Our approachGuidanceCoaching ProgramsPartnersAbout
SubscribeTake the Capability Snapshot
Guidance›Digital & AI›How can I use AI to decide what to do next in my business?
Digital & AI

Part of the guide: AI for small business

How can I use AI to decide what to do next in my business?

Use AI to widen and pressure-test your options, then decide yourself. It's reliable for gathering what you already know into one place, naming trade-offs you have missed and arguing against your preferred answer. It is unreliable at knowing which constraint is actually holding the business back, because that depends on context it can't actually see.

By Samantha Hurley7 minute readUpdated November 2026

The short answer

AI is useful where it has strengths and poor where it doesn't.

It's useful for laying out the options, for finding the trade-offs and second-order effects you haven't thought of, for arguing the opposite case to the one you favour, and for turning a decision you've made into something written down that the business can follow.

It's poor at telling you which problem to solve. That requires knowing what is actually happening in your business, which parts are load-bearing, and what you want the business to make possible. None of that is in the model unless you put it there, and fr a lot of businesses, most of it isn't written down anywhere.

So a good sequence to follow is: you choose the question, AI widens and stress-tests the thinking, you make the call.

Why AI is good at some of this and poor at the rest

The reason is worth understanding, because it tells you where to draw the line.

A model is strong at producing and recombining possibilities. It has read a great deal about how businesses work, so asking it for the ways a pricing change could go wrong will reliably return several you hadn't considered before. That is genuine value in a matter of minutes.

It's weak at judgement though, because judgement requires knowing which of those possibilities matters in your specific situation. It will also produce confident reasoning for a bad option if you've framed the question badly, which can cost owners real money if it's put into action.

There is a measured version of this trade. In a 2026 MIT field experiment, Ju and Aral found that people working with an AI agent produced about 50 per cent more output and higher-quality text, while the output became more homogeneous and human review of quality and diversity was the thing that constrained it. The mechanism is the useful part: AI raises throughput and narrows variety, which means the reviewing is the actual real work.

Applied to deciding what to do next, that means the value is in your questions and your review, and the risk is in accepting the first well-written answer you're given.

A four-step way to use AI on a real decision

Pick one decision currently taking up your attention. Then work through this.

1. Write the decision down as a decision

Most decisions arrive as a feeling. "We need more leads." "I should hire." Write it instead as a choice between named options, with what each one would require and what you would have to stop doing.

If you can't write it as a choice, that in itself is the finding. You have a symptom rather than a decision, and the next step is to diagnose rather than use AI. How do I identify the real constraint in my business? is the method for that part.

2. Give it your context before you ask the question

The difference between a useless answer and a useful one is almost always context. Paste in the real numbers, the actual constraint, the size of the team, what you have already tried, and what you want the business to look like in two years.

This is the point at which owners discover the context isn't written down anywhere. That's one of the reasons we support business owners to build a Business Brain: once the business keeps its own pricing logic, standards and decision rules in one place, every AI conversation starts from your business rather than from a generic one.

3. Ask it to argue against you

This is the step that earns the most and we don't often do (probably because we like how good it feels to get complementary answers all of the time!). Give it your preferred option and ask for the strongest case against it, the conditions under which it fails, and what someone who disagreed with you would point at.

Then ask what information would settle the disagreement. That question turns a conversation into a small test you can actually run.

4. Decide, record the reasoning, set a review date

Write down the decision, the reason, what you expect to happen and when you will check. The checking is what turns a decision into learning.

The questions that do the work

These are the plain language prompts worth keeping.

  • Here is my situation and my numbers. What am I treating as fixed that might not be?
  • If this works, what breaks first?
  • What would have to be true for this to be the right decision?
  • What are three options I have not considered, and what would each cost me?
  • Argue the opposite of what I have just told you, as strongly as you can.
  • What is the smallest test that would tell me whether I am right?

Notice that none of them asks the model what to do. They ask it to widen the field and attack the reasoning, which is what it is good at.

What to keep human

Four things shouldn't be delegated, whatever the tool.

Which problem you are solving. The most visible problem is often a symptom of a constraint elsewhere, and a confident answer to the wrong question is worse than no answer.

Anything involving a person. Pay, performance, whether a role is working. Use AI to prepare, never to conclude.

Anything you can't check. If an answer contains a figure, a legal position or a tax consequence, verify it at the source. Treat the output as a draft from a well-read colleague who has never met your accountant.

The decision itself. Keeping the decision is also how you keep the capability. Each decision you work through yourself makes the next one faster, and that compounds in a way that outsourcing the thinking doesn't.

Where client data, staff data or anything confidential goes into a tool, decide the rules first. AI governance for a small business, without the bureaucracy covers the practical version, and responsible AI for a small business covers the wider picture.

When you need a person instead

AI is a poor substitute in certain situations. When you don't trust your own understanding of the problem is one, because the model will agree with whichever framing you give it. When the decision is one you have avoided for months is another, because the obstacle is not information. And when you need to be held to it is another one, because a tool has no stake in whether you follow through.

If the problem is working out where to start rather than how to execute, the AI Opportunity Mapping Session exists for exactly that. It is a 60-minute one to one that produces a documented map of where AI genuinely serves your business, and the fee is credited toward a program within 30 days where the program matches the recommendation.

Frequently asked questions

Can AI tell me what to do next in my business?

It can tell you what the options are and what each would involve, which is most of the useful work. It can't reliably tell you which one is right, because that depends on your constraint, your capacity and what you want the business to make possible.

What is the best AI tool for business planning?

Any of the current general-purpose assistants will do this well enough. The quality of the answer depends far more on the context you supply and the questions you ask than on which tool you choose. Start with what you already have.

How do I stop AI just agreeing with me?

Ask it to argue the opposite case explicitly, and ask what evidence would change the answer. A model will follow the framing it is given, so the framing is your job.

Should I give AI my real business numbers?

Decide the rule before the moment arrives. Many owners are comfortable with their own aggregate figures and not with customer or staff data. Write the rule down once, in plain language, and apply it consistently.

What should I do before using AI for decisions at all?

Make sure the business has its own context written down somewhere usable. Without it, every conversation starts from scratch and the answers stay generic. What order should a small business adopt AI in? sets out the sequence.

Is this worth the time for a small business?

For the kind of decision that has been sitting unresolved for weeks, half an hour of structured questioning usually moves it forward. Is AI worth it for a small business? gives the honest version of the wider question.

Where to go next

If you want the broader picture first, start with the AI for small business guide.

If you would rather find out which part of the business to work on before adding any tool, the Capability Snapshot takes under ten minutes and returns your growth stage, your priority capability gap and your AI readiness.

Find your first move with AI

A 60 minute AI Opportunity Mapping Session, with your documented map within 48 hours.

See the session