AI Is Making Product Teams Faster. That's Not the Interesting Part.

AI is making product teams faster. That's not the interesting part. The interesting part is what happens when you move fast without being clear on the problem you're solving.

I've seen a lot of technology waves. The failure mode is always the same: teams reach for the tool before they've properly understood the problem.

AI doesn't fix that. It amplifies it.

So before we talk about what AI enables, and it enables a lot - let's start with what it doesn't change, because that's the thread that runs through everything else.

Governance Is More Important Now, Not Less

There's a version of the AI story that goes: the tools are so powerful now that the old disciplines don't matter as much. Move fast, iterate, let the machine figure it out. I've seen that approach play out, and the results are predictable.

What you put into the discovery phase is what you get out the other end. A problem framed through real customer insight, stress-tested assumptions, and a clear picture of the competitive landscape produces fundamentally different AI output than a vague brief and a good intention. That's the variable most teams aren't controlling, and it's the one that matters most.

Not which AI tools to use. Not how to write better prompts. Whether you're in control of the problem definition before anything gets built. Everything else follows from that.

The Spec Is Evolving. The Prototype Is Now Part of It.

With that foundation in place, here's what AI genuinely changes, and it's significant.

Specs aren't going away. Anyone who tells you they are hasn't thought it through. The spec still does something irreplaceable: it defines the boundaries. What you're building, for whom, what's explicitly out of scope, what done actually looks like. Those constraints don't become less important when you're moving fast, they become more important. Without them, the prototype becomes the only reference point, the scope expands to fill whatever AI can generate, and you end up shipping something nobody agreed to build.

What's changing is that the spec and the prototype are converging. The spec used to be the primary artefact - a written document handed to engineering and hoped to be interpreted correctly. Everyone has sat in the meeting where it meant something different to every person in the room. That's the part that's breaking down.

The written spec still sets the boundaries. The prototype makes them tangible. Used together, they surface misalignments faster, give stakeholders something real to respond to, and compress the feedback loop in a way that wasn't possible before. The discipline of writing the spec, being precise about scope, constraints, and what done looks like, is now the thing that determines whether the prototype is a useful validation tool or just an impressive-looking distraction.

Get the spec right and AI speed is your biggest asset. Skip it and you'll find out the hard way.

One Size Fits All Is Over

For most of the last two decades, enterprise software worked on a simple model: build a core product, add configuration options, and hope that covered enough of what your customers needed. Customisation was expensive, slow, and usually reserved for your biggest accounts. Everyone else got what was on the shelf.

Generative AI is dismantling that. The pattern that's emerging, sometimes framed as the shift from software as a service to service as software, is that you build a core base of capabilities, and AI handles the tailoring for each client, each context, each use case.

This has real implications for how you think about your product strategy. Your roadmap can no longer just be a feature list. It needs to be clear about what belongs in the core - the stable, defensible capabilities - and what should be handled by the intelligence layer around it. That's a fundamentally different kind of product thinking, and most teams haven't made that mental shift yet.

The Product Manager's Job Is Changing

All of this points to a shift in what good product management actually looks like. The PM role has always been about judgment - deciding what matters, making tradeoffs, staying close to the customer problem. That doesn't change. What changes is where the time goes.

The best product people I see right now are operating more like orchestrators than specifiers. Less time writing detailed feature requirements, more time doing three things well: defining the problem with enough precision that AI can help solve it, working with subject matter experts to understand the workflows that need to be automated or augmented, and making the calls that AI genuinely can't make - what matters, what doesn't, what the customer actually needs versus what they said they wanted.

Teams that get this division right - using AI aggressively for the things it's good at, staying fiercely human on the things it isn't - will outperform teams that either over-rely on AI output or under-use it out of habit or caution.

The Portfolio Opportunity Nobody's Moving Fast Enough On

Most organisations I talk to are focused on the future - what agentic AI means for their product strategy in two or three years. That's a valid and important question. But there's a near-term opportunity sitting right in front of them that they're moving too slowly on.

Existing product portfolios can be meaningfully enhanced right now. Problems that your customers have been living with for years - because solving them used to be expensive, complex, or slow - are now solvable in weeks. This is where the immediate investment should be: identifying the highest-value problems in your existing portfolio and iterating quickly with AI-assisted development to solve them.

That's not instead of thinking about the agentic future. It's in parallel with it. The teams doing both - enhancing now and designing for an agentic core - are the ones that will be best positioned. The agentic shift changes how capabilities get built at a fundamental level. You're not designing features anymore, you're designing workflows. And the subject matter experts inside your customers' organisations become central to that process in a way they never were when you were building conventional software.

The Thread That Runs Through All of It

Every one of these shifts - faster prototyping, tailored software, the changing PM role, the portfolio opportunity, the agentic future - is enabled by AI and constrained by the same thing: how well you understand the problem before you start building.

That's not a new idea. It's the oldest idea in product development. What's new is the stakes. When building was slow, a poorly defined problem meant a delayed product. When building is fast, a poorly defined problem means a fast mess - shipped quickly, hard to unpick, solving things nobody needed.

• Rethink what a spec is for. It still defines the boundaries - scope, constraints, what done looks like. What's changed is that a working prototype should sit alongside it, not replace it. The two together surface misalignments faster than either one alone.
• Think in layers. What belongs in your core product? What belongs in the intelligence layer? That's now a strategic question, not just an architecture one.
• Move on your existing portfolio now. Don't wait for the perfect agentic strategy - there are high-value problems you can solve in weeks with what exists today.
• Stay human on the things that matter. Problem definition, customer judgment, strategic tradeoffs. AI doesn't do those. You do.

The oldest idea in product development hasn't changed: understand the problem before you reach for the solution. What's changed is the cost of getting it wrong. When building was slow, a poorly defined problem meant a delayed product. When building is fast, it means something shipped quickly that nobody needed - and a team that doesn't understand why.

That's the problem Product Clarity is built around - not replacing the judgment, but making sure the judgment is grounded in the right structure before anything gets built.