The Intelligence Gap

An answer can be articulate, detailed and persuasive while still being wrong for the circumstances. It may omit a constraint, rely on an assumption or recommend something the business cannot deliver.

The intelligence gap considered here is the distance between what AI can produce and the human judgement needed to make good use of it. The idea draws on two discussions in Steve Bolton’s Think With Me: the responsibility that remains with the person, and the thinking a person continues to practise when some work moves to AI.

An answer needs a purpose

Consider the business situations in the book: a proposed hire, a late-paying client, a competitor’s price change or a proposal due on Friday. AI can contribute research, options, drafts and scenarios. Each situation still needs a person to establish what matters.

A hiring analysis may compare salary with expected revenue. The owner also needs to examine the assumptions, the business’s capacity to support the role and the consequences if the forecast is wrong. A late payment may require a conversation informed by years of working together.

Judgement gives those decisions direction. It establishes the problem, the relevant context, the evidence worth trusting and the trade-offs that can be accepted. The quality of the language on the screen cannot settle those matters by itself.

Notice where the thinking goes

In The Mirror Effect, Steve examines what happens when people move mental work outside themselves. Lists, diaries and calculators already carry tasks that once depended more directly on memory or calculation. AI extends the choices about what to delegate.

Those choices change the work that remains. Receiving a draft may shift effort towards editing. Receiving several options may make comparison more important. Asking for objections creates value only if somebody examines the objections and decides what follows from them.

The book resists a simple claim that AI automatically makes people more or less capable. Its practical concern is which parts of thinking are still being exercised. Accepting an answer without judging it provides no practice in making that judgement.

Keep the consequential work visible

A useful review can make three things explicit: what is known, what is assumed and what remains to be decided. This creates a basis for discussion beyond whether the output sounds convincing.

For an ordinary proposal, that might mean identifying the client’s confirmed requirements, checking a delivery estimate with the responsible colleague and marking a question that still needs to be asked. AI can help organise those distinctions. Confirmation has to come from the relevant evidence or person.

It also helps to consider a serious alternative. What would someone who disagreed notice? Which assumption would change the recommendation? Where would a different question lead?

Capability that carries into the next decision

The result of a useful exchange can include a completed document. It can also include a clearer understanding of the problem and a better question to carry forward.

That second result deserves attention. Can the person explain the decision, defend the evidence and identify what could make it wrong? Can they recognise when another person’s expertise is needed, or when the work should remain entirely human?

For Steve, better human thinking is the opportunity. The gap begins to narrow when a capable system is met by people who continue to frame, question, verify and decide.