 🙋‍♀️
We're undertaking some research on how businesses are using AI in their workflows. Please fill out this survey if you have 10 minutes, and pass it on to anyone else who’s using AI to improve workflows in their business!
https://scaleupquestions.typeform.com/operationalAI

📉 Progress in September - new product launch! 📈
We’re launching a platform that enables companies to automate their manual processes, without needing software engineers. They can train an AI to automate their work just by doing it on Artanis. There’s a close analogy with training a new employee to take over a process:

1/ You first do some examples in front of the new employee to show them how it’s done.

2/ After they’ve seen enough examples, they do the job themselves and you review their work. You provide feedback on their output to train them.

3/ When you no longer need to provide feedback, they’re good enough to do the job themselves. You stop reviewing their work and delegate ownership to them.

How does this work? Watch our 2-minute demo video to find out!

🏹 Goal for October - de-risk top of funnel 🏹
We need to prove willingness to pay for our new platform. The ultimate metric we need to grow is # of customers. We’re setting up a new B2B outbound funnel to achieve this:

1. Identify prospects
2. Book discovery calls to qualify them
3. Write proposals for qualified leads
4. Close the deal

However, in October our main goal is booking 25 discovery calls. This is because i) the biggest risk in this funnel is booking discovery calls, and ii) sales cycle length makes it unrealistic to close a new customer in October. For the next few updates, we’ll report on:

Number of discovery calls
Number of customers
Monthly revenue

🙏 Shout-outs 🙏
Special thanks for September go to:

Sami T - for kind words plugging our Substack
Thomas W - for the warm intro
Sakib A - for advice on personal brand marketing
Johann & Srecko - for a great B2B sales workshop
Mariam A - for advice on growth stacks

En Taro Tassadar,
Artanis Team

🙋 Ways you can help - intros to operational leaders 🙋
We’d like to speak with people with the following profile:

1/ Operational role in their business e.g. COO, Director of Operations
2/ The business has >£10m annual revenue
3/ They’re looking to use AI to improve the efficiency of manual processes

Please get in touch if you know anyone who fits the bill!

📉 Progress in August - slowing down 📈
Our metrics from August are…

Method customers: 2 (-)
Total customers: 3 (-1)
Monthly Revenue: £27k (-£29k)
Team Size: 5 (-)

The highlight was signing a new customer. They’re a strong fit for our thesis around using AI to improve operational efficiency, and we’ve made a good start to our pilot.

However, we missed our growth target. This was partly due to churn from a long-term customer, for reasons outside our control. However, we also failed to convert one of our pilots into a continuation deal. The AI we developed wasn’t critical enough to their business to make it worth further investment.

We set targets each month to force ourselves to evaluate strategy if we miss them. This is the third month in a row that we’ve not met our targets. While August is usually slow, there are underlying problems we need to address. We’re now going to change direction more quickly.

🧐 Why are we revising our strategy? 🧐
It’s been nearly six months since we raised. In some ways, it’s been a great six months! We’ve built a strong founding team that’s working well together and learned a lot about delivering AI projects. Unusually, we’ve done this without spending any of our capital due to high revenue.

However, we’ve been too cautious. We’ve been unwilling to let go of consulting revenues and have avoided taking the risk of productising. While being cash generative is nice, we raised VC funding so that we could take risks instead. We need to do so more aggressively going forward.

We’ve also had three months of missed targets in a row. Our strategy of i) acquiring customers for AI consulting projects, then ii) converting them to longer-term partnerships, hasn’t been leading to growth.

The main issue has been post-pilot churn. This was partly expectation misalignment: some customers were only looking for a one-off project, followed by a handover. Also, our initial consulting customers were largely startups whose business models weren’t yet stable.

Another issue has been slow distribution. We positioned ourselves as a service, which was hard to sell outbound. Service buying decisions are usually made via referral or, for bigger companies, from a whitelist of known providers.

🏹 Goal for September - confirm new strategy & relaunch 🏹
Our top priority is setting a new strategy to address the problems above. We know that it will broadly involve focusing more on our AI platform and longer term deals, rather than one-off projects. But we’re in the process of ironing out the details, so it’s too soon to set target metrics. We’ll do that in the next update!

🙏 Shout-outs 🙏
Special thanks for August go to:

Katie - for some great sessions
Finn W - for being proactive with intros
Rod F - for a very thoughtful response
Helene - for the team photos
Andy F - a strong start and complementary beer

En Taro Tassadar,
Artanis Team

🙋 Ways you can help 🙋
We’ve been posting more on our Substack and would appreciate likes, comments and shares!

Check our new posts out below and let us know what you think:

1. 90% accuracy often brings no value
2. The Easy Way vs The Hard Way in AI
3. Why “Bikeshedding” kills so many AI projects

📉 Progress in July - mixed signals 📈
Our metrics from July are…

Method customers: 2 (+1)
Total customers: 4 (-1)
Monthly Revenue: £56.3k (-£19.8k)
Team Size (proxy for cost): 5 (-)

We completed the pilot with one of our customers and have moved on to a longer-term deal to continue developing their AI projects using our Methodology. This was our primary goal for the month, so was a big win!

We also successfully finished up a second pilot. We’re still negotiating a longer-term deal so we won’t yet mark that as a “Method customer”.

Another customer moved down to a maintenance contract, which has impacted MoM revenue.

New business has been a little slower for the last few months. We’ve just started working with a marketing agency to sharpen our positioning and improve top of funnel.

We’re also starting to pitch an entry-level offering, which we’re calling “lightning projects”, where we focus solely on building AI (and not custom system integrations, which adds significant time/cost to projects):

• Lower pricing for customers - shorter sales cycles, delivers quick wins, clear upgrade pathway to full engagements

• Faster learning for Artanis - allows us to focus solely on AI (and not software engineering) to maximise learnings on our tool/methodology

If this appeals to you or someone you know then get in touch! 

🏹 Goal for August 🏹
Grow method customers from 2 to 3. Sign one lightning project customer.

🙏 Shout-outs 🙏
Special thanks for July go to:

Henry M - for the intro to Katie
JPH - for the coaching perspective and intro
Srecko D - you know why
Adam D - for connecting us with a great opportunity
Oliver K - for the intro to Peter
Duncan C - for plugging our Substack!

En Taro Tassadar,
Artanis Team

🙋 Ways you can help - operational directors 🙋
We’re doing another round of market research. We’d like to meet with COOs/Ops Directors at scaling businesses who’re tasked with resolving operational bottlenecks. Please get in touch if you know anyone who fits this profile!

🔬 Focusing on Methodology 🔬
June was the first month with the full-time team in place, so we had our first offsite!

The main outcome was defining our “product” as our methodology for delivering AI projects, rather than just software. We currently spend a lot of our time ensuring customer success by being forward-deployed. We’ve already identified one area that can be productised - labelling - but there are also other points in the development process to productise. 

Continuing to be forward-deployed enables us to learn as much as we can about what matters: building a repeatable methodology for delivering AI projects - and with each project, we’re able to deliver that more effectively.

The market norm is still that money goes into flashy AI demos that make it into a press release, but not into production. We have lots of experience getting AI projects into production, and this methodology is a major market need. Going forward, we’ll put more time into documenting our methodology internally and publishing more content externally.

📉 Progress in June - evening out 📈
Our metrics from June are…

Method customers: 1 (-)
Total customers: 5 (-)
Monthly Revenue: £76.1k (+£22.9k)
Team Size (proxy for cost): 5 (-)

We changed our primary metric during the offsite, when we decided “SaaS” customers wasn’t the right focus. Instead, we wanted to focus on customers who are following our methodology to monitor their AI in production. We now define “Method customers” as those with AI deployed in production and following our methodology to maintain it.

This is quite a harsh metric. At this stage, we’d rather err on the side of being too harsh with our metrics, as it’s important to get the fundamentals right before scaling.

Getting through some pilots has still been slower than we’d hoped. However, we made good progress on one that we’re optimistic will become a great case study for us.

🏹 Goal for July 🏹
We’re rolling over our goal from June, to grow from 1 to 2 “Method Customers”. We feel very close with a couple of pilots, but are aware it’s now two months running, having missed a similar target. We’re going to give it another month before re-evaluating whether we’re doing something wrong.

🙏 Shout-outs 🙏
Special thanks for June go to:

Laura R - for pulling together a great offsite
Zach / Aash / Maurice - for providing website feedback
Cait C - for the intro, or at least the attempt!
Dom S - for being a class act
Rawan A and Adam D - for the intros
Alamin S - for hosting our lightning talk
Hannah G + Kassi R - for the heroic data labelling efforts
Andrew M - for bringing great energy in your first month

En Taro Tassadar,
Artanis Team

🙋 Ways you can help - feedback on our new website! 🙋
We’ve made a big change to our website at https://artanis.ai. Primarily, it now features more information about who we are and what we do, with fewer Starcraft references. We expect visitors to have some referral already, rather than be cold, so it’s designed for info rather than conversion. We’d appreciate any reviewers!

📉 Progress in May - a bit slower 📈
Our metrics from May are…

SaaS customers: 1 (unchanged)
Total customers: 5 (unchanged)
Monthly Revenue: £53.2k (-£18.8k)
Team Size (proxy for cost): 5 (unchanged)

We didn’t hit our growth target in May for two reasons:

1. We’ve been slower to get through the pilot phase with a couple of customers than we’d hoped. This delays revenue, as we report invoiced rather than contracted revenue.

2. At our stage, we’re still in “figuring it out” mode, particularly regarding customer profile. Our main strategic goal is to find a scalable business model, rather than grow revenue. While in this phase, it’s important to remain lean & robust to changes. We’ve written about how the pre-PMF startup should not be a rocketship, but a snail.

We’re no longer sure that getting customers onto our SaaS platform is the right metric. We don’t want to do this at the expense of delivering value to customers from AI. At this stage, it’s more important to develop our methodology. Methodology isn’t just technical and we have more to learn about other aspects, like stakeholder alignment.

For example: rushing a customer onto our platform before we have access to their data wouldn’t add any value. Instead, it’s better to focus on getting access to their data first. We’d taken this as a given when working with startups, but it’s harder when working with bigger companies.

🔬 Two Types of Customers 🔬
After some customer discovery work, we’ve found two broad profiles of businesses wanting to do AI projects:

1. “Product AI”. These are usually startups whose entire business is a new AI product. Getting their AI to work is a top-of-mind problem, because they don’t have a business otherwise! They’re usually venture-backed and don’t yet have product-market fit.

2. “Operational AI”. These are established businesses wanting AI to improve margins on existing operations. Getting their AI to work is a less pressing problem, as it’s not core to their business. However, they have more stable processes and business models which make for better long-term customers.

Due to our network, most of our customers so far have fallen into the “Product AI” bucket. They enabled us to develop our methodology quickly and were highly engaged. However, we’re starting to work on more “Operational AI” projects, and it could be a big longer-term opportunity.

🤔 Challenges - Enterprise Software Advice 🤔
We’re starting to work with some bigger customers, rather than just startups. While we’re strong on the technical side, we’re finding stakeholder alignment more challenging. It would be great to hear from other founders, or forward-deployed engineers, who’ve integrated new software into enterprise environments for advice.

🏹 Goal for June 🏹
Our current goal is to get another “Operational AI” customer into production. We have one particularly promising customer that’s still in pilot. If we succeed, this will take us from 1 to 2 on Operational AI customers using our methodology.

We’re also doing an offsite in June, which may lead to a change in strategy/metrics. It’ll be our first offsite with a full team, so everything is on the table for that!

🙏 Shout-outs 🙏
Special thanks for May go to:

Dhruv S - for your transparency
Srecko - you know why
Geordie - for the referrals
Anna W - responding to our last call for help
Jonny, Chris & Anna - for being outstanding humans
Lloyd H - for the surprise gift!

En Taro Tassadar,
Artanis Team

🙋 Ways You Can Help - Advice on Consulting → SaaS 🙋
We’re starting to move from a pure services business to SaaS. This has thrown up some challenges, both strategic and operational. Connecting with founders who’ve transitioned from services to SaaS for advice calls would be helpful.

📉 Progress in April - first SaaS customer 📈
Our metrics from April are…

SaaS customers: 1 (+1)
Total customers: 5 (-2)
Monthly Revenue: £72k (+£10.5k)
Team Size (proxy for cost): 5 (+1)

Our main metric has changed to customers on our SaaS platform. We went from 0 to 1 in April, with a first customer now using Artanis to write Policy and label their data! As a reminder, here’s our post on how Policy is the missing core of the AI stack. We still deliver the rest of the AI stack as a service, but it’s much smoother when customers can label their data without us.

The strong revenue numbers (technically a $1m run rate!) include wraparound service. We started with an enterprise customer that requires more integration work, so we needed to expand our capacity with a contractor to support this. It’s not clear yet whether these revenues are recurring, although it’s a good sign!

We lost a couple of customers in April, one due to their financial situation, and the other due to changes in their roadmap. We couldn't see a path for them becoming SaaS customers in the near term, so we’re not concerned, and it has enabled us to narrow our focus.

🔬 New posts in How to Build AI that Actually Works 🔬
We’ve written new posts on:

1. How to build AI when the ground truth is subjective. Labelling data is often the hardest part of creating an AI product. The solution to this is treating both Policy and labelling as an iterative process called the Policy-Data Loop.

2. Don’t write prompts. Write Policies. Prompt engineering is arguably a workaround for the current shortcomings of foundation models. As they improve, prompts will become less important. Instead, the enduring IP of your AI will come from the Policies that you write.

🤔 Challenges - aligning Sprint planning with customers 🤔
We’ve found it hard to align our Sprint planning with customer Sprints. We try to do our weekly planning first thing on Monday. However, this is out of sync with customer Sprints, which can often be later in the week. It would be helpful to hear from anyone who’s faced a similar challenge!

🏹 Goal for May - continue transition to SaaS 🏹
Our main goal is to get more customers onto our AI platform. In May, we’d like to go from 1 to 2. Pretty simple in theory!

In practice, not all of our current customers will be a good fit for this, as a lack of high-quality labelled data needs to be block their AI projects. Some are still early in the development cycle, or other problems are more pressing.

🙏 Shout-outs 🙏
Special thanks for April go to:

Christine F - for the opportunity to speak at Experian
Lloyd H - for the intro to Dimitris
Greta A - for the invite to teach the new Balderton Launched cohort about AI
Shabana - for helping out at short notice
Stephen W - for helpful perspective over lunch
Dom/Hannah/Tom - for being the ideal development partner on the new platform!
Srecko D - for reminding us to be patient
Ibraheem R and Alamin S - for hosting and inviting us to speak about founder life
Oliver W and Mamal A for swapping war stories
Benji F for advice on product tooling

En Taro Tassadar,
Artanis Team

🙋 Ways You Can Help - Subscribe to our Substack 🙋
We've been posting about AI and startups over the past year on Substack. This has recently picked up steam and generated inbound from customers/investors. We're therefore doubling down on our Substack, so please subscribe to follow our AI and startup content here.

👩‍💼 A Wild Founding Team Appears! 🧑‍🔬
Pleased to announce that we filled out our founding team in March:

Laura Rosenberger became COO of Naked Wines, a public company doing ~£100m ARR, by age 28. She then founded and ran a venture-backed wine business, Laylo, for four years. She recently decided to transition into AI and will lead GTM/strategy/ops at Artanis. Lastly, she has an MPhys in Physics from Oxford!

Olly Styles has an AI PhD from the University of Warwick. He then co-founded Atlas AI, so longtime readers of these updates may be very familiar with his story already! Following the demise of Atlas, he joined a large edtech scaleup as their first ML hire. We’re very glad he’s decided to have another crack at a startup.

Andrew Manderson has an AI PhD from Cambridge. He spent several years at the Alan Turing Institute, where he met Sam. After his PhD, he led data science at a climate-tech scaleup. He quickly became “business critical”, which is unheard of for a data scientist, so we’ll be waiting until June for him to join.

The team now includes four AI PhDs AND someone who knows how to run a business. We’re unlikely to hire more in the next few months, and feel very lucky with how this has happened so quickly!

📉 Progress in March - capacity improving 📈
Our metrics from March are:

North Star: 7 customers (+1)
Monthly Revenue: £61.5k (+£9k)
Team Size (proxy for cost): 4 (+2)

We’ve been able to onboard new customers, due to recent hiring. We’re now at all-time highs for both customers and revenue, which is a great feeling. We now need to manage the transition from consulting to SaaS, while still maintaining a premium service.

🏹 Goal for April - start transition to SaaS 🏹
The problem we’ve unearthed from our consulting is people don’t get the outputs they want from their AI because they’re inconsistent with how they define what they want. We’ve solved this by making them define their Policy, which is the missing core of the AI stack.

We’re now building a platform where customers write Policy (i.e. what they want their AI to do) and label their data. We then provide model development and deployment as a wrap-around service. This is a much more scalable model than pure services, which involved labelling customer data ourselves. However, there’s risk around whether we can build a platform that meets this goal.

Our next major milestone is one customer consistently using our platform to write Policy/label data successfully, which means without our intervention. Therefore, customers on our SaaS platform (currently 0) is our new primary metric. We will continue to track total customers, revenue & team size.

🙏 Shout-outs 🙏
Special thanks for March go to:

Ed S - for saying nice things about us
Greta A - for the many opportunities to speak at Balderton
Sivesh - for believing in us so early
Fergus B - for giving first
Brad H - for critical short-notice legal help
Kari & Rachid - for being understanding
Laura & Srecko - for teaming up against Sam on legal risk
Vini C - for a very helpful reference and a fun chat
Sam C - for sage advice about the limits of transparency
Hannah G - for an intense-but-hopefully-productive afternoon

En Taro Tassadar,
Artanis Team

🙋 Ways You Can Help - Subscribe to our Substack 🙋
We've been posting about AI and startups over the past year on Substack. We enjoy writing these posts, even if few people read them! Recently, we got the first tangible evidence that this makes business sense too with a reader greatly accelerating our first raise. We're now doubling down on our Substack, so please subscribe to follow our AI and startup content here.

🔫 We've pulled the trigger and raised our first VC round! 🔫
We're thrilled to announce we've raised our first VC round! The penny dropped for us in January about how to scale and we decided it was time to pull the trigger on raising. We thought we were in for a long slog, but luckily the lead investor (Lucid Capital - more on them below) moved very quickly. They invested ~£1m and left ~£250k space for strategic follow-on investors, which is now mostly filled.

This will fund a 6-person founding team for 18 months. The goal is to become cash-generative AND hit at least seed metrics, putting us in a strong position to raise again. The main risk is whether we can transition from consulting to SaaS, which we'll focus on fully once the founding team is in place.

🤝 Why we picked Lucid Capital 🤝
We met Srecko for a random intro call before we decided to fundraise. Srecko read all 25 posts on our Substack, then debated us about one of them - how dare he!! Anyway, we quickly cooled off, then realised we were strongly aligned on both AI and startups. We also just got the vibe that he was a brutally honest person that we could work with very transparently.

We'd actually never heard of Lucid - their fund only launched in 2024. However, Srecko made such a strong impression that we called him immediately after we decided to raise, and they ran a very quick process. They both showed flexibility on things that mattered to us, but challenged heavily at the right times. For example: Srecko called me after their investment committee to tell me it was "not a good call" (exact wording redacted) and we needed to make it right.

All three Lucid partners are ex-operators and sharp, straightforward people. They also understand AI to a degree well beyond the vast majority of VCs - we all agree that there's a bubble in AI. Interestingly, they wrote posts pointing toward our thesis on their Substack before we even started Artanis. We're thrilled they're backing us and very excited for the road ahead!

📉 Progress in February - demand is still growing, but we're at capacity 📈
Our consulting metrics from February are:

Primary metric: 6 customers (+2)
-> of which 2 on paid waitlist
Secondary metric - revenue: £52.5k (+£12.5k)
Secondary metric - team size: 2 (unchanged)

We're still turning down work—our "army of 2" has its limits—but we have started a paid waitlist where new customers can pay a deposit to guarantee a certain start date. However, we've made good progress on hiring recently and hope to resolve the capacity issue very soon. Fingers crossed for some big hiring announcements in March!

🙏 Shout-outs 🙏
Special thanks for February go to:

Akash - for playing it straight
Chris W - for going into bat for us
Ed & Ryo - for believing super early
Laura / Andrew / Olly - you know why
Henry M - for continuing to earn superstar ally status
Sivesh S - for accelerating us so much
Greta A - for trusting us to teach the next Balderton
cohort about AI!
Srecko - for everything, but particularly sparring!
Zoe / Jamesin / Will - for the invite to speak at a great event!
Irfan - for the serendipitous intro to Laura
Alex K - for saying nice things about us to Andrew
Pascual - for the DD support
Sam C - rapid intros to Paul & Chris
Harry S - for letting Jamie observe
Jamie F - for a very fun afternoon!

En Taro Tassadar,
Yousef & Sam

Balderton Capital is running a demo day for our incubator cohort in Kings Cross. This is good timing, as we've built a ton of conviction recently in our vision and will pitch Artanis to the public for the first time! Sign up here if you'd like to attend demo day on the evening of Tuesday 4 March.

🧐 Special Edition - Big Decision Coming Up 🧐

We've got a big decision coming up. We need to grow the team, as we've had to turn down several customers recently. We've also solidified our thesis on the AI space (see below) and have a product vision we'd like to start building. We'll take a call soon on whether to keep bootstrapping, or fundraise to build our vision faster. To VC, or not to VC...?

🎯 We've found the missing core of the AI stack - Policy 🎯

Our thesis finally clicked for us in January. The reason people can't trust their AI products, and only human-in-the-loop use cases are being deployed, is they can't control the output. And they can't control the output because they've been ambiguous in the instructions they give the AI. We've solved this problem with Policy: The Missing Core of the AI stack. For more info, read our 5-minute thesis on Substack.

📉 Progress in January - demand growing, team turbulence 📈

Our  consulting metrics from January are:

Primary metric: 4 customers (unchanged)
Secondary metric - revenue: £40k (-£10k)
Secondary metric - team size: 2 (-1)

The metrics above are slightly misleading because we've been turning down work recently. This has been a mixture of i) inbound leads and ii) existing customers wanting bigger contracts. As we're a new business, we've had a hard line around needing to prioritise quality over quantity. But we're optimistic about capacity going forward, given recent progress on our thesis and product vision.

We also decided that Jerome is best placed as an advisor, rather than a co-founder, as we wanted to build different businesses. We're still retaining him, as he's accelerated us a lot already, but it means Yousef and Sam are back to being an army of 2! And if you need some extra technical/commercial firepower for your startup, reach out to jeromeminney@gmail.com.

🙏 Shout-outs 🙏

Special thanks for January go to:

Sivesh S - for teaching us the dark arts
Christine F - for the great connection to Jonny
Harry D - for an inspiring offsite
Zahid M - for fundraising advice
Bertie V - for being a "good sport" and boosting our vision
Laura R - for the intro to Ben S
Jonny DMM - for the exciting new project (+ "base of data" anecdote)
Rod F - for support behind the scenes
Al R & Marcus S - for being serial founder role models 
Sophia, Ben, Dom, and Haseeb - for the extremely kind words about us
Ed S & Ryo H - for first ticket enthusiasm
Jerome M - for handling it all very well
Callum NF - for an exciting opportunity
Srecko D - you know why
Henry M - for being an intro superstar
Pascual & Jeppe -  for playing it straight
Imran R - for organising our PyData talk

Sorry if we missed anyone off the list, January was a biggie!

Cheers,
Yousef & Sam

🙋‍♂️ Ways you can help: people building AI products 🙋‍♂️

We're looking to speak with people at companies meeting the following criteria:

1. An LLM is a core component in their main product

2. Their product is already in the market and has some traction

3. They have a modelling team of at least 5 people and one of their team has a background in AI/ML rather than software engineering

Please get in touch if you know anyone meeting these criteria.

📉 Progress in December - learning about the labelling market 📈

In December, we started discovery conversations among companies selling data labelling services. We met our initial target of confirming similar pain points across at least three companies in the segment. We booked eight calls and have written about the conclusions in more detail below. We’ve also proceeded to demo with two prospects.

Pain points were mostly in line with our priors. However, 3 out of 4 companies said they wouldn't buy third-party labelling tooling because it's too close to their core IP. We've therefore decided to alter our focus to speaking with companies building AI products and who are labelling data in-house, rather than selling labelled data.

Separately, our consulting metrics were:

Primary metric: 4 customers (+1)
Secondary metric - revenue: £50.0k (+£14.5k)
Secondary metric - team size: 3 (this is a proxy for cost)

We’re very proud to hit £50k monthly revenue within 7 months of starting Artanis 🎉 While not our primary metric, it’s proven we can go-to-market in the B2B AI space. (And it pays the bills, of course)

🔬 What we learned about the labelling market 🔬

We learned some super interesting stuff about the cutting-edge players in data labelling:

1. Most of their business is now labelling difficult tasks (e.g. answering complex financial questions) rather than simple ones (e.g. is this image a cat or a dog). Volumes have shrunk, but they can pay up to $1,000 per labelled query!

2. Mislabelling is very costly, as it leads to models being trained on bad data, and they're spending heavily on QA to reduce labelling mistakes.

3. They draft very detailed labelling guidelines to reduce ambiguity and errors. However, it's hard for their labellers to remember all the guidelines, so they still make mistakes. The tradeoff between the length of guidelines and ambiguity is an open problem.

4. Sourcing labellers with the right domain expertise for these difficult tasks is very expensive. One company told us they needed to source 200 lawyers within a fortnight!

🧐 Challenges - competing business ideas 🧐

We have multiple compelling go-to-market strategies and we’re finding it hard to choose the best one. We’re keen to hear from others who’ve been in founding teams with competing ideas - how do you think co-founders should resolve strategy disputes?

🏹 Goal for January - validate at least one customer segment 🏹

This is pretty similar to last month's goal. We want to confirm pain points across at least 3 companies building ML models that are labelling data in-house, and that they're willing to buy third-party labelling tooling. If we can do this, we'll progress from discovery to sales.

📣 Shout-outs 📣

Thanks for December go to:

Bertie V - our intro MVP for 2024!
Sivesh S - spreading the gospel about us
Anthony F - for connecting us to Josh
Nandu A - for sending curious investors our way
Karun S - for the chat about labelling data at Arsenal FC!

Cheers,
Yousef, Jerome & Sam

🔬 Artanis: the story so far 🔬

For new readers, we've started publishing all previous updates online. If this is hitting you fresh and you'd like more context, go to https://artanis.ai.

📣 Big news: we've found a third cofounder! 📣 

Very pleased to announce that Jerome Minney has joined as a third cofounder! We first met Jerome at a house party in October, where he mentioned his idea for a product to help people label data where domain expertise was needed. This aligned very strongly with our own experience of the biggest problem in the applied AI market, so after a few weeks of discussions we decided to start working together.

Jerome brings complementary expertise in growth (i.e. he's much better at talking to customers than Yousef or myself). He's worked as a growth consultant for ~10 startups, and ran his own software consultancy for 5 years before that. He's already accelerated our go-to-market and we're excited to see how the next few months shake out!

🙋 Ways you can help: people working in data labelling 🙋

We're looking to speak with people who are trying to improve AI model quality. We'd love to connect with anyone fitting the following profiles:

1. Working in an internal modelling team at a company building in AI
2. Working in an internal data / evals team at a company building in AI
3. Working in a team buying labelled data from other companies
4. Working in a team selling labelled data to other companies

Please get in touch if you know anyone fitting those profiles.

📉 Progress in November - steadying the ship 📈

Our goals were to i) retain 3 service customers and ii) come up with a new plan during our offsite. We feel good about progress on both fronts.

Primary metric: 3 customers (-1)
Secondary metric: £35.5k monthly revenue (+£8.5k)
Secondary metric: 2.25 team size (Jerome was here for 1/4 of the month)
-> This is replacing delivery hours as our "cost metric", so higher is worse

Yousef and I had a romantic few days at our offsite in Hampshire. We felt strongly aligned that labelling data accurately was the biggest problem across our past AI projects and that we'd like to work toward a scalable solution for it. We came up with some ideas, ranging from building an internal tool for our consulting clients to selling a standalone labelling product. We're looking forward to working through these (and other) ideas with Jerome.

🎯 Goal for December - validate at least one customer segment. 🎯

Now that our team is 50% larger, we've got more capacity to work on longer-term projects. Right now, our best guess is in the labelling tooling space. Our first target for validation is going to be:

1. Similar pain points confirmed across at least three customers in the segment.

If we can achieve that, we'll aim for a first signed customer the following month. We'll continue to report consulting metrics too, but are not aiming for growth there.

🕵️ Challenges - relationships in three-person founding teams 🕵️

There's a ton of upside potential from bringing in a third co-founder. However, we're aware there are risks too and the relational dynamics are more complicated than a two-person team. If you've been part of a three-person founding team before, we'd love to hear from you about your experience.

🙏 Shout-outs 🙏

Thanks this month go to:

Sophia - for a glowing testimonial
Haseeb - for an interesting new project requiring domain expertise
Ibrahim - for the invite to speak at the Buildathon
Henry & Tom H - for help with getting out the spam naughty corner
Haz H & Tom H - for hosting the party we met Jerome
Natalie, Rod, Joachim, Mo, Ryo, Ed S - for helpful offsite advice
Sam C - for telling it how it is
Caitlin C - for the nicest surprise call of the month
Anon ex-client - you know why!

Jerome, Yousef & Sam

🙋 Ways you can help: tips for running an offsite 🙋

We're doing our first offsite this week, to have a reset and review our longer-term vision for Artanis. There's only two of us, so it'll be very romantic. Anyway, let us know if you've got tips from past experiences with offsites/awaydays.

📉 Progress in October -  📈

In October, we put our nascent product on ice and set a goal to retain 3 service customers.

Primary metric: 4 customers (unchanged since September)
Monthly revenue: £27k (+£15k)
Monthly cost: 150 hours building AI for clients (+23)

We also set aside time to reflect on the six months since we started. We had an initial goal of getting to 10 customers as a service before making a serious effort to productise. We assumed that customers would each start with an initial high-intensity build phase, followed by lower-intensity maintenance deals after that. This would have enabled getting to 10 on a small team, as most customers would be on maintenance contracts.

Things didn't work out this way. Customers who finished the pilot wanted to continue at the same intensity, usually to build new models or features around the existing model. This was good for revenue, but we capped out more quickly than expected on capacity. By July, servicing 4 customers was close to a full-time job, then August tipped us over the edge. Like this one, we're going to write up some more detailed learnings in November.

🎯 Goal for November - retain 3 service customers and come up with a new plan 🎯

We're wrestling with the right vision for Artanis. When we first started the company, we were pretty set on working toward something scalable. However, we've been enjoying the services over the past six months and it's been nice to have a tangible impact this quickly. We also haven't come up with many compelling product ideas yet - we feel we still learn something from each new project.

We've set aside some time for an offsite to come up with a revised vision. That's the main goal for next month, rather than a growth target. If we can end November with 3 happy service customers and a new plan that we've got conviction in, that'll be a good month.

🔬 How to build AI that actually works: final post 🔬

We've written the final post in our playbook for building AI that works reliably. This post covers how to monitor your model after launching, cautioning that your data will "drift" over time away from the data you initially fit the model with. The particular topics we cover are:

1. Needing to keep labelling a sample of data after launching

2. Smartly tilting that sample towards likely failure cases, but...

3. NOT being too smart for your own good with fully automated monitoring

4. NOT chasing the holy grail of having users monitor for you

This series of posts provides a high-level playbook for those without a deep AI background to build better systems. It's been fun to write, so we're going to keep writing posts on related topics. Do let us know if there are any AI issues you'd like us to write about!

🕵️ Challenges - the risks of "building in public" 🕵️

After we published the last update, we got a call from a potential customer who we'd progressed to drafting a contract with. They were worried about our uncertainty on strategy - they'd previously been burned by a company who built software for them, and then didn't provide longer-term support after changing direction. We had an open conversation, and ended up putting the deal on hold.

This has made us think about the way we write these updates. We've come down on the side of continuing to be transparent. Starting a business is messy and it's more fun to write openly about that. Our take is that telling the truth won't always get us what we want short term, but usually works out best in the long run. (although John Maynard Keynes may disagree...)

🙏 Shout-outs 🙏

Special thanks to the following for helping out in September.

Antoine N - for the intro to Matthew
Ashley HL - for the "firm" advice to prioritise an offsite
Greta A - for putting on a great B2B sales session at Balderton
Benji F - for boosting us to your network
Mehdi B - for the referral
Lorenzo S - for a very similar referral!

Thanks,
Yousef & Sam

🙋 Ways you can help: potential 3rd co-founders 🙋

We're considering taking a 3rd co-founder to add capacity and fill in some gaps. We're looking for people who:

1. Actively want to found and know what the role entails e.g. because they've been in an early-stage startup before

2. Have a strong background in data science and/or early-stage sales

3. Want to work in-person most of the time

Please get in touch if you know anyone who meets these criteria.

🔬 How to build AI that actually works: new post 🔬

We've written the third post in our playbook for building AI that works reliably. This post covers how to fit models in the real world, where there are added complexities you don't experience during university projects or on Kaggle. In particular, we dive into:

1. Defining what "good enough" looks like (the market decides)

2. Using a basic model to check data is correctly labelled (it's often not)

3. Iterating, in a methodical way, towards good enough (logging your experiments)

📉 Progress in September - first month in the red 📉

Our goal for September was to retain 1 customer for our new SaaS product. Unfortunately, we didn't manage it.

Primary metric: 4 customers (-2 since August)
-> of which SaaS: 0 (-1)

Monthly revenue (secondary): £12k (-£10k, some late invoices)
-> of which SaaS: -£1k (-£2k, includes a £1k refund)

Monthly cost (secondary): 127 hours building AI for clients (-26)
-> this doesn't include time spent on SaaS

We were squeezed for time, given our existing commitments to services customers, and dedicated <25% of our time to the product. We also made no time for sales. Ultimately, we didn't deliver and we refunded our first SaaS customer. We're now thinking carefully about what to do next.

In retrospect, we made a big mistake in our strategy. We thought launching a SaaS product would provide a more scalable growth path. While this is true in the long run, the 0 -> 1 stage felt like launching a second business. In hindsight, it clearly is launching a new business!

🕵️ Challenges - flip-flopping on strategy 🕵️

We're having a bit of an identity crisis between providing services and building a product. We enjoy the services work and it's clearly a viable business. But we also haven't found a clear path to scaling it, which is our ultimate mission. This uncertainty makes it hard to set sensible targets, and has led to mistakes such as:

1) Trying to grow a services business at the rate of a product startup, then...
2) Launching a product while trying to maintain services, as a team of two

Normal for an early-stage startup? Or headless chicken? Perhaps they go together more often than we think...

On the plus side, we started a 6-month incubator with Balderton Capital. It's been nice sharing a space with a bunch of other early-stage AI startups, some of whom may be reading this now!

🎯 Goal for October - retain 3 service customers and reflect 🎯

October's goal: retain 3 service customers

We've decided to deprioritise product work for now. We got our fingers burned recently and want to ensure delivery for current customers, who we're grateful to for deepening ties with us. We've done quite a lot since starting Artanis and the end of October will be our 6-month anniversary, so it feels like a good time to reflect on some themes we've observed and chart a new path forward.

🙏 Shout-outs 🙏

Special thanks to the following for helping out in September.

Ben L - for helping scope an interesting project
Maurice B - for giving our writing a shout-out
Nicky F - for some sage advice on not knowing
Mo N - for the intro to Alex
Aniruddha R - for the fun intro to Arsenal FC!
Ben C - for your patience

Thanks,
Sam & Yousef

🔬 New SaaS product  - explainer video 🔬

We've put together a 3-minute explainer video for our new SaaS product. It's a bit rough, but hopefully gets the point across. Let us know what you think!

🙋 Ways you can help: connecting with non-technical founders 🙋

We're looking to meet more people that match the following profile:

1. Non-technical founder building an AI product using their domain expertise e.g. in health, education, finance

2. Their startup is small, 1-5 people

3. They've raised little or no funding, so hiring more engineers is a big risk

Please get in touch if you know anyone who may meet these criteria

📉 Progress in July - first SaaS customer 📈

We signed up customer #1 for our SaaS product! They've asked to remain anonymous, so we'll refer to them as "#1". 

#1 has commercial experience in the football sector. They've found an opportunity to use AI to automatically highlight clips from videos of lower-league games. However, they haven't raised funding and don't have a technical co-founder. They tried using ChatGPT and other no-code tools, but none could analyse video accurately. After much trying, they felt out of options to build their AI. When we told them Artanis would be just £1k/month, they were keen to sign up so we launched early for them!

We should also report on our usual metrics. It's getting messy between service work and the nascent SaaS product, so we need to think about how to report going forward...

Primary metric: 6 customers (+1 since July)
-> of which SaaS: 1 customer

Monthly revenue (secondary): £22k (+£1.5k)
-> of which SaaS: £1k

Monthly cost (secondary): 163 hours building AI for clients (+27)
-> less relevant for SaaS

🕵️ Challenges - we're overstretched 🕵️

August was an intense month. We delivered for 5 service customers, alongside building for our first SaaS customer. We needed to work late shifts/weekends to ensure we didn't drop the ball, which didn't feel sustainable. We got several warnings that we wouldn't be able to reach 10 customers, on a team of 2, while going to market as a service. For those who tried to warn us: we were wrong and you were right!

Relatedly, another challenge was losing a customer for the first time. It wasn't dramatic - they just repeatedly didn't pay their first invoice in time. We decided to cancel the project which wasn't nice, but made a bit easier by feeling overstretched.

🎯 Goal for September - stay at 1 SaaS customer 🎯

Hang on, aren't startups meant to grow? Aren't we meant to be a rocketship?!

We take burnout risk seriously. Startups are a long game, so sustainable working is important and we don't want a repeat of August. We're going to take our foot off the gas in September, and prioritise doing a good job for existing customers.

We have two routes for staying at 1 SaaS customer. We can either retain #1 by making the product work well for them, or sign up a second customer. We'd grow if we managed to do both, but don't want to hold ourselves to that.

🙏 Shout-outs 🙏

Special thanks again to everyone below.

Sivesh S - for taking us onto the Balderton program
Rod F - a very overdue shout-out for several intros!
Mounir M & Emma B - for a very interesting technical project
Andy R & Matt C - for scouting a nice deal for us this month
Adam D - for the intro to Jason, despite some blunt feedback from us...
Alex H & Kieran - for warning us, it eventually got through!

Thanks,
Sam & Yousef

🙋 Ways you can help: connecting with a new customer profile 🙋

We have a new ideal customer profile (ICP):

1. Non-technical founder building an AI product in a domain where they have expertise

2. Their startup is small, 1-5 people

3. They've raised little or no funding, so hiring engineers is a big risk

Let us know if you can connect us with anyone who may fit the new criteria.

📉 Progress in July - getting to 5 customers 📈

A reminder that our focus is market risk: will startups pay us to build their AI? Our milestone for de-risking this is 10 customers.

Our goal for July was to grow from 4 to 5 customers. A bright spot is that our second customer signed a 6-month extension following the pilot. We also hit our target, but we let our ICP fray by taking a customer whose first project doesn't involve a heavy AI component.

Primary metric: 5 customers (+1 since June)
Monthly revenue (secondary): £20.5k (-£2.5k since June)
Monthly cost (secondary): 136 hours building AI for clients (+15 since June)

Our "cost" metric is a concern. We're a team of 2, and we learned we're not going to be able to serve 10 customers while remaining purely a service. It's just not scalable enough. We don't want to hire, as our mission isn't to scale a consultancy, so we're going to be changing course in August!

💡 Critical insight - our users are domain experts, not engineers 💡

In all our projects, the main stakeholder has been the domain expert (i.e. the person closest to the data) rather than the CTO. For example: when building an AI tutor for marking English essays, our main stakeholder is the human tutor currently marking those essays. Our value is encoding the knowledge of a domain expert into an AI system that reliably does their job.

We've learned that domain experts with the right tech are more likely to build good AI systems than engineers. Therefore, the user of Artanis should be the domain expert. It's easier to launch new products when the customer is the user, so we're changing our ICP to align with this.

🕵️ Challenges - changing course 🕵️

We're making two main changes in response to issues with our ICP and scalability:

ICP change: our old ICP was the founder of an AI startup that has software engineers but no AI specialists. This has been a tough sell: they often see the quality of their tech as their IP, so want to build it with an internal team. However, non-technical founders often see their domain expertise as their IP, rather than their tech. This is a much more natural fit for us.

Product Launch: we'll struggle to reach 10 customers operating purely as a service, due to lack of scalability, so we're launching a SaaS tool. Artanis will help non-technical founders with domain expertise build AI products without needing to hire engineers. Our SaaS product will perform three main functions for the domain expert:

1. Encode their domain knowledge into an AI product that works reliably. 

2. Deploy and host the AI for them so that other people can use their product.

3. Allow them to update their product in response to requests from their users.

You may question why we're changing tack when the services business has started well. Sometimes we too wonder "Is this just self-sabotage?" Perhaps entrepreneurship often requires a degree of masochism...

🎯 Goal for August - 1 customer for our new SaaS product 🎯

We're changing our goals to focus on the new product. The main goal will be to sign up 1 customer for that, rather than a custom project. If we do that, and avoid churn, we'll grow from 5 to 6 customers. To avoid confusion, we'll separate our SaaS metrics from our consulting work.

🔬 How to build AI that actually works: new chapter🔬

We've written a second post in our playbook on building AI, covering the need for performance evaluation. A common mistake is building a model, seeing it works "most of the time", then launching. When it's time to iterate, people don't know if they're making things better or worse and end up feeling like they're playing whack-a-mole. 

A better approach starts with defining "success" by building a labelled dataset, before spending any time building a model. You can't reliably improve your model unless you put the time into building this dataset. We acknowledge that defining success can be tricky with language model outputs, so we provide an example of how to do that too.

🙏 Shout-outs 🙏

Big thanks to the following for responding to our previous call for help!

Shingai A - for a fun chat about crazy stuff we've seen in AI startups
Christine F - for the intro to Oli
Matt C - for a few good referrals off the back of a fairly random meeting
Walid BM - for giving us a strong customer reference
Myles G - for the links into Elevate
Ryo & Olly - for helping calm our nerves about self-sabotage!

Thanks,
Sam & Yousef

🙋 Ways you can help: save us from doing cold outreach! 🙋

First, a huge thanks to everyone who responded to last month's call on intros (see bottom of this email for shout-outs). We're still looking to schedule more calls to learn about how startups/SMEs approach AI projects, with companies roughly meeting the following profile:

• They're already building AI or ML-based features

• They haven't got an AI or ML PhD on the founding team

• They have some funding or revenue already

Please let us know if you can connect us to companies meeting some of these criteria.

📉 Progress in June - getting to 4 customers 📈

First a reminder that our current focus is market risk: will startups pay for AI / ML without hiring an in-house team? Our next major milestone for de-risking this is 10 paying customers. 

Our goal for June was to grow to 4 paying customers. By some miraculous luck, rather than things going to plan, we just about scraped our way there! This may be the only time in Artanis history that we meet our target two months in a row...

Primary metric: 4 paying customers (+2 since May)
Monthly revenue (secondary): £23k (+£9k since May)
Monthly cost (secondary): 121 hours building AI for clients (+20 since May)

Why was this lucky? Our go-to-market plan A hasn't yet yielded any customers. Instead, Sam had a week where he took about 40 calls, none of which led to a deal. However, a new tenant in our workspace overheard and approached to say they've just raised, they're looking to build out their AI, and it sounds like we're obsessed with this. One thing led to another, and we had a deal!

🕵️ Challenges - mixed signals on go-to-market 🕵️

Our go-to-market plan A is taking longer than we'd hoped. We spoke with ~60 "allies", hoping they would connect us with potential customers. These initial calls yielded about 10 relevant leads. Some of these may convert to new customers in July, but we've not yet seen a concrete success story.

Some startups hiring their first AI engineer have a hard red-line that this must be a permanent employee, as they may build a team around this first hire. Building a team seems more important to them than getting their AI built quickly and reliably. They'd rather spend months struggling to find the right person than consider alternatives, such as us!

Our first customer has signed a 6-month extension, which is great news! However, our prior was they'd mostly want us to maintain/improve existing functionality. Instead, their roadmap involves more intense building than expected. We're being fairly compensated, but if other customers want similarly intense long-term deals then we will need to rethink whether we can support 10 customers as a 2-person team.

🎯 Goal for July - grow to 5 customers 🎯

July's goal is growing from 4 to 5 customers. We're hoping to retain our 4 current customers and acquire at least 1 new customer.

We'll keep going with our "allies" GTM for another month. While it's not yet led to a new customer, it's showing promise with several fairly well-qualified leads. We're also having fun talking with people we know about Artanis and, to be honest, we're more likely to succeed with a GTM strategy that we enjoy. We therefore want to fully rule it out before moving onto a more cold outbound strategy.

🔬 How to build AI that actually works: our new playbook 🔬

We've started writing a playbook on building AI that works accurately and reliably. The first post covers how to break a high-level problem down into smaller steps. Most of these steps should be solved with reliable functions, not AI, in order to make the AI problem smaller and easier to solve.

To break the problem down well, you must understand the real-world system (not the computer system) that generates the data. Crucially, this involves spending a lot of time talking to the humans who are close to the data! 

Unfortunately, this won't be as concise as "5 sick use cases of ChatGPT!!". Building AI that actually works is hard and can't be boiled down to a 3-minute read. But if you're keen to learn more, then please read on - we're aiming for one new monthly post over the next 6 months.

🙏 Shout-outs 🙏

Huge list this week, big thanks again for everyone who responded to the last call to arms!

Furkan E - for being this month's top intro hero, and the great shout on using https://spiky.ai
Sophia R - for the enthusiastic referral to other founders and being an A++ customer!
Nathan C - for the shout-out on Fiverr and intro to Greg
Kurt H - for a couple of great intros to Rishab and Richard
Richard L - for flagging Founders Factory to us
Dan N - for connecting us to Sergey, was a very interesting convo!
Patrick M - for firing off quite a few intros for us
Camilla D - for opening us up to your portfolio cos
Jayshan R - for linking us with Pavir
Will H - for the intro to Jack
Andrew J - for another fun catch-up (and some good intros ofc)
Arnaud - for connecting us to Matis
Ruchni V - for connecting us with Stephanie (and giving me a laugh with never taking "no")
Bertie V - for several solid plugs, and having the attention to watch our 3-minute video ;)
Peter M - for the intro to Charles
Ryan ON - for the good craic, always, and the connection to Patrick
Ludivine C - for the solid connections to Talip & Abs
James Stewart - for a very serendipitous lead with John

Thanks,
Sam & Yousef

🙋 Ways you can help: meeting companies building in AI / ML 🙋

We want to improve our understanding of how startups/SMEs approach AI and ML projects. We've got a lot more to learn about the market; setting up conversations is the fastest way. Our ideal profile is:

• They're already building AI or ML features

• Their founding team may include software engineers, but aren't AI experts

• They haven't yet hired an ML team in-house

• They have some revenue or funding already

Please get in touch if you know any companies meeting some of these criteria.

🔬 How will an automated ML engineer work?🔬

We got this question a lot in May, so we've put together a 3-minute vision video for our eventual product. If that doesn't sell it, it also features AI, bad jokes and our semi-professional voice acting! Please give it a watch and let us know what you think.

📉 Progress in May 📈

Our current focus is market risk: will companies pay to outsource their AI/ML? In May, we went to market as a service. If customers signed up, we built their AI ourselves i.e. as consultants. 

Our main goal was 2 paying customers. Pleased to say we managed that, so we don't have to chalk up a failed target in our first month as a company!

Primary metric: 2 paying customers ✅ (+1 on paid waitlist)
Revenue (secondary): £14k on a cash-flow basis
Cost (secondary): 101 hours building AI for customers (we want to keep this below 160)

🕵️ Challenges 🕵️

May wasn't all plain sailing - we faced some challenges with going to market:

• Many people believe software engineers and ML engineers are fungible. It's a very different skillset, which seems similar because they both write code. We haven't yet found a concise way to communicate the risks of software engineers trying to implement ML projects.

• We haven't found a repeatable channel for getting in front of our target customer, both of May's customers came inbound. While this shortens the sales cycle, inbound is not repeatable.

• A previous (pre-Artanis) client didn't convert to a longer-term deal, as they still hadn't launched the AI we built. We need to be mindful that building AI that isn't core to a company's value prop is more likely to lead to churn.

🎯 Goals for June 🎯

June's goal is to grow from 2 to 4 customers - we're still concerned about market risk and our milestone for de-risking is 10 customers.

Plan A is outbound warm via our networks (see above: ways you can help). We want to set up calls with people who know/trust us ("allies"), so we can i) collect feedback on Artanis and ii) get referrals to potential customers. We're first aiming for ~50 calls with "allies" and if we can't get 10 decent referrals, we'll change to a different strategy such as cold outbound.

🙏 Shout-outs 🙏

Sophia R - for being a great first customer
Ashley H & Callum N - for a quick process
Tom H - for being May's intro GOAT (and the best commercial data scientist in the UK)
Samatha W - for the inspirational source material for our demo video
Antoine N - for connecting us to Dr. Vivie

Thanks,
Yousef & Sam

🧐 What's this email about? 🧐

This is the first monthly email update for Artanis, our new startup! Artanis is an automated AI engineer. It lives in your Slack channel and builds AI-powered features for you, so you no longer have to spend hundreds of thousands to build AI that actually works. If this doesn't make sense to you, you're not alone - check out our 3-minute explainer video here.

Our mission is to make reliable AI-powered features accessible to all businesses. Imagine being a cafe in the 90s: big tech companies were hiring expensive teams to build websites, but this was way out of your price range. Fast forward to today, and every cafe can now launch a website with Squarespace. We want to make AI just as accessible as websites are today.

🙋 Ways you can help - keep this update relevant 🙋

We've added you because you've expressed interest in the new startup. If that's still the case, then please add sam@artanis.ai and yousef@artanis.ai to your contacts. This will help future updates avoid your spam folder. If you don't want these updates, just let us know and we'll take you off the list. :)

🔬 Who's building this automated AI engineer? 🔬

Artanis has two co-founders: Yousef and Sam. Rather than giving bios, we thought it would be fun to point out some strikingly similar shared experiences:

1. We both founded "AI for fitness" startups prior to Artanis. Despite raising VC backing, both startups failed. We made the same first-time founder mistakes, most importantly picking a market we didn't understand (fitness) rather than our real expertise (AI). Luckily, we share a masochistic streak that's made having another try irresistible.

2. Prior to startup life, we both did PhDs - Yousef's PhD was in CS and Sam's was in AI. We both enjoyed the autonomy of research, but wanted to pursue something with more tangible impact. In retrospect, it's pretty crazy that two CS/AI PhDs decided to found fitness startups.

3. We share common values and ways of working. Most importantly, we focus on telling the truth and doing the right thing, then letting the outcomes sort themselves out. Sometimes easier said than done!

🎯 What's our next major milestone? 🎯

Our next major milestone is 10 paying customers and we're not yet planning beyond that. We don't need a fully scalable self-serve product to get there. Instead we'll be going to market with a service-based approach, where we build a lot of the AI for customers ourselves, and build tools to speed this up behind the scenes.

Each month, we're going to report on:

Primary metric: # of paying customers.

Secondary: Revenue (£). This stops us from cheating by cutting prices too low.

Secondary: Cost (hours building AI for customers). This stops us from cheating by hiring to meet demand - we want to build a product, not a consultancy.

📉 Goals for May 📉

Each update, we'll state a goal for the next month. If we miss this goal, we have to consider pivoting and justify the decision either way. For example: we may miss a monthly sales goal because Sam was sick for 2 weeks. In this case, we wouldn't necessarily need to pivot.

Our goal for May is 2 paying customers. That's it!

🙏 Shout-outs 🙏

Irfan R - for making a great connection for us, despite Sam voicing some "blunt" views
Ben L & Alister R - for the opportunity to get involved in your new venture

Thanks,
Yousef & Sam

https://artanis.ai/