The AI Disconnect: Individual Heroes, Organisational Vacuums

Let’s cut through the noise. The narrative of 2025 has been that “AI changes everything.” Our data suggests something different: AI is changing individuals, but it is barely scratching the surface of organisations.

We surveyed 53 product leaders across the UK, North America, and APAC. The results paint a picture of a split reality. On one side, we have product people using AI to save themselves from burnout. On the other, we have organisations that demand innovation but refuse to build the systems to support it.

Here is the cold, hard reality of AI readiness in 2025.

The “Shadow Efficiency” Boom

The adoption numbers look fantastic on a slide deck.

• 91% of respondents report some level of AI integration.

• 96% save at least one hour per week.

• 73% save more than three hours per week.

But don’t mistake this for organisational maturity. This is survival.

While usage is high, 64% of respondents are still “exploring tools individually,” while only 6% are scaling AI across departments. This isn’t strategy; it’s shadow IT. Product people are using ChatGPT and Claude to claw back time from administrative bloat, but because this usage is siloed, the organisation learns nothing. You are saving time, but you aren’t building equity.

You Can’t Automate Chaos

This is the most damning statistic in the entire analysis:

• 77% agree AI improves the product lifecycle.

• Only 42% agree their team follows a clear product lifecycle.

The Insight: You cannot accelerate a process that does not exist.

If your team doesn’t have a defined way of working, AI will simply help you generate subpar PRDs, rapid discovery with low-signal input or accelerate siloed thinking. The blockade to AI isn’t technology; it is the lack of a clear operating model. It’s no surprise that 32% of respondents are unsure how AI even applies to their role—not because they lack imagination, but because their role lacks definition.

The “Innovation Theatre” of Leadership

Leadership sentiment is high from the people we surveyed. Everyone wants to be an “AI-first” company. But looking at the blockers reveals a transfer of anxiety from the C-Suite to the team:

• 45% cite “poor integration” as a major blocker.

• 42% cite a “lack of time or space to experiment”.

The Hard Truth: If you tell your team to “embrace AI” but do not clear their calendar or integrate their tools, you are engaging in innovation theatre.

Buying Copilot licences without investing in middleware or training time is like buying a Ferrari to drive in a school zone. It looks good, but you aren’t going anywhere.

We Are Optimising the Wrong Things

Where is the time actually going?

• High Usage: Writing PRDs, synthesising research, and drafting updates.

• Low Usage: Experiments, post-launch analysis, and forecasting.

We are using supercomputers to focus on automating or providing assistance to achieve mundane and manual steps (everything left of prioritisation).

While streamlining documentation, automating research and updates is valid, the high-leverage work—deciding what to build, running experiments and measuring if it worked—remains largely untouched by AI. We are optimising the administrative periphery of the job while leaving the strategic core exposed.

The 2026 Prediction: The Great Divergence

Based on this data, the gap between the “Tool Users” and the “System Builders” will widen in the coming year.

• The Feature Factory will use AI to ship more output, creating more bloat and higher churn.

• The Product Led companies will use AI Agents to synthesise research, pin-point want to build, automate low-value decisions, freeing up oxygen for deep work.

Your Next Step: Stop looking for the perfect prompt. Start fixing your process and mapping how you build today. If you want to be “AI Ready,” define your PDLC. Give your team the time to experiment. And stop treating AI as a magic wand for broken structures and systems. We still have to do the hard work to get the gains.

FYI: We will be publishing our thoughts on how you go from 50 > 18 steps in your PDLC soon.