Why your AI productivity gains might be going nowhere.

Faster rowers. Slower boat.

THOUGHT PIECE | AI STRATEGY

Introduction

Why your AI productivity gains might

be going nowhere

Ben Hunt–Davis and his GB rowing crew built their path to Olympic gold around a single, relentless question: will it make the boat go faster? Not will it make me feel stronger. Not will it impress the coach. Will it, specifically and measurably, make the boat go faster?

It’s a framework that has since been adopted by boardrooms across the world, and rightly so. Clarity of purpose. Ruthless prioritisation. No wasted energy. But I want to borrow the metaphor rather than the lesson, because the AI wave now sweeping through organisations is producing something the rowing team never would have tolerated: a boat full of faster rowers going nowhere faster.

“

Scaling individual output is not the same as removing the constraints that slow the organisation down.

THE PRODUCTIVITY TRAP

Ask any knowledge worker today whether AI has made them more productive, and the answer is almost universally yes. Emails drafted in seconds. Summaries generated from hour–long documents. Code written, slide decks assembled, research synthesised, all faster than before. The individual performance gains are real – or at least they feel like they are.

And that matters. Getting tools into people’s hands, building familiarity, lowering the activation energy around AI – this is not wasted effort. It is the foundation everything else is built on.

Ask the CEOs and CFOs of those same organisations whether AI has made the business materially faster, and the answer is considerably less clear.

This is not a paradox. It is what happens when you optimise the parts without ever looking at the whole. Speed up a step that isn’t the constraint and you don’t move any faster – you just build a faster queue in front of the same wall. The work arrives more quickly. The bottleneck holds. The boat does not accelerate.

THE QUESTION BOARDS SHOULD ACTUALLY BE ASKING

The instinct is to measure AI success in terms of headcount productivity – output per person, tasks automated, time saved. But most of these numbers are self-reported, loosely defined, and difficult to verify. Organisations are making significant investment decisions on the basis of how productive people feel, not how productive they are.

The right question is not how much faster are our people working. It is what has actually changed for the customer, the cost base, or the speed of the business.

If you cannot answer that, you are not measuring AI performance. You are measuring activity – and not very accurately at that.

Getting to a real answer requires looking at the full length of a process – from customer need to customer outcome – and being honest about where it actually stalls. Where are decisions slow? Where do handoffs fail? Where does work sit waiting for something that has nothing to do with how fast any individual is moving? A bottleneck relieved at step three means nothing if step seven is unchanged. And that is the step that drives the most business value.

WHAT THE TRANSITION ACTUALLY DEMANDS

The organisations that will extract genuine business velocity from AI are those that treat it as a system design problem rather than a productivity tool. That means starting with the outcome, the equivalent of the gold medal, or in commercial terms the customer experience, the speed to revenue, the cost to serve, and working backwards to identify what actually constrains it.

It means investing in the connective tissue: integrated data, shared process ownership, cross–functional decision rights. AI cannot overcome organisational fragmentation. In many cases it makes that fragmentation more visible, as the bottlenecks that remain become starker against a backdrop of everywhere else moving faster.

And it means the board and executive team taking a position. Not on which AI tools the business should use, but on which outcomes AI is expected to move, and how progress against those outcomes will be measured. Without that clarity, AI investments will continue to generate impressive individual productivity statistics and frustratingly modest business results.

Hunt–Davis and his crew did not ask whether training was making them feel like better rowers. They asked whether it was making the boat go faster. The distinction sounds simple. In practice, it demands exactly the kind of clear–eyed, system–level thinking that separates organisations that extract transformational value from AI from those that end up with very busy people and a boat that has barely moved.

The question for your leadership team is not are we using AI? It is: what, specifically, has it made faster, and does that actually matter?

HOW WE APPROACH THIS

This is precisely the problem that Enterprise AI and ADAPTOVATE have built their partnership to solve. Enterprise AI is an AI platform company that helps organisations re–design the way they work with AI, cutting through the technology noise to focus on where intelligent systems can reshape processes, not just accelerate tasks. ADAPTOVATE is an agile transformation consultancy that specialises in benefits realisation, customer journey mapping, and the hard work of reorganising people, processes and technology to produce results that actually show up in the numbers that matter.

Together, the approach is deliberately different from the prevailing model of licence deployment and adoption dashboards. It starts with the outcome, maps the process constraints that prevent it, and applies AI where it removes those constraints, not where it is simply easiest to implement.

CASE STUDY

DAISY: reimagining development approvals end to end (click here to find out more)

One of the clearest examples of this approach in practice is DAISY, the Development Application Intelligence System built by ADAPTOVATE on the Enterprise AI platform. Development approvals in Australia are a process notorious for bottlenecks: applicants navigating complex regulatory requirements, councils spending the majority of planner time chasing missing or incorrect documentation, and approval timelines that frustrate communities and stall housing supply.

DAISY was not designed to make planners faster at their existing tasks. It was designed to remove the bottlenecks that make the entire process slow. By guiding applicants through eligibility checks, flagging constraints before lodgement, and automatically validating documentation against council requirements, it attacks the problem at its source: incomplete and non–compliant applications, rather than downstream symptoms.

20% faster DA processing times

49% increase in service capacity

200 enquiries resolved per week without human intervention

The boat went faster because the process was redesigned, not because individual rowers were given better oars.

A FINAL THOUGHT

None of this is a criticism of the organisations living it. The pull toward visible, measurable, individual progress is entirely human. When a new technology arrives, the natural instinct is to put it in people’s hands and watch what happens. That is not naivety – it is how adoption has always worked, from the spreadsheet to the smartphone. And the organisations encouraging their people to embrace AI tools today are ahead of the curve, not behind it. That step is real, it counts, and it is the necessary precondition for everything that follows.

The organisations deploying AI tools today are not getting it wrong so much as getting it first. The productivity gains at an individual level, however imperfectly measured, are the foundation. The coordination and process redesign that unlocks business–wide velocity is what comes next.

Every major technology shift has followed this pattern. The businesses that move from the first chapter to the second are the ones that pause, look honestly at where the system is still slow, and ask the harder question. That is not a criticism. It is an invitation.

You can download this document here.