Our Actual AI Toolkit
Series note
This is Part 3 of our four-part Field Notes series on how we use AI at Gather & Grow. In Part 1, we shared what building with AI has taught us. In Part 2, we looked at how a small studio can use AI without losing the human thread. Here, we are opening up the practical stack behind the work.
Our Actual AI Toolkit
This is the practical tour.
If you are a founder or leader wondering where to start with AI in your own business, it can be tempting to begin with the tools. Which platform should we use? What should we automate? What is everyone else doing?
Those are reasonable questions. But before we choose tools, we try to look at the work.
Where does context live? Who owns decisions? Where do handoffs break down? Where is feedback slow, vague or overly dependent on one person? Where are people waiting for permission or clarity before they can move?
Your AI stack will only be as useful as the operating habits underneath it.
Here is what is in ours, and what each part helps us practise.
Claude and Cowork: clearer briefs and better delegation
Claude is where much of our thinking happens now: drafting, structuring, problem-solving and building the automations we have written about elsewhere in this series.
Cowork is where that thinking turns into coordinated work across the different threads we are running at once: client deliverables, content, community programmes and internal projects.
Rather than each of us holding the full picture in our heads, or scattering the work across half-updated documents, Cowork lets us delegate real chunks of work and trust that they will come back with shape.
This has made us more specific about what good delegation actually requires.
A fuzzy brief still creates fuzzy work, even when the collaborator is AI. The more clearly we can name the outcome, context, constraints, audience and quality bar, the more useful the work becomes.
That is a leadership practice as much as a technical one.
Notion: shared context and one source of truth
If Claude and Cowork are where the thinking and doing happen, Notion is where the work lives.
Every client engagement, meeting transcript, content idea and programme-in-progress sits in one connected workspace. That matters more than it sounds like it should.
None of our automations, especially the Story Harvester, would work without a single source of truth underneath them. Messy, scattered documents cannot be read by an agent in the same way a well-structured workspace can.
This is where AI has made our foundations more visible.
If the workspace is unclear, the automation is unclear. If ownership is fuzzy, the tool exposes it. If decisions are sitting in someone's head, the system cannot help move them forward.
Notion has become the place where shared context turns into usable work.
Riverside FM: capturing what is already useful
We use Riverside FM for the video content we create, both for clients and for our own marketing.
The AI audio cleanup has worked well for us, with one click. Captioning is built in too, which means we have removed a separate step from the workflow.
The feature we lean on hardest is Magic Clips. It listens through a longer recording and finds the moments where someone says something that lands, then pulls those moments out automatically.
For a studio built on capturing real stories from real leaders, that matters.
The value is not only speed. It is pattern recognition. It helps us see what people come back to, what feels useful enough to share and where the energy is in a conversation.
The tool does some of the finding. We still make the judgement.
The New Terrain producer skill: feedback and creative structure
Our podcast, The New Terrain, runs on a similar stack.
We have built a producer skill inside Claude that reviews episode transcripts, suggests where a long conversation should naturally split into parts, flags moments of crosstalk for our editor and gives us structural feedback on flow and pacing.
Riverside then supports the finishing touches: highlight clips, show notes and polish.
Between the two, an episode can move from raw conversation to something ready to shape without either of us needing to sit through hours of footage from scratch.
Again, the point is not that the tool replaces our judgement. It gives our judgement a better starting point.
The real toolkit
None of this is especially exotic. It is Claude, Cowork, Notion and Riverside FM.
What makes it work is that each tool supports a behaviour we already care about: clearer thinking, better delegation, shared context, faster feedback and more useful capture of what is already happening in the work.
That is the deeper point.
A toolkit is only as strong as the behaviours around it. AI can help a team move faster, but it will also ask that team to become clearer about what it is asking for, who owns what and how decisions get made.
That is where the practical work sits.
Keep reading
In Part 4, we step back from the toolkit and make the bigger argument: AI adoption is a leadership practice, because it depends on clarity, feedback, ownership and the way teams change their behaviour.
In this series
Part 1: What Building With AI Has Actually Taught Us
Part 2: How a Small Studio Uses AI Without Losing the Human Thread
