The product triad has been the default model for building software for over a decade. One product manager. One designer. One engineer. A neat triangle of ownership, each point accountable for their domain, all three aligned in theory and — if you've worked in any real product team — frequently misaligned in practice.
It was a good model for its time. That time is ending.
Not because the people aren't capable. Because the model was built around constraints that no longer exist — and we're still running it like they do.
What the Triad Was Really Solving For
The triad wasn't a philosophy. It was a workaround.
Software used to be genuinely hard to build. Research took time. Design required specialist tools and specialist training. Code had to be written by hand, line by line, by people who'd spent years learning how. So you divided the labour. You gave each domain to someone who'd spent years inside it, and you created coordination rituals — standups, sprint planning, design reviews, handoffs — to stitch it all back together.
The triad was a response to scarcity. Scarcity of skill, scarcity of speed, scarcity of people who could hold more than one domain at once.
That scarcity is dissolving.
What AI Has Actually Changed
This isn't about replacing designers or engineers. Anyone saying that is missing the point.
What AI has changed is the cost of crossing domain boundaries.
A product manager can now prototype an idea in an afternoon without knowing how to use Figma. A designer can build a working version of their concept without writing a line of code. An engineer can conduct user research, synthesise findings, and produce a document the team can act on — without a dedicated researcher on the team.
The walls between disciplines haven't disappeared. But the toll for crossing them has dropped dramatically. And that changes what a team can look like, how fast it can move, and — most importantly — who needs to be in the room.
We've felt this directly. Research that used to take a week now takes an afternoon. Prototypes that used to require a developer handoff get built by the person who had the idea. Competitive analysis that used to sit in a backlog gets done the day the question comes up.
When you remove the bottlenecks, the structure built around the bottlenecks stops making sense.
The Problems We're Actually Solving
Before we talk about what replaces the triad, it's worth naming what was always broken about it.
Handoffs create distance from the problem. Every time work moves from PM to designer to engineer, something gets lost. Context, intent, nuance. The person closest to the decision is rarely the person building the solution. The person building the solution is rarely in the room when the customer talks.
Ownership silos create defensiveness. When your identity is tied to your domain — "I'm the designer, this is my work" — feedback feels personal and collaboration feels threatening. The triad encourages this by design.
Coordination overhead scales badly. The more people in the triangle, the more time spent aligning them. In a small team this is manageable. As the team grows, the coordination becomes the work.
Speed is constrained by the slowest handoff. You move at the pace of the last person to receive the brief. When that person is blocked, everyone waits.
None of this is about individual people failing. It's about a structural model that optimises for specialisation at the expense of speed and coherence.
What We're Moving Toward
We're not replacing the triad with a free-for-all. The disciplines still matter. Design thinking, engineering rigour, product strategy — these aren't going away. What's changing is how we hold them.
Fewer, more complete people. Instead of three specialists who each own a slice, we're looking for people who can hold the whole problem — and use AI to fill the gaps at the edges. A designer who can prototype in code. A PM who can run their own research. An engineer who thinks deeply about the user. These people have always existed. AI makes it viable to build teams around them.
Decisions made closer to the customer. The person who talks to users should be able to act on what they learn — not write a brief and hand it to someone else. We're shortening that loop deliberately. Faster feedback, less translation, fewer opportunities for the signal to get diluted.
AI as the fourth member of the team. Not a tool you open occasionally. A constant collaborator — doing the desk research before the meeting, synthesising the session notes after it, turning the rough wireframe into a working prototype, flagging the edge case nobody spotted. When AI is genuinely embedded in the workflow, one person can do what three used to.
Documentation as the coordination layer. Instead of aligning through meetings, we align through clear, living documents. Decisions recorded, context preserved, anyone who joins late able to catch up without a 45-minute call. Claude Projects has become our institutional memory. It doesn't replace communication — it makes communication more deliberate.
What This Doesn't Mean
It doesn't mean small teams are always right. There are problems that genuinely need deep specialisation — complex systems, accessibility at scale, performance engineering. The triad still makes sense there.
It doesn't mean generalists replace specialists. It means the ratio is shifting. And it means the way specialists collaborate needs to change — less handoff, more overlap, more shared ownership of outcomes rather than outputs.
And it doesn't mean AI makes the human judgment less important. If anything, the opposite. When AI can handle the mechanical parts of each discipline, what's left is the harder work — knowing what to build, why it matters, and what good actually looks like. That still needs people. It just needs fewer of them, working differently.
The Honest Part
We're still figuring this out. The new model isn't clean yet. There are things the triad gave us — clear accountability, predictable process, defined roles in a job description — that don't have clean replacements.
But the discomfort of figuring that out is more productive than the comfort of running a model that no longer fits.
The teams that will build the best products in the next five years won't be the ones who hired the most specialists. They'll be the ones who figured out how to stay closest to the problem, move fastest from insight to working software, and use AI not as a novelty but as genuine infrastructure.
That's what we're working toward. We don't have all the answers yet. But we're asking the right questions.
If you're navigating similar questions about how your product team is structured in the age of AI, I'd genuinely like to hear how you're thinking about it. [Reply here] or find me on LinkedIn.