Process Market Fit - by kingsley - thoughts with data
thoughts with data
SubscribeSign in
Process Market Fit
kingsley<br>Jul 01, 2026
Share
Since Glitch was founded we’ve had Solution Market Fit. We knew our customers’ pain (underperforming ads, impossibly complex ad platforms and poor audience measurement).<br>Thanks for reading thoughts with data! Subscribe for free to receive new posts and support my work.
Subscribe
We immediately did many things that didn’t scale, manually reviewing ads, setting up conversion tracking by hand, and long in depth calls with clients to understand their business and audience and work on optimising everything. In the background we built our app, that could reliably create campaigns that outperformed humans, by as much as 5x all while reducing the time to create ads.<br>All of which reads like the playbook for getting PMF. The truth is we thought we hit PMF several times. My definition of PMF is probably a higher bar than yours, if someone doesn’t sue you if you try to turn off your product, it’s not PMF. Yet we strongly thought we were there.<br>The following should feel familiar to anyone working in AI, we hit PMF sometimes. The product worked sometimes. I don’t mean it crashed, I meant that the output was sometimes good enough and often not. And it changed over time without us modifying the code.<br>Coming from the traditional software universe this doesn’t happen. At first we assumed this was a hardening problem. That we needed to get the bugs out of the system, write better evals and create more rigorous harnesses. We realised the real issue was that our product did not match the capabilities of the model or just as dangerous overestimated them.<br>However the models marched on, what worked in 2024 did not work in 2025, was not good enough in 2026. The capabilities of AI increased, and the intelligence. The idea that you can build a product that will continue to have PMF between model increments is bullshit.<br>Product Market Fit was coined during the great scaling period of software. It is a term that really means Cloud. Your software was not notably moving forward, but you were no longer constrained by infrastructure. PMF was simple, give your clients the capabilities they want, the capabilities that were possible with a static core software offering then put it in the hands of anyone on the planet. Easier said than done, but software’s core capabilities were not changing month on month. This is no longer possible.<br>People talk now about Service as Software. We’re seeing the creation of a new industry in real time where the core job is to understand what is done solely by AI, solely by a human or requires a skilled operator . What was physically impossible last week could be trivial this week. And what worked last week, may now be holding the model back.<br>Everyone’s PMF is currently a facade. If you say you have PMF in AI and your software is not regulated, I don’t believe you. If you’re not making a pacemaker you don’t have PMF.<br>The good news is that you can have Process Market Fit. If your company were to stop trading your customers would sue you. Your customers don’t care, to them there’s no difference. They get billed and they get the results they want.<br>Your company, your culture and your product needs to survive and thrive when the AI models uproot the game every quarter. If you’d asked me 2 years ago if that were possible I’d have laughed, but AI is in also the key to this. When shipping at AI speed, the key is to be that skilled operator for your own business. Your AI processes must also be updated to handle new capabilities.<br>For now the process is clear, every two months AI will manage more of your business. The speed at which you adopt and integrate this will determine if you have Process Market Fit or if you are swimming against the tide.
Thanks for reading thoughts with data! Subscribe for free to receive new posts and support my work.
Subscribe
Share
Discussion about this post<br>CommentsRestacks
TopLatest
No posts
Ready for more?
Subscribe
© 2026 kingsley · Privacy ∙ Terms ∙ Collection notice<br>Start your SubstackGet the app<br>Substack is the home for great culture
This site requires JavaScript to run correctly. Please turn on JavaScript or unblock scripts