Today we are publishing Made by Many's Agentic Playbook. It was written for our own people and our own agents. This is why we are sharing it, and what you will find in it.
Gonna start by reminiscing. Can’t help it. It’s my age.
When we started Made by Many in 2007, we shared almost everything. We almost certainly over-shared. We were, and remain, unreasonably opinionated and I think that’s been our saving grace. But it wasn’t just that.
We learn from sharing. We were after all trying to describe an emergent practice area that had never existed before. That worked best as a conversation between peers all trying to do the same thing. The market was moving fast. We learnt a lot.
We were also born in the over-sharing era, when everything was a conversation anyway. “Markets are conversations”, as the Cluetrain Manifesto famously put it in 1999. And social hadn’t yet gone toxic yet. Within the studio, we used Twitter as an internal messaging app - broadcasting a constant stream of whatever we were thinking about and larking about with each other in the open.
Then came Slack. Maybe we grew up and started taking ourselves too seriously. We lost our mojo a bit. The rules of sharing changed and all the noise overwhelmed all of the signal. Social technology got stolen. Hijacked by the marketing-industrial complex.
Whatever it was, we stopped sharing.
I think some of it was simply the pace of change slowing, and there was less to say. But here we are again. Things are moving at least as fast as they were in those early years. If anything, faster. So, we are going back to the habit.
What changed
Our first response to AI was the wrong one. Like a lot of firms, we assumed we had to invent something new to sell, and we spent time trying to work out what it was. What happened instead was that people across the studio started using the tools on real work, and it became obvious that they were a better way of doing what we already do. Making has always been how we learn. Now that the tools have caught up, everyone here can do it. Every time we have drifted away from that idea we have been stung, so the playbook is built around it.
Most of what our industry says about AI is about compression: the same people and the same process, ten per cent faster. The research tends to agree, with study after study putting the productivity gain somewhere between 5% and 15%. Almost all of those studies measure an old process with agents added to it. We are more interested in what happens when you change who gets to make things. A strategist can test an idea as a working prototype the same afternoon, and a designer can build in production code, so what they design is what ships. Engineers get to spend their time on the hard problems. Last week two of our strategists, neither of them an engineer, started building a tool on our CRM's API in a shared codebase, and within a day they were working the way engineers work. That is the change we care about.
None of this is a lab exercise. We are already working this way inside one of our biggest and most complex clients, and it is working. We are learning something every day, mostly about what it takes for the method to hold up inside a large organisation, with its systems, its governance and the people whose jobs it changes. That is far more useful to us than getting it running in a brand new start-up, where there is nothing to work around. What we think we are proving is an emerging consulting model, one that brings making together with design to change the way a business works, with the client's own people doing the work alongside us.
What is in it
The playbook has four parts. The first is the principles we work to, three of them, covering what people do and what agents do, how a project's lessons change the next one, and why redesigning just your tooling and not your process isn’t enough. The second is how we make: the product lifecycle every project and prototype runs on, from plan to operate, written as skills our agents run. Each stage says what a person does and what an agent does, and each skill carries nineteen years of how the studio works. The third is the knowledge our agents consult, our published positions on design systems, accessibility, search and AI discoverability, and designers who build. The last is the learning programme we are using to bring the whole studio with us, from basic AI fluency to making something real and putting it into production.
How to read it
This is a working document that we are using internally to join up the change happening.
It’s not marketing, which is part of the point. A lot of what is being published about AI at the moment is marketing presented as method. In contrast, this is the material our own people are using, put where anyone can read it, and it is closer to an engineering blog than a campaign. If you are the person inside an organisation trying to work out how any of this would work in practice, it is written for you as much as for us. The site is built by the suite it documents, and the suite itself is on GitHub.
What comes next
This year we have used our own method on ourselves, and at times it has been uncomfortable. We have made things to find out, got some of it wrong, and changed direction. The playbook will keep moving in the same way. Positions will change when projects prove them wrong, gaps will become drafts, and the learning courses will open to the studio first and more widely after that. Over the coming weeks people from across Made by Many will write about the parts they know best, starting with Andrew Flett, who has built most of it.
If you read it and disagree with something, tell us. That was always the best part of working in the open.
Read the Agentic Playbook.
