The Verifier's Discount

The Verifier's Discount is what happens to the value of human work when the machine produces and the human checks: the accountability stays with the person while the pay and the status are repriced downward.

The mechanism is subtle because the verifying is real work, and frequently harder than producing. It is also invisible in the output. Nobody can see the error that was caught, only the document that was fine, so organisations end up paying least for the work they depend on most.

SuperSkills uses the term to describe this pattern, without a claim of first use.

Why verification gets repriced#

Producing is visible and checking is not. A person who writes a report has produced something. A person who catches an error in it has produced nothing anyone can point at. Reward systems track artefacts, so one of these gets promoted.

It looks like administration. Reviewing has the surface features of low-status work: reactive, procedural, done to someone else's output, easily described as a sign-off step. That surface is misleading, and the misreading is expensive.

The failure is invisible until it isn't. An organisation that under-resources verification looks identical to one that resources it properly, for as long as nothing goes wrong. The two only separate at the moment of a bad outcome, by which point the pricing decision was made years earlier.

Bainbridge's irony compounds it. Automating the routine leaves the human with monitoring, the hardest residue, while removing the practice that built the competence for it. The job gets harder and looks easier at the same time.

The claim this page actually makes#

Verification is not a separate activity from expertise. It is expertise, applied.

Knowing that an answer is subtly wrong, before you can articulate why, is a tacit judgement built from having done the work. That is why it cannot be delegated downward to someone junior with a checklist, and why buying more of it cheaply does not work: you are not buying a process, you are buying accumulated competence.

Which produces the pricing error. Organisations price verification as a task and it is a capability, and capabilities do not respond to procurement.

Where the evidence sits#

Direct evidence on how verification is compensated does not exist, and this page is an argument rather than a finding. The supporting evidence is about the difficulty of the work rather than its price.

The Vaccaro meta-analysis of 106 experiments found human-AI combinations underperforming the better party alone, with losses concentrated in decision tasks, which is the reviewing configuration. Yu and colleagues found the effect of AI assistance on radiologists running from strongly positive to strongly negative between individuals, unpredicted by experience, so the quality of verification varies enormously between people doing nominally the same job. And Autor argues that AI's distinctive opportunity is to extend the reach of expertise, which only holds if the expertise is still there to extend.

What would settle it: wage and role data showing whether verification-heavy roles are being repriced relative to production roles. Nobody publishes it.

It is now a compliance question too#

Article 14 of the EU AI Act, in force since 2 August 2026, requires that people overseeing high-risk systems can detect anomalies, remain aware of automation bias, interpret output correctly, and disregard or override the system.

Those are capability requirements attached to named individuals. An organisation that has priced verification as administration, staffed it accordingly, and recorded it as a control has a documentation problem as well as a capability one. See meaningful human oversight.

Price it as skilled work#

• Pay for it as skilled work. While it is priced as residue you will keep getting the amount of it that residue buys.
• Put time in the plan. Verification with no allocated time does not happen under pressure, which is when it matters.
• Apply the capability test. Could the person verifying this have produced it themselves, well enough to notice if it were wrong? If not, record the control as absent rather than satisfied.
• Count the catches. Organisations count output and not errors caught, so one is visible and the other is not. A simple log changes the conversation.
• Never describe sign-off as verification. Sign-off accepts accountability for an outcome. Verification establishes whether the content is correct. A senior person can do the first without being able to do the second.

Put it to work#

The operational version, stage by stage with a downloadable working grid, is the Delegation Boundary Map. It sets the verification requirement for every stage, including the capability test.

The ownership question that follows, and which in most organisations has no answer, is who owns verification when AI does the work? The control-failure version is who supervises work they cannot do themselves?

Related SuperSkills research#

On why review positions fail, human in the loop is not a safeguard. On the underlying erosion, capability debt and deskilling. On what makes verification possible, tacit knowledge. On the pattern, drift versus design.

Key sources

• Vaccaro, M., Almaatouq, A. and Malone, T. (2024). When combinations of humans and AI are useful.
• Yu, F. et al. (2024). Heterogeneity and predictors of the effects of AI assistance on radiologists.
• Bainbridge, L. (1983). Ironies of Automation.
• Autor, D. (2024). Applying AI to Rebuild Middle Class Jobs.

About this research#

Written by Rahim Hirji, author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company.

How this research works · Reviewed quarterly · Found an error? Tell me and it is corrected on the page.