What Building With AI Has Actually Taught Us
Series note
This is Part 1 of a four-part Field Notes series on how we use AI at Gather & Grow, and what it is teaching us about leadership, ownership and change. In this first piece, we are starting with the shift that changed the most for us: moving from using AI as a tool to building AI into the rhythm of the work itself.
What Building With AI Has Actually Taught Us
We started using AI because it helped us move faster.
We kept building with it because it showed us where our thinking, systems and ways of working needed to become clearer.
That has probably been the biggest surprise. The more we have built with AI inside Gather & Grow, the less it has felt like a story about tools. The tools matter, of course. But what AI keeps reflecting back to us is much more human: clarity, trust, ownership, feedback and the way work actually moves through a team.
We have been meaning to write about this for a while. We have been building with AI for the better part of two years now, and along the way we have picked up a few lessons we hope will be useful to the leaders and founders we work with.
We do not think we have cracked it. This change is still moving. We are still learning as we go. That is why it feels like the right time to write it down.
Where it started
Like many people, we started with ChatGPT.
It became our first real hub: project after project, conversation after conversation, figuring out what this new way of working could actually do for a growth studio.
A huge amount of our early confidence came from Sebastian and Mary at Haavn. They helped us become genuinely strong users of ChatGPT and AI more broadly, back when most people were still using it for quick experiments. We credit them for a lot of what followed, including our eventual move to Notion as the operating system underneath the business.
From there, our thinking and our toolkit kept expanding.
For a long time, Claude was something I turned to the way many people do: a smart assistant for drafting, thinking things through and getting unstuck. That is still true. But somewhere along the way, the relationship changed.
I stopped only asking it to help me write, and started building things with it that now run inside the studio.
That shift matters. A tool is something you pick up when you need it and put down when you do not. What we have now feels more like infrastructure. It does real work in the background, and it has changed what we notice, what we remember and how we make decisions.
The Story Harvester
The clearest example is something we call the Story Harvester.
Every day, Leah and I are in conversations with leaders navigating real friction: a founder realising they have been avoiding a hard conversation, a leader noticing they have been operating instead of leading, a team finding its way back to trust.
Those moments are exactly what our work is built around. There is so much learning inside them, and for a long time, too much of it disappeared as soon as the meeting ended. We would finish a call, move on to the next thing and lose the story that carried the insight.
So we built something to catch them.
Every night, an automation scans our Notion workspace for meetings from the last 24 hours, reads the transcripts and pulls out the single most story-worthy moment from each one. It writes a short, anonymised story in our voice and adds it to a running Story Bank.
Nobody has to trigger it. We wake up and the stories are there, ready to shape into something useful.
It sounds small when you say it out loud, but it has changed how we work. The raw material for our content used to depend on someone remembering to write it down in the moment. Now it has a place to land.
It has also kept us close to what is true for the people we work with. Reading back through weeks of stories helps us see where the real friction is: where decisions are still sitting with one person, where ownership is fuzzy, where teams are saying yes to change but still behaving in familiar old patterns.
That shapes how we coach. It shapes what we build next. It also helps Leah and me stay connected across very different client conversations, without relying on a hurried catch-up to understand what is happening.
In that sense, the Story Harvester is more than a content system. It is a listening system. It helps us notice what growth is asking of the leaders around us, before those patterns harden into assumptions.
The unglamorous one
The other example is less poetic, but just as useful.
An agent now runs on my inbox every hour. It reads incoming email, labels it by project, archives the noise and drops draft replies into my drafts folder when something needs a response.
I still read and send everything myself. The judgement stays with me. But the blank page problem, the thing that used to eat twenty minutes before I had even started replying, is mostly gone.
A year ago, neither of these systems existed inside the studio. Now they both run without me thinking about them, which is really the point: the useful shift has not been having another place to generate ideas, but building support into the rhythm of the work itself.
What actually changed
The first version of anything we built was never the version that stuck.
The Story Harvester went through several rounds of teaching it what counted as a genuine story rather than a forgettable summary. The inbox agent needed correction before its drafts sounded like me and not like a generic assistant.
That has been useful in itself. AI has made the thinking work more visible. It has shown us where our standards were unclear, where our systems needed more structure and where work was still depending on memory or effort rather than design.
That is the lesson I keep coming back to.
AI has not taken the thinking work out of running a studio. It has given us more reason to do that thinking properly.
The useful work begins when AI helps you notice what was already happening.
Keep reading
In Part 2, we look at what this means for a small studio in practice: how AI helps us carry more without losing the human thread, and why shared context matters more than simply adding another tool.
In this series
Part 2: How a Two-Person Studio Uses AI Without Losing the Human Thread
