The AI Productivity Paradox: How Stigma Is Holding Organizations Back

Why is it that the obvious uses of artificial intelligence (AI) to make work more productive does not seem to be obvious? Around a billion people are using Generative AI, in every sector, but the evidence for productivity gains was sparce.

In researching how productivity was measured, the traditional number was Gross Domestic Product (GDP). But this was too abstract and general; too removed from the real world of actual people working in real organizations. It was clear that people found these tools productive and their utility clear. What was actually happening on the ground was less obvious.

First, I realised that the AI adoption (fastest in the history of tech adoption) was unlike any other – it was bottom up. The script had been flipped. Individuals found it useful but did not necessarily welcome it within organizations. The senior management, whether in universities, public sector or corporations, were nervous, even negative, about its worth. This led to a bottom-up, AI-on-the-sly phenomenon, where the use of it by students, employees and managers was widespread but hidden.

This is a paradox. But what other paradoxes existed in this ‘productivity problem’?

Behavioral paradoxes

We are creatures of habit and full of biases and these create many behavioral paradoxes. These are:

· Confirmation Bias - narrows our behavior and learning; we keep repeating what works for us and miss newer, better methods.

· Status Quo Bias - makes switching costs feel larger than they are, so we stick with processes we are familiar with and know well.

· Negativity Bias - makes us overweigh risks, so adoption slows under exaggerated fear of loss, blame or failure.

· Anthropomorphic Bias - where we misread AI as a person (trusting it too much or fearing it), leading to underuse.

Organizational paradoxes

Beyond the individual, there are some general paradoxes that are widespread in most organizations.

· Procrastination - makes us push work downstream; deadlines become the outer limit - and do you know anyone who submits their taxes early or always wrote their essay weeks before submission? We will do anything to delay work – coffee and biscuit distractions are always calling!

· Paralysis by analysis – this turns pedantry into delay; and overwriting displaces execution and getting things done.

· The busyness paradox - simply keeping ourselves busy and pretend that by being busy is productive. Calendars fill up with meetings and performative effort. A variation on this is Parkinson’s Law, where slack time expands process; tasks inflate and things generally expand to fill out the times in schedules.

· The boiled frog paradox – where organizations gradually deteriorate in productivity, refuse to adapt and use obsolete processes which persist until crisis forces change, or even the cut of a department, layer of management or the closing of the entire organization.

Technological paradoxes

· The Solow paradox - explained why productivity gains don’t show up quickly; measurement lags reality and sometimes slows adoption itself. There is often an infrastructure lag, as there was with the internet but with AI the internet and devices are already there.

· Legacy-systems lock-in – a more problematic paradox where old IT architectures prevent more productive technologies like AI from being adopted. AI may even add to the problem by not being integrated with legacy systems, amplifying bureaucracy rather than removing it.

· Moravec’s paradox – a more technical paradox where what seems easy is hard and the hard is easy. This may limit progress, in say embodied AI in robots, which is fiendishly difficult.

Economic paradoxes

Some paradoxes have been recognized in economics at a larger scale.

· Jevons paradox – explains that efficiency increases usage, so that saved time becomes more demand, more scope, more work, not fewer hours.

· Friedman’s make-work paradox - points to the often-stated brake on productivity, where workers may use spades to dig huge ditches to create employment. Friedman’s famous retort was – why not use spoons. In other words, incentives can prefer employment theatre over efficiency; institutions preserve labor to simply protect roles and budgets.

· The Easterlin paradox – states that higher output doesn’t guarantee wellbeing. Chasing productivity without meaning can erode motivation and engagement (and thus performance). This is why it is important not to be too cut and dry on productivity. There are human dimensions that need to be considered when driving for pure economic productivity.

· The Pollyanna paradox - when techno-utopian optimism backfires and breeds complacency. People plan for magic, neglect implementation and underinvest in the hard, boring work of deployment.

The impact of the paradoxes

We’ve taught rocks to think, reason and research at blistering speed; software agents can line up tasks, hand work off and glue processes together. Yet the bottlenecks are still there. Human attention is finite and worse, our institutions are rigid. We still run countries, companies and classrooms with primitive technology built for a slower century. So exponential capability collides with linear procedure. The result is not a productivity miracle but a new kind of drag with delays, compliance bloat and a timid culture that calls itself ‘responsible’ while quietly resisting change.

Taken together, these paradoxes act like a full set of brakes. They blunt learning, stall adaptation and reduce world-class tools to glossy hyped tech. The uncomfortable truth is that productivity doesn’t live inside the technology. It emerges from the whole setup, how people think, how work is organized, what infrastructure exists, what gets measured and what gets rewarded, all aligned to turn technology into real results.

Technological escapism vs productive intelligence

The internet promised so much, but has it delivered?

I would argue it has delivered one thing in spades – technological escapism. Its focus on the attention economy with consumerism built on selling us ads, fractious social media, filters, selfies, online gambling etc. is now giving way to something far more useful and productive.

AI promises abundance, agency, rapid gains in healthcare and education, robots, accelerated research and intelligence on tap. This is an altogether more pleasing future.

So, what do we do about AI adoption?

Refusing AI outright is as naïve as worshipping it. Resistance is futile. It is not a fad you can ignore, as it is already threaded through workplaces and daily life. The real question is not whether it goes back in the box, but who sets the terms of its use.

· For governments, that means planning for the churn in jobs before it becomes a crisis. This needs serious reskilling at scale, safety nets that work at speed and tax systems that do not collapse if wages stop being the main tap.

· For organizations, it means making adoption a strategy, not a series of irrelevant pilots. It needs clear direction from the top, relentless experimentation from the front line and no lingering in the bureaucratic swamp where committees debate while competitors automate.

· For individuals, it means upgrading your role with more validation, checking, challenging, improving and signing off AI outputs with judgement. This is what I call ‘Experts-in-the-loop’.

None of this works if we pretend it will just happen or may go away of its own accord. If democracy has a job here, it is to stop an era of abundance turning into an oligarchic era. The benefits of productivity must land broadly, not narrowly.

AI is already removing some roles, bending others, and inventing new ones. It can trigger shocks, but it can also deliver astonishing gains, especially in healthcare and education. I’m betting on the better path where the benefits improve life for us all.

Save 30% on AI and Productivity with code AHR30.