Skip to content
Join 7,000+ leaders following Alastair's work on LinkedIn.

The Verification Dial: A Simple Rule for When to Check AI's Work

AI Essentials
Updated 8th August 2026 Published 14th July 2026

The question I hear in every workshop

"If I have to check everything the AI produces, where's the saving?"

It's a fair question. The Verification Tax - the time you spend checking AI output is a tax on every hour it saves you. And if you're checking every word of everything, forever, the tax rate is so high you might wonder why you bothered.

You already know how to solve this. Your auditor solved it decades ago.

An auditor doesn't check every transaction in your ledger. They sample, and the intensity of the sampling is set by risk. Nobody calls that careless - it's the professional standard for managing exactly this problem at scale. We can borrow the same move for AI. The Verification Dial runs on four rules.

Rule 1: stakes set the floor. Ask yourself how bad an undetected error would be, and whether anyone would catch it before it does damage. For high-stakes work, the dial is welded at 100% and it never moves. A solicitor filing a submission to court checks every citation on the thousandth AI draft, exactly as she did on the first. That's not a failure of the technology. That's the job.

Rule 2: track record moves the dial. For routine work, you start at 100% and earn your way down. Say you use AI every Monday to turn your rough meeting notes into clean minutes. For the first few weeks, you read every line. After a month of clean output, you check every second one. A month later, one in five. You're not trusting blindly - you're sampling, like the auditor.

Rule 3: the dial belongs to the workflow, not the tool. This is where people get burned. Six months of happy experience with a chatbot tells you nothing about a brand new task. A finance manager who has dialled back to spot checks on expense categorisation still goes back to 100% the first time she asks the same tool to draft board report figures. New task shape, new dial, back to the start.

Rule 4: any error resets the dial. Find one mistake in a spot check and you're back to checking everything, earning the reduction all over again. Harsh? Maybe. But it's the rule that lets you sleep at night.

One caveat: spot checks only work if they're honest. Random, on a schedule - and if an output looks dodgy, checking it doesn't count as your sample. (That's just common sense wearing a lab coat.)

You never eliminate the Verification Tax. You bring the rate down - to nearly nothing on the routine work where the hours actually live, while it stays high on the work that deserves it.

The one-page version: The Verification Dial in our Framework Library - the four rules on a single card you can keep to hand.

Related Articles

// ai essentials · aug 2, 2026

AI Skills for Teams

Why your team gets five different answers from the same AI If you manage a team that has started using AI, you have probably noticed something odd. Everyone has access to the same tool. Everyone is doing broadly the same job with it. And the output varies wildly depending on who typed the request. One […]
→ read more
// ai essentials · jul 19, 2026

Everyone's calling it brain rot. (It's not)

Oxford University Press named "brain rot" its Word of the Year in late 2024 with more than 37,000 people voting. Usage of "brain rot" had risen 230% in twelve months. The definition Oxford gave was "the supposed deterioration of a person's mental or intellectual state, especially viewed as the result of overconsumption of material (now […]
→ read more
// ai essentials · jul 16, 2026

Where Does It Run?

A question we've all answered before Every manager has answered this question already, just not about AI. When you bring in a contractor, you decide: do they work on our site or theirs? With our equipment or theirs? Nobody leaves that to chance, because the answer settles who is responsible for what, what the contractor […]
→ read more
// ai essentials · jul 16, 2026

Not All Cloud AI Is the Same: The Three Tiers of AI Privacy

The question that stalls every AI rollout Somewhere in every workshop, someone asks whether it is safe to put their data into AI. It is the right worry and the wrong question, because "AI" is not one place. The convenience of a cloud tool and the strictness of GDPR only feel like opposites while you […]
→ read more

Is your business AI ready?

  • Get honest, practical AI advice
  • Find out where AI saves the most time
  • If we're not a fit, I'll point you somewhere useful
Alastair McDermott

25 mins · Free · No obligation

Book a Focus Call