TL;DR. Most small businesses should start without a consultant: pick one task, try the AI already inside your tools on real examples, and measure it. Hire help when the task spans several systems, touches sensitive data or money, or the cheap option fails on your real cases. Buy one fixed-price step with numbers at the end. Skip strategy decks, long retainers and anything tested only on clean sample data.

AI consulting for a small business should mean one thing: getting one task that eats your week done faster, on your real data, at a cost you can see. Not a strategy. Not a transformation. One task, measured before and after, with a clear price for the next step. Often you can get there without a consultant at all, and I'd try that first.

I run Skcript Applied Labs, which sells AI consulting. So take my advice to try it yourself first as an honest one: it is what I would tell a friend.

Does a small business actually need AI consulting?

Not always. A lot of useful AI now ships inside software you already pay for. Your email, your accounting tool, your helpdesk and your office suite probably have an AI feature switched off somewhere. For simple jobs, that is enough.

You need outside help when one of these is true:

  • The task crosses systems. Reading an email, checking it against the accounting tool and updating an order in a third place is integration work, not a setting.
  • The data is sensitive. Customer records, health data, payroll or contracts need care about where data goes and who can see it.
  • Mistakes cost money. Payments, quotes, refunds and anything a customer relies on need a person approving the right steps and a test that proves the AI is good enough.
  • The cheap option failed. You tried the built-in feature or a popular tool and it did not hold up on your messy real cases.

If none of those apply, save the money. If you are weighing a product against outside help, I go through when to buy, use a no-code tool or do nothing in do you need an AI consultant.

Which task should a small business automate first?

The best first task has four traits. It happens often, every day or every week. It is heavy on text or documents. It takes real hours from someone whose time is worth more elsewhere. And you can tell a right answer from a wrong one.

Common first picks in small businesses:

  • Supplier invoices and receipts. Read them, match them to orders, enter them in the accounting tool.
  • Customer emails. Sort them, pull out the order number or the problem, draft a reply for a person to send.
  • Quotes and order entry. Turn an emailed request into a quote or an order in your system.
  • Onboarding paperwork. Collect and check documents from new customers or staff.

Bad first picks: anything that happens a few times a month, anything where the rules can simply be written down, and anything where one mistake could seriously hurt a customer. A rule or a better form is cheaper and easier to check. I keep a longer list in when not to use AI.

What can you try yourself before hiring anyone?

A week of your own effort can save you a consulting bill, or make the one you pay much smaller.

  1. Write the task down. What comes in, what has to happen, what goes out, and who does it today.
  2. Measure today. How many a week, how long each one takes, how often something goes wrong.
  3. Collect 30 to 50 real examples. Include the messy ones: bad scans, half-filled forms, the email in two languages. Write down the right answer for each.
  4. Try the AI you already have. On a business plan that does not train on your data. Run your examples through it and count the right answers.
  5. Decide. If it gets nearly all of them right and fits how your team works, you are done. If not, you now have exactly what a good consultant will ask for on day one.

That set of examples with the right answers is worth keeping either way. It lets you check any tool or vendor later on your own work instead of their demo. We explain how we test models this way in how we test AI models on your work.

What should AI consulting for a small business include?

If you do hire help, the first engagement should be small and end in a decision. Look for these in the proposal:

  • One task, named. Not "AI across the business."
  • A starting measurement. Time, volume, error rate and cost, written down before anything is built.
  • Tests on your real cases. Not a demo on sample data.
  • More than one model compared. On accuracy and cost per task, on the same cases.
  • Built inside your tools. Your team should not have to learn a new app for one task.
  • A person approves what matters. Anything the AI was not tested on goes to a person, as it does today.
  • A fixed price for the next step, or a plain recommendation to stop.
  • You own it. The code, the test cases and the numbers.

How much should a small business spend on AI consulting?

Enough to answer one question well: does AI do this task on our real cases, and what will it cost to run? Not more.

The simple test is payback. Take the hours the task eats each week, put a cost on them, and ask how long the time saved would take to cover the first step plus a year of running costs. If the answer is years, pick a bigger task or skip it.

Avoid three things on a small budget. Open-ended discovery with no end date. A monthly retainer before anything is live. And a price that covers a demo but not the work to make it real. I explain what drives AI consulting prices, and the costs nobody quotes, in how much does AI consulting cost in 2026. Our own fixed prices are on the AI consulting pricing page.

What are the risks for a small business, and how do you guard against them?

  • Customer data in the wrong place. Consumer AI accounts can use your data in ways business plans do not. Use business plans, and check where the data is processed.
  • Lock-in. Build so you can switch models. A test set from your own work makes switching a measured decision, not a guess.
  • Nobody owns it. Name one person who knows the task and checks the results. Without an owner, the system drifts and people stop trusting it.
  • Paying for a demo. Ask whether what you pay for now carries on to production, or gets rebuilt.

What should a small business skip?

  • AI strategy decks. A small business has one or two tasks worth automating, not a portfolio. You don't need a roadmap to find them.
  • Readiness assessments with no build. If the output is a score and another proposal, you have paid to be sold to.
  • Custom chatbots for the website unless customer questions really are eating someone's day. Most small sites get a handful a week.
  • Annual contracts before a test. Run any product on your own examples first, on a monthly plan if you can.
  • Training your own model. The leading models on business plans are good enough for almost every small-business task. The work is in the task, the test and the fit, not the model.

How do you choose an AI consultant as a small business?

Look for a small senior team where the person on the first call does the work. Ask which number they will move, which models they will test and on what, and what you keep if you stop. Walk away from anyone who starts with a strategy deck before seeing your task. The full checklist is in how to choose an AI consultant.

What does a good first month look like?

Week one: pick the task, measure it, collect real examples. Week two: test the AI you already have, and decide whether you need help. Weeks three and four, if you do: one fixed-price step on that one task, ending with before and after numbers and a price to go live.

That is roughly how our 10-day AI pilot works, compressed into ten working days with our engineers doing the building. If you want to see the whole offer, it is on our AI consulting page. Bring one task and a few real examples, and we'll tell you honestly whether it is worth doing.