Skip to content
3 role hubs · 8 tasksData privacy

AI consultants for small business: what to sort out first, and what to ask before you sign

Most small businesses can try AI on one back-office task without a consultant. If you do hire one, agree the task, the data they will see, the contract and who checks the output before any work starts.

Short answer: if you can name the task, the data it touches and the person who will check the result, start on your own. Bring in a consultant when you can’t answer those three questions or don’t have the hours to set things up, and when you do, treat them as someone who will handle your data.

Do you need an AI consultant, or one well-chosen task?

Most people skip this question. They go straight to “who should we hire?” when the more useful question is “which repetitive job would we most like to hand over?”

Pick the task first. Look for something repetitive, with a clear right answer, that a person can check in a few minutes. A first draft of a policy, tidying a messy spreadsheet layout and writing the formulas for a reconciliation are all good candidates.

If you can describe the task in one sentence, you probably don’t need a consultant yet. If you can’t, a short conversation may help, and that conversation doesn’t need to come with a contract. AI for small business: start with the back office goes through how to choose a first task, and choosing AI tools by task and data covers the same ground from the tool side.

What a consultant can and cannot do that you can’t do in an afternoon

A good consultant can save you set-up time. They can look at how your team actually works, suggest where AI fits, build prompts and templates, and show people how to use them. That is real work, and not every small team has the spare hours.

What they can’t do is own your numbers or your decisions. The figures come from the system of record (the ledger, the HR system, the bank). AI drafts words, structure and formulas. A person in your business checks and approves. A consultant who talks as though the AI will “just handle” the figures is describing a different arrangement, and one you should question.

They also can’t take on your responsibility for risk. Under UK GDPR that stays with your business, as the next sections explain.

Sort these four things out before the first call

You can do all four in an afternoon, and having them on a page makes the first call shorter and more useful.

  1. The task. One sentence, written down.
  2. The data. What would the consultant need to see, and does any of it identify a real person?
  3. The contract. Who writes it, and does it say what happens to your data during and after the work?
  4. The checker. A named person in your business who reads the output before anyone relies on it.

Without these, the first call tends to become a sales conversation where the consultant defines the problem for you.

Questions to ask about your data (and why they matter under UK GDPR)

Start here, because this is where small firms are most exposed.

A consultant who handles personal data on your behalf is generally acting as your processor, and you are the controller. The ICO’s guidance on contracts says that whenever a controller uses a processor, there must be a written contract (or other legal act) in place. It also says controllers must only use processors that can give sufficient guarantees they will protect the data properly, and that controllers remain primarily responsible for overall compliance. The detail is in the ICO’s guidance on contracts between controllers and processors. The ICO notes it is reviewing that guidance after the Data (Use and Access) Act, so read the current version.

That turns into practical questions:

  • What exactly will you see, and can the task work with less?
  • Where will our data be held, and who else can access it?
  • Will any tool you use keep our data or use it to train a model?
  • What happens to our data when the work ends?
  • Will you use anyone else (a sub-processor) to do the work?
  • Can all of that go in the written contract?

The ICO’s guidance on AI and data protection covers accountability, lawfulness, fairness and data minimisation when AI is used with personal data. Data minimisation is the most useful principle for a small business: personal data should be adequate, relevant and limited to what is necessary. If the task works with names and identifiers stripped out, strip them out before anything is shared.

If the work will involve personal data, also ask whether a data protection impact assessment is needed. The ICO says a DPIA must be done before processing begins where it is likely to result in a high risk, and its list of high-risk examples includes AI combined with another risk factor. Use the ICO’s DPIA guidance to screen your own task. If data protection is new to you, the ICO’s advice for small organisations is a sensible place to start.

Questions to ask about the work itself

  • What will you deliver, and what will it look like when it’s done?
  • Can you give us a written scope and a total price?
  • What do you need from our team, and roughly how many hours?
  • How will errors in the output be caught, and by whom?
  • What will our staff be able to do without you afterwards?

The last question matters most. If the honest answer is “not much”, you are buying ongoing dependence on the consultant. That can be a reasonable choice, but make it deliberately.

Prices vary widely, so we don’t quote any. Get the scope and total in writing and compare proposals on the same terms.

Warning signs in a proposal

  • The task is vague, but the technology is described in detail.
  • Your data gets little or no mention.
  • It promises the AI will replace a role or “guarantees” accuracy.
  • It doesn’t say who checks the output.
  • No contract is mentioned at all.
  • You are pushed to sign quickly.
  • It asks for broad access to your systems when the task needs only a small slice of data. That runs against data minimisation, so ask why.

Two or three of these in one proposal is a reason to slow down or walk away.

Who signs off the output in your business

This part always needs a person inside your business. Neither the consultant nor the tool can do it.

Choose someone who knows the subject well enough to spot a wrong answer. For a reconciliation formula, that’s someone in finance (see AI for bank reconciliation). For a draft policy, it’s someone in HR or a manager who knows the rules (see AI for HR policies). For everyday admin, it’s usually whoever runs the office (see AI for office managers). The AI for pages cover more roles and tasks.

Write down the person’s name and what “checked” means for this task, for example “totals agree to the ledger” or “policy wording reviewed against current Acas guidance”. That turns review into a defined step.

A short checklist to take to the first call

Put these on one page, in your own words:

  • The task in one sentence.
  • The data the consultant would see, and whether it identifies real people.
  • A request for a written contract covering data use, storage, sub-processors and deletion.
  • A request for a written scope and total price.
  • The name of the person who checks the output, and what checking means.

If a consultant can work comfortably within that page, you have a sound starting point.

Questions

Questions people ask. Every answer open.

Short answers, drawn from this post. The sources are listed above.

Something missing, or out of date? Tell the author.

Email hello@usingaias.com

This post 5

What does an AI consultant actually do for a small business?
Usually they help you pick tasks, set up tools, prompts and templates, and train your staff. They should not own your numbers or decisions. Those come from your ledger, HR system or bank, and a person in your business checks and approves anything the AI drafts.
Is an AI consultant a data processor under UK GDPR, and do I need a contract?
If they process personal data on your behalf, they are generally acting as your processor. The ICO says that whenever a controller uses a processor, there must be a written contract (or other legal act) in place. As the controller, you must only use processors that can give sufficient guarantees, and you remain primarily responsible for overall compliance. The ICO notes that this guidance is under review after the Data (Use and Access) Act, so check the current version.
Do I need a data protection impact assessment before a consultant sets up an AI tool?
Possibly. The ICO says you must do a DPIA before you begin processing that is likely to result in a high risk to individuals. Its list of high-risk examples includes innovative technology such as AI when combined with another high-risk criterion. If the tool will only see anonymised or non-personal data, a DPIA is less likely to be needed. Use the ICO's screening checklists to decide for your own task.
How do I tell whether a small business needs a consultant or can start on its own?
Try to write down one task in one sentence, the data it uses and the name of the person who will check the result. If you can do all three, you can probably start on your own. If you can't, a short scoping conversation may help before you commit to a contract.
How much does an AI consultant cost?
Prices vary widely, and we don't quote figures. Ask each consultant for a written scope and a total price, then compare proposals on the same terms.