Inside Purple's innovation department,
AI isn't a tool. It's a teammate.
We handed the admin of innovation to AI and kept the human parts for ourselves.
It's Friday afternoon, and a progress report just landed in the team's channel. Every active innovation project — what moved this week, what's stuck, what's next — written up in plain English, posted without anyone asking for it. No project manager sent a reminder. No engineer stopped building to fill in a status template. Nobody chased anyone.
That's the part that surprises people first, so it's worth sitting with: nobody chases anything. The update writes itself because, in this department, writing the update is an AI's job — and it just does it, every week, like any reliable teammate would.
We didn't adopt AI. We gave it a seat.
It's easy to say a company "uses AI." Most do now — a chatbot here, an autocomplete there. That's not what happened at Purple. We built a department where AI is a working member of the team, sitting inside an operating system that runs end to end. An idea comes in, gets a fair hearing, gets claimed by a team that wants it, gets built, reports on its own progress, and feeds what it learns back into planning — with barely a human touching the admin in between.
The humans are still very much in charge. We make the calls that matter: which ideas are worth doing, how to build them, when something's good enough to ship. What we handed over is the overhead — the capturing, the chasing, the duplicate-checking, the status-writing that used to eat the day and quietly kill good ideas before they got a chance.
One idea, all the way through
The whole department is built around a single loop, and once you see it, the place stops looking like a pile of tools and starts looking like one organism. Follow one idea around it.
Intake. You have an idea. Instead of it dying in a meeting or an inbox, you just talk to an AI — today the Idea Intaker, and soon Project Portal, the single front door we're building for innovation work — which asks the right questions, quietly checks whether someone's already proposed something similar, and files it as a properly structured entry on our innovation register. You have a conversation, and your idea is now real and on the board.
Judge. A small human forum decides — back it, send it for a proof of concept, or park it with a reason. This part is deliberately human. Taste and judgement are exactly the things you don't automate.
Claim. When an idea gets the green light, the system automatically opens the project up — no manager hand-assigning tasks. Teams browse what's approved and pull the work they want. People build best when they've chosen the thing.
Build. The work gets built AI-first, end to end. This is where speed shows up: things that would traditionally take a long stretch of calendar time come together in a fraction of it. We're deliberately not throwing a headline multiple at you here — we're measuring the real number from the actual build history and will publish it once it's derived rather than asserted.
Report. Every Friday, the Progress Reporter writes that plain-English summary and posts it. Separately, progress is pushed into the company's goal-tracking, so the quarterly objectives stay current without anyone reconciling spreadsheets.
Plan. All of that — what shipped, what stalled, what got learned — flows back into what the department takes on next. The loop closes, and the next idea starts its lap.
How it actually works (for the curious)
Under the hood it's refreshingly unmagical. The department runs on four AI agents plus a couple of small automations, all on one Jira board we call the Innovation Registry. Decisions are human; the admin is automated. That's the whole trick.
The Innovation Forum — actual humans, Chris and Jaco — decides what happens to each idea; on approval, the epic is created automatically. This quarter the department has twelve connected platforms in flight under five objectives — in flight, being built, not all shipped yet.
See it working
Where this is heading: idea capture is moving into Project Portal, so submitting an idea, tracking where it's got to, and discovering approved work to claim will all live behind one front door — the same loop, an even smoother way in.
Why this is different
Plenty of places have an "innovation process." Usually it's a form, a committee, and a graveyard of good ideas nobody had time to chase. The difference here isn't that we move faster because people work harder. It's that the system absorbed the boring, repetitive parts, so the human energy goes where it's valuable: having ideas, judging them well, and building them.
That's the model Purple built. Everything else we'll show you in this series — each platform we launch — is another organ of this same body. This is the body.
The leverage
If the admin of innovation costs nothing, the binding constraint is no longer coordination overhead — it's the quality of the ideas and the judgement applied to them. A far better constraint to be limited by. The department can carry more in flight than its headcount would historically allow, progress is always visible, and the model is repeatable. The question isn't "is this clever?" It's "where else would this loop pay for itself?"
Belonging
This is where your ideas go now — and they get a real hearing, fast. You have a conversation, your idea lands on the board, and a human forum looks at it within the week. The system respects two things: your time (no chasing updates) and your idea (judged on its merits in the open, not on whether you were in the right meeting). Bring it the next one.