What is the human cost hidden inside your AI business case?

The deck says the efficiency gains pay for themselves in fourteen months, and you signed it, and a quieter part of you is wondering what those numbers cost the people behind them. The meeting moved on. The slide had a clean payback curve and a green tick. And on the train home the figure that stayed with you was not the fourteen months. It was the face of the person whose role sat under one of those automated lines, and the sense that the model had quietly priced them at zero.

So here is the answer your case has not given you yet. The human cost of AI efficiency is the value that walks out of the building when the people who remain stop offering their best, and a standard efficiency case does not measure it because it was never asked to. The case counts hours removed. It does not count the engagement, the retention, and the discretionary effort that those hours were carrying. You felt the gap on the train. That feeling is data.

Why does a clean AI business case feel like it is hiding something?

Because it is, and not through anyone's bad faith. A business case is built to be defensible to a board, so it counts what is easy to count: roles, hours, licence fees, a payback curve. The people line becomes a number that only goes down. What the model cannot see is the effect of the change on everyone who stays. That is the part you felt and could not name on the train.

You are not alone in the unease. In BCG's 2026 survey of chief executives, 61% said their boards are rushing AI transformation, and around a third said their boards overestimate the human capabilities AI can replace. Read that twice. A third of the people running these companies believe the body governing them is wrong about the very assumption that carries the savings. That is the quieter part of you, sitting in a thousand other heads, rendered as a percentage.

What does the human cost of AI efficiency actually look like in the numbers?

It looks like disengagement, and disengagement is already the most expensive thing in your operation that nobody put on a slide. Gallup measures it every year. Employees who are not engaged or actively disengaged cost the world $8.9 trillion in lost productivity, around 9% of global GDP. Only 23% of people are engaged. The rest are present and producing a fraction of what they could. And in 2024 global engagement fell from 23% to 21%, only the second decline in twelve years, a two-point slide Gallup costs at $438 billion. The drag exists before you automate a single task. An efficiency programme that lands badly does not start from zero. It adds to that number.

Then there is the fear, and fear has a price the spreadsheet never books. PwC's 2025 Global Workforce Hopes and Fears Survey (a study of 49,843 workers across 48 countries) found that more than a third of the global workforce feels overwhelmed at least once a week, and that the people who saw AI as a threat to their security were markedly less likely to report it lifting their performance. Gartner's 1Q26 Global Labor Market Survey of 12,004 employees and managers across 40 countries puts a multiplier on it: employees with a positive outlook on AI are 3.4 times more likely to be highly productive. Read that as a cost. When a workforce reads its own future in your roadmap and recoils, you do not lose a few hours; you lose a multiple of them. The model booked the salary as a saving. It did not book the recoil as a cost.

What turns a clean case into a true one?

The category of fix here is not a better spreadsheet. It is treating the People pillar of your operating model as the binding constraint, the way a good engineer treats the weakest joist, rather than as the line with the most give. The evidence sits in the failure data. BCG finds 74% of companies struggle to achieve and scale value from AI, and only 26% have moved past proofs of concept to tangible value. Their advised split of resources is the tell: 10% on algorithms, 20% on technology and data, and 70% on people and process. Your case probably inverts that split. The slide weight sits on the technology, where only a tenth of the value lives. The bottleneck is no longer the technology. It is whether the people on the other side of the change are coherent enough to make it real.

There is a clue to that human bottleneck in the research on judgement under load. In a study of experienced parole judges, the share of favourable rulings fell across each session from roughly 65% toward zero, then snapped back to 65% after a food break. Same judges, same law, depleted minds. Decisions narrowed under strain. Push a depleted workforce through a transformation and you do not get the upside on the slide; you get the conservative, defensive, lights-on behaviour of people running on empty. The deeper autonomic research points the same way: higher heart rate variability (the natural beat-to-beat variation in your pulse, a marker of how well the nervous system regulates itself) tracks with stronger executive function and self-regulation, and it falls under sustained stress. Capacity is physical. A case that assumes steady human performance through upheaval is assuming a body that does not behave that way.

So make the human cost a line you commit to defend. Not to slow the case down, but to make its numbers honest enough to actually arrive.

1. Name the people effect as an assumption. Write the engagement and retention impact into the model as an explicit figure, sourced and visible, so the board signs the whole picture rather than a curated half.
2. Reweight the effort to match where value lands. Move the plan toward the 70% that BCG places on people and process. Budget the change capability, not only the build.
3. Protect the people who carry the discretionary effort. Identify who holds the institutional knowledge and the goodwill, and design their role in the new model first, so they read the roadmap as a place that holds a future for them.
4. Re-time the payback to human capacity. Stage the change to a rhythm a depleted workforce can sustain, so you collect the upside instead of the defensive behaviour that swallows it.

Do this and the fourteen months stops being a hope dressed as a forecast. It becomes a number you can stand behind, because the people it depends on are inside it. That is the toward move available to you here: a case that is both efficient and true, signed by a leader who already knows what it costs and has chosen to pay it on purpose.

About the author

Thomas Green

British technology futurist, AI keynote speaker and advisor. Thirty years across enterprise technology and AI strategy, helping leaders navigate the future of work. The futurist who died.