The Paradox of Progress: Why AI Makes Us Busier, Not Freer

Over the past three years I have repeatedly made a promise to audiences that I can no longer keep.

In almost every talk, workshop and AI accelerator I have run, I've something along these lines, "AI gives you time back to be more human." The idea that technology absorbs the dull, the dirty and the dangerous, liberating us for the work that only humans can do, sits at the heart of everything I do in the world of AI transformation. Liberate people from the routine. Return them to the irreplaceable.

It's a hopeful promise. It motivates teams to experiment, it gives boards a reason to invest. And most importantly, it gives the whole AI endeavour a sense of moral ethics beyond cost reduction. Inspirational, but right now, not realistic. In practice, I'm seeing something else is happening. In almost every organisation I work with, AI isn't making people more human. It's making them more busy.

AI isn't making people more human. It's making them more busy.

Championing the flourishing of humans in AI-native organisations is my life's work. So when I kept seeing the opposite, I went looking for why. There is a 160-year-old paradox that explains it.

Jevons paradox and the illusion of progress

In 1865, William Stanley Jevons published an economic paper about coal production. James Watt's steam engine had been made more efficient, hence less coal per unit of work was required. The reasonable assumption was that Britain would burn less coal overall. But Jevons inspected the data and found something peculiar. Although the price per unit of coal production was decreasing, overall coal consumption was increasing. In simple terms, the more efficient the steam engine, the more coal Britain demanded.

The contrast is precise. If the steam engine needs to use less coal per job, and the number of jobs stay constant, coal consumption should fall. Although the mathematics is correct, the assumption is wrong. When the cost of doing work falls, the appetite for jobs expands. Things that were previously too expensive to attempt suddenly become viable. Cotton mills that had never afforded steam power now ran looms twenty-four hours a day. Foundries that had never justified the heat of a steam furnace began to smelt iron through the night. New industries emerged and old ones scaled. And the variable that everyone treated as a constant turned out to be the most elastic thing in the system. When technology makes a resource more efficient, the total consumption of that resource increased.

When technology makes a resource more efficient, the demand for of that resource increases, not decreases.

This is Jevons Paradox, an idea which holds across a remarkable range of contexts. For example;

• More fuel-efficient aircraft led to more flights, not fewer.

• LED lighting used a fraction of the electricity of older bulbs, yet global electricity consumption for lighting increased.

• Build a wider motorway to ease congestion. Within years, traffic fills it.
  

Resource efficiency does not reduce demand. It stimulates new appetite for the resource.

Nowhere is this 1865 paradox more visible than in the world of AI. At the outset, the promise of this technology was to make us more human. But in almost every business I am working with, what I am seeing is that teams that are more busy, not more human. Not yet anyway.

What is AI doing to the nature of work? In February this year, researchers from the Haas School of Business moved into a technology company to answer a seemingly simple question. What is AI doing to the nature of work? For eight months they observed forty workers going about their days. Their hypothesis was that due to the deployment of AI, they would find people with more time. What they found was the opposite. The workers were doing more because of AI, not less.

What they found was the opposite. The workers were doing more because of AI, not less.

Working faster. Taking on broader scope. Extending into hours that had previously been their own. Taking on work that nobody had asked them to do. The company offered AI tools and mandated nothing. Workers were choosing to do more because AI made doing more accessible, and in the researchers' own words, intrinsically rewarding. New tasks emerged that would never have been attempted before. Old ones expanded in scope and ambition. Long-forgotten idea lists suddenly resurfaced to be tackled with AI.

This is Jevons paradox running exactly as predicted. The resource became cheaper to use. Appetite grew to meet it.

What to do about being more busy

I'm aware that this is not the most motivating piece of thought leadership. But I am not on a motivational crusade, I am on an educational one. So here are some thoughts from a classically trained accountant who has segued into the beautifully complex world of AI.

When we adopt AI, we face a choice between two things. Efficiency vs effectiveness. Doing things right and doing the right things. AI has made organisations dramatically better at the first without real gains second. You can now do the wrong things even though you can do them very fast. But does that count as success?

A filter for ROI

The most useful thing my classical training gave me was a filter for ROI. In business, every activity must either generate revenue, reduce cost, or builds an asset either of those things in the future. If it does none of those things, stop! Its likely consumption dressed up as productivity.

If it does none of those things, stop! Its likely consumption dressed up as productivity.

Hence, before any AI initiative, I have my teams answer four questions from first principles.

1. What is the job this AI work is actually trying to do?

   
   

2. Does using AI increase revenue or decrease cost and can you clearly tell me how?

   
   

3. Does using AI build something that will generate value in the future?

   
   

4. Who benefits from our using AI and how do we know?

   
   

The answer to that last question must name a person, not a process. If your answer describes a workflow rather than a human being, you have not answered the question. This is also the question most teams never reach, because they stop at internal efficiency. For example, which customer gets something quicker or with better quality? The anthem to my work is that technology must be in the services of humans, not the other way around.

These four questions are the only thing that cuts through the feeling. If a team cannot answer all four, we do not proceed. The late Rabbi Jonathan Sacks said something that captures this moment.

The single most important distinction in life is to distinguish between an opportunity to be seized and a temptation to be resisted.

AI has given us more opportunities than any generation in history. It has also given us more temptations than any generation in history. The ROI filter is how you tell the difference.

AI is the greatest workplace temptation

Jevons Paradox describes what efficiency does to appetite. It does not determine what fills the expanded capacity. That is a leadership choice. In most organisations, that choice is not being made deliberately. Jevons is making it instead by filling the room with busyness dressed up as productivity. Jevons may explain why so many organisations are not seeing the returns on AI investment they expected.

AI may be the greatest temptation ever placed in front of a workforce because it makes busyness feel indistinguishable from purpose.

Back to my promise that AI make you more human. I still believe this is true, but only with strategic ROI choices and deliberate trade-offs. AI can return humans to the craft that is uniquely theirs. But it is not magic nor does it forgo the need for critical thinking. Leaders must critically decide what to do with freed capacity and hold that decision steady against the pressure to simply produce more.

Take care of your craft.