Who Owns the Wait? The Coordination Tax in Regulated Work

Individual productivity is up almost everywhere. Enterprise value is not. In McKinsey's 2026 global survey, 80% of firms said AI improved personal productivity, but only 37% could attribute any EBIT impact to it. The gap is not inside the work. It is between the steps.

Most AI has been pointed at the boxes on the process map, the individual steps where a person was already reasonably efficient. 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.

This field note is about that gap: what the coordination tax is made of, why forty years of process work kept missing it, and why in regulated finance and legal it takes a specific shape that can be engineered out one interface at a time. You can download the full field note as a PDF, or read the argument below.

AI has been pointed at the boxes. The cost is in the arrows.

McKinsey put the coordination tax at 35 to 60% of knowledge-work time: aligning, verifying, reconciling, waiting and handing over. It has three parts, and cost programmes only see one of them: the visible headcount (planning teams, PMOs, governance forums), typically 3 to 5% of revenue. The larger parts, the latency in cycle time and working capital, and the revenue that never arrived because the organisation could not coordinate fast enough, sit off the income statement entirely. In one disguised manufacturer, waiting beat working by nine to thirty-six times across a single workflow.

We break down the numbers in The 80/37 gap: why AI lifts productivity but not enterprise value.

Every era of process work optimised the box and left the arrow alone

Lean and Six Sigma, business re-engineering, ERP, process management, RPA: five eras, each solving a real problem, each optimising the step and leaving the interface human. No previous technology could sit at the handoff and exercise the judgement to verify, reconcile and route work across competing constraints. That is McKinsey's point. Ours is narrower: in regulated work the judgement at the interface was rarely the hard part. Proving it was exercised is.

An agent at a handoff does three things: verify (does the upstream output meet the downstream requirement), reconcile (resolve discrepancies between competing constraints), and route (send work onward, escalating only what exceeds the confidence threshold). Those map one to one onto the controlled workflow we build.

More on this in Verify, reconcile, route: what an AI agent does at a handoff.

In regulated work, the coordination tax is a proof tax

In an industrial case the tax is throughput and working capital. In regulated finance and legal it is mostly proof. A handoff does not stall because nobody can make the judgement. It stalls because nobody can evidence that the upstream step was done correctly, by an authorised person, against an approved source, at a known time. The reviewer re-performs the work not from distrust, but because the record does not let them do anything else.

That changes the design requirement. Verify, reconcile and route are necessary but not sufficient. Each action has to emit its evidence as it runs: source, rule applied, who signed off, timestamp. If the audit trail is a report someone assembles afterwards, the handoff still waits for it. If it is a property of the system that executed the step, the next step can start. Most AI in regulated firms dies at the point where someone has to sign for it. The fix is not more review; it is a workflow that produces what the reviewer needed to see, at execution time, every time.

Read In regulated work, the coordination tax is a proof tax.

Weeks, not months: start with one interface

When McKinsey took one interface apart, the work split three ways: 65% routine verification against known rules, 25% constrained agent judgement, 10% people. Their programme redesigns the whole workflow over six to twelve months with a COO sponsor. Ours redesigns one interface, leaves the proof behind, and moves to the next.

The routine 65% can be built, evaluated and running on one interface in weeks, with the record attached. Before it runs untouched, it runs in parallel against the process it replaces, on real volume, until the difference is explainable line by line. Then the threshold moves. The connectors, definitions, permissions and evidence model from the first interface are reused by the second, so each build costs less than the last. That is the same compounding argument as the company brain.

The through-line

This is the same finding we have made from two other directions. The company brain argues that tools plateau because context never carries between them. The operating model for AI where the output has consequences argues that a prompt improves a step, not a workflow. McKinsey's interface tax is the cost of that plateau, measured. The value is between the steps, and what the steps are waiting for, in a regulated firm, is proof.

Which handoff in your firm waits longest? We map the interfaces in a fortnight, build the first one inside your stack, and prove it in parallel before it runs alone. 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.