The 80/37 Gap: Why AI Lifts Productivity but Not Enterprise Value
The 80/37 gap is the distance between two numbers in McKinsey's 2026 global AI survey: 80% of respondents say AI improved their personal productivity, and only 37% attribute any EBIT impact to their organisation's use of it. Individual work got faster. Enterprise value did not follow. The reason is that value is not created inside the steps, where AI has mostly been deployed. It is created, and lost, between them.
Most AI has been pointed at the individual workflow step, where a person was already reasonably efficient. Making an efficient step faster produces a real personal gain and very little at the level of the business. The larger cost sits at the interfaces, where work passes from one team, function or system to the next, and nobody owns those. The forecasting team owns the forecast. The allocation team owns the allocation. The gap between them, where the forecast waits to be reconciled, is unowned and unmeasured. That is where the 37% leaks away.
This post breaks down the numbers behind the gap. It is a chunk of the wider argument in Who owns the wait?, our field note on the coordination tax in regulated work.
The value is between the steps, not inside them
McKinsey puts the total at 35 to 60% of knowledge-work time spent on coordination: aligning, verifying, reconciling, waiting and handing over. That is not a rounding error on top of the real work. In many knowledge functions it is a larger share of the day than the work everyone considers their job.
The finding matches what we have argued from two other directions. Tools plateau because context never carries between them, which is the case we make in the company brain. A prompt improves a step and not a workflow, which is the case we make in the operating model. McKinsey's interface tax is the cost of that plateau, measured. AI has been pointed at the boxes on the process map. The cost is in the arrows.
What the coordination tax is made of
The coordination tax has three parts, and only the smallest one shows up on the income statement.
- Visible: about a third of the total. People and structures dedicated to coordination: planning teams, PMOs, governance forums, review meetings. Typically 3 to 5% of revenue in industrials. This is the part cost programmes already target because it is the part they can see.
- Latency: hidden. Cycle time and working capital. Every day a handoff waits carries a cost: inventory held longer, receivables collected later, capacity committed but not yet earning. It is felt rather than counted.
- Foregone: largest and least visible. Revenue that never arrived because the organisation could not coordinate fast enough. Launches delayed by handoffs, share lost to slow pricing decisions. Nothing on any statement records it, so nothing in the business is accountable for it.
The consequence is uncomfortable. A cost programme aimed at coordination will find the visible headcount, cut some of it, and report a saving. It will leave the latency and the foregone revenue untouched because they never appeared in the model. The tax that matters is the one nobody is measuring.
Waiting beats working, nine to thirty-six times over
McKinsey took one workflow apart at a disguised manufacturer and timed each interface: how long the work waited to be picked up, against how long the work itself took once someone started.
| Interface | Waiting | Working |
|---|---|---|
| Sensing to planning | 1-2 days | 2-4 hours |
| Planning to supply | 3-5 days | 4-8 hours |
| Supply to scheduling | 1-2 days | 2-4 hours |
| Scheduling to procurement | 1-3 days | 1-2 hours |
| Procurement to execution | 2-5 days | 1-2 hours |
| Execution to fulfilment | 1 day | 2-4 hours |
| Total | 9-18 days | 12-24 hours |
Read down the two columns. The work is measured in hours. The waiting is measured in days. Across the whole workflow, waiting beats working by roughly nine to thirty-six times. McKinsey estimated the total interface tax for this single workflow at $140m to $240m a year, with the visible portion under 7% of it.
That last figure is the whole point of the table. Cost programmes cut the 7% they can see. The other 93% is waiting, sitting in the arrows between the boxes, invisible to the income statement and unowned by any team.
What this means for a regulated firm
The manufacturer is a clean illustration, but the shape is general. Substitute a fund accountant for the scheduler and a partner for the plant manager and the arithmetic holds. A reconciliation that takes an hour of real work waits days for a checker. A client letter that takes an afternoon to draft waits a week for a reviewer who is really waiting for a paper trail. The days of latency around the hours of work are the tax, and in a regulated firm they have a specific cause we take up elsewhere.
The response is not another cost programme aimed at the visible 7%. It is to put something at the interface that can do the routine verification, reconciliation and routing that keeps the work waiting. What that something does at a handoff is set out in Verify, reconcile, route. Why, in regulated work, the wait is specifically a wait for evidence is set out in the coordination tax is a proof tax.
The gap between 80% and 37% is not a sign that AI does not work. It is a sign that it has been aimed at the wrong target. The productivity is real. The value is waiting in the arrows.
Which handoff in your firm waits longest? Download the field note, or tell us where the work waits and we will tell you whether it is a proof problem.
Figures and the industrial example are from McKinsey & Company, "Cutting the 'coordination tax': how agentic AI can reshape workflows", Industrials Practice, September 2026.