Do I Need an AI Consultant? A Five-Question Test That Takes 10 Minutes

You’ve got three AI subscriptions, a ChatGPT tab you open twice a week, and a 10 pm habit of typing “do I need an AI consultant” into Google because everyone else seems to have worked something out that you haven’t. Every result was written by someone who sells consulting.

Here’s the honest version. Some businesses need outside help to get AI working. Plenty don’t. They need a better tool, set up properly, and one free afternoon. What matters is knowing which group you’re in before you pay to find out the hard way.

This post gives you a five-question test that takes about ten minutes. At the end, you’ll know whether to buy a tool, get a scoped plan, or bring in someone to build the system and run it with you. You’ll also know what good help looks like, so you don’t pay for a report that ends up in a drawer.

Key Takeaways

• Do I need an AI consultant? Usually only if the problem crosses people and systems. A single repetitive task is a job for a tool.
• Founders who hire an AI consultant too early pay for a report. Founders who stick with tools too long pay in wasted months.
• If the knowledge an AI needs lives only in your head, no off-the-shelf tool will fix it. Someone has to get it out and structure it.
• When a tool stalls after week two, the tool usually wasn’t the problem. The workflow around it was, and that’s where help earns its fee.
• Score yourself on five yes or no questions. Zero or one yes: buy a tool. Two or three: get a scoped plan. Four or five: get a builder.
• The best AI consultant for a small business builds and supports the system instead of just advising on it. Advice on its own rarely ships.
• Ask who owns the system in month three. If the honest answer is you at 9 pm, you’ve bought yourself another job, not a solution.

Do I Need an AI Consultant, or Just a Better Tool?

Most founders I talk to treat this as a money question. Is it worth paying someone, or can I do it myself for twenty bucks a month? That’s the wrong way to look at it. The real question is what kind of problem you have. Tools and consultants fix different kinds of problems, and buying the wrong one is how businesses end up with a drawer full of logins and nothing to show for them.

What a better tool fixes

A tool fixes a task. Writing the first draft of a quote. Summarising a meeting. Turning a voice note into a job sheet. Cleaning up a messy spreadsheet. If the problem sits inside one person’s day, and starts and finishes in one place, a tool is usually enough. ChatGPT Plus or Claude Pro costs around US$20 a month, and for one repetitive task that could be the best twenty dollars you spend all year.

What an AI consultant fixes

A consultant, a good one anyway, fixes a decision. Which of the forty things you could automate should go first? What does it cost you to leave each one alone? Which tools work together and which will fight each other? That’s worth paying for when you’re stuck because you don’t know where to start, not because you’re short of software.

The catch is that plenty of consultants stop at the decision. You get a slide deck, a roadmap and a handshake. Then the roadmap goes into the same folder as the SOPs nobody ever opened.

What a builder fixes

The third option gets less airtime: someone who does the diagnosis, builds the system, and stays on to train it and fix it as the business changes. That’s the work I do, so take my view with that in mind. It fixes a different problem again, which is the way work moves between people, systems and your own head.

Most stuck businesses aren’t stuck on a task. They’re stuck on the hand-offs. The lead that came in on Friday that nobody owned. The quote that sat for three days waiting for your sign-off. The new hire who asks you the same twelve questions every week. A subscription can’t fix a hand-off on its own, because the tool doesn’t know who’s meant to do what next.

So before you decide anything, work out what kind of problem you’ve got. That’s what the test is for.

The Five-Question Test

Grab a pen and think of the one thing in your business that annoys you most. Not a general feeling of being busy. One specific thing that keeps going wrong or keeps landing on your plate. Run that problem through these five questions and count your yeses.

Question 1: Does the problem cross more than one person?

If the problem starts and ends with one person, it’s a task. If it passes through two or three people, or between your team and a customer, it’s a workflow.

Take a typical trades business. An enquiry comes in by phone. The office manager takes the details. You price the job. An apprentice books the site visit. Someone chases the deposit. An AI tool that writes quotes in half the time is nice, but it does nothing if the quote then sits in your inbox for three days because nobody else can approve it.

That’s the pattern behind most “AI didn’t work for us” stories. The tool did its bit. The hand-off on either side of it stayed broken.

Score a yes if your problem involves a hand-off between at least two people, or between a person and a customer.

Question 2: Have you already tried a tool and watched it stall?

Most founders have. Week one is exciting. By week three, it’s another tab. By month two you’re paying for something you’ve stopped opening.

That stall tells you something. It usually means the tool was never the issue. In my experience, it comes down to one of three causes. The tool didn’t know enough about your business to give useful answers. Nobody owned it. Or it sat off to the side, away from where the work actually happens, so using it meant extra steps nobody had time for. I’ve written more on why this happens in the guide to AI adoption for SMEs.

If you’ve been through that cycle once, a second tool on its own is likely to go the same way.

Score a yes if you’ve paid for at least one AI or automation tool that you no longer use, or use far less than you meant to.

Question 3: Does the knowledge the AI needs live in your head?

An AI is only as good as what it knows about your business. Your pricing logic. Which clients get the gentle approach. Why you never quote below a certain margin. Which supplier always runs late and needs chasing on a Wednesday.

If that’s written down somewhere, a tool can read it. If it lives in your head, someone has to get it out and shape it so the AI can use it. Founders tend to underestimate this part. It isn’t technical work. It’s interviewing, listening and writing things down in a structured way.

Can you do it yourself? In theory, yes. In practice, you’re the worst person to do it, because the things you know best feel too obvious to write down. You skip the steps you’ve done a thousand times, and those are exactly the steps the AI needs.

Score a yes if a new staff member would need to ask you more than a handful of questions in their first week to do the job properly.

Question 4: Does the fix need to touch more than one system?

Your CRM, your accounting software, your calendar, your inbox, your job management tool. If the fix has to read from one and write to another, it’s an integration problem, not a tool problem.

Tools connect to each other far more easily than they did a few years ago. Someone still has to set the rules, test the odd cases, and decide what happens when a customer gives half an address, or the calendar is double-booked. That design work decides whether the thing works on a Tuesday afternoon with real customers.

Here’s a real example. Dr Claire’s dental practice had two receptionists and was still missing close to half its calls. A better phone app wasn’t going to fix that. It needed AI that answers every call, works out what the patient needs, and puts the details in front of the team in the right place so nobody falls through the gaps. Once that was in, booked appointments went up 44%. That’s a systems fix, and it needed several systems talking to each other.

Score a yes if the fix needs at least two of your systems to share information.

Question 5: Who owns it in month three?

This is the question most people skip. Systems drift. Prices change, staff change, a supplier updates their software, and something quietly breaks. Someone has to notice, fix it and keep improving it.

If the honest answer is “me, at 9 pm, when I remember,” then you haven’t bought a solution. You’ve bought another job. It’s the same trap that turns every tool into an abandoned tab. Building it is the first step. Keeping it running week after week is what decides whether it was worth doing. If you’ve already got something running, the 90-day AI automation checkpoint covers what to review.

Score a yes if nobody on your team has the time and interest to look after an automated system every week.

What Your Score Says About Whether You Need an AI Consultant

Add up your yeses. Your answer to “do I need an AI consultant” depends far more on that number than on your industry, your size or how much you know about technology.

Zero or one yes: buy a better tool

Your problem is a task. Don’t hire anyone. I’d tell you the same thing on a call.

Pick the one task that eats the most time. Choose one tool. Give it a proper go: write a one-page brief about your business that you paste in or save as instructions, save the prompts that work, and use it every day for two weeks. Most people who say AI didn’t work for them gave it four half-hearted attempts with no context. Do it properly once, and you’ll know.

Two or three yeses: get a scoped plan before you buy anything else

This is where most founders land. The problem is bigger than a task, but you’re not sure how big. You need a diagnosis: someone to map how work moves through the business, rank what to automate by time saved and effort, and put a number on what doing nothing is costing you.

Keep it short and bounded. The output should name a first build in plain English, with a clear result and a rough timeline. If you finish with a forty-page document and no obvious first step, you’ve paid for homework. Before you sign anything, these questions for your accountant about an AI automation contract are worth ten minutes.

Four or five yeses: you need someone who builds and stays

Your problem isn’t a task or a decision. It’s how the business runs. Advice alone won’t land here, because you don’t have the bandwidth to act on it. That lack of bandwidth is probably why you scored this high in the first place.

What you need is someone who gets the first system live in weeks, then trains it, improves it and supports it as the business grows. Start small anyway. One bounded system that works beats a grand programme that’s always six weeks from launch. Each working system makes the next one easier, because the AI already knows your business and your team already trusts it.

Why the middle score is good news

If you scored two or three, don’t read that as being stuck in between. It means your problem is well defined and there’s an obvious first step. The businesses that struggle are the ones that score five and keep buying tools, or score one and hire a consultant to tell them what a free afternoon would have shown them.

What a Good AI Consultant for a Small Business Looks Like

If your score says you need help, the next risk is hiring the wrong kind. The market is crowded with people who learned the words last year.

The numbers back up the caution. MIT research reported by Fortune found that around 95% of generative AI pilots at companies delivered no measurable financial return. The ones that worked tended to start with a specific business process and fit the AI into it. The ones that failed started with the technology and went looking for a use.

That gives you a simple filter. Good help starts with your work. Bad help starts with their tools.

Green flags

They ask about your week before they ask about your software. What did you do on Monday? What took longest? What did you handle that you shouldn’t have had to?

They name a first build in plain English. Something like “every new enquiry gets a reply within two minutes and lands in your CRM with the job details filled in.” Not “we’ll implement an agentic framework.”

They price an outcome, not open-ended hours. You should know what you’ll have at the end and roughly when.

They can show you something they run in their own business. Anyone who builds these systems for a living should have their own running every day.

They’ll tell you when you don’t need them. If someone runs you through a test like this one and says, “Honestly, just buy the tool,” that’s the person to call back when you do need help.

Red flags

They open with a list of tools and platforms before they understand your business.

The discovery phase keeps going and has no build at the end of it.

The word “transformation” comes up a lot, and a finish line never does.

They can’t say who looks after the system after launch, or the answer is vague.

They won’t put you in touch with a client.

If you’ve been burned before, rebuilding after a bad AI automation vendor covers what to do differently next time, and the AI consultant red flags guide goes further into the warning signs.

Consultant, agency or partner?

The labels get mixed up, so here’s how I see them. A consultant advises. An agency usually builds what you ask for, then moves on to the next project. A partner diagnoses, builds, and stays on to train and support the system as you grow. None of them is automatically better. It depends on your score. If you scored four or five, though, advice on its own won’t get you there, and a one-off build without support will drift by month three.

If You Go the Tool Route, Do It Properly

Plenty of readers will score zero or one. Good. Here’s how to give a tool a fair test so you’re not back here in six months wondering what went wrong.

Pick one task worth at least two hours a week

Not the most interesting task. The most expensive one. Writing quotes, answering the same five customer questions, writing up meeting notes, chasing invoices. If it doesn’t take at least two hours a week, the gain won’t be big enough to change your habits. The guide to admin overload in small business is a good place to find candidates.

Give the tool some context

Write one page about your business. What you sell, who you sell to, how you talk to customers, what you never do. Save it as standing instructions so you’re not explaining yourself from scratch every time. That one page is the difference between generic answers and answers that sound like you.

Put it where the work happens

If you quote from your phone on site, the tool needs to work on your phone on site. If your team lives in the inbox, it needs to be in the inbox. Every extra step between the work and the tool cuts how often it gets used.

Measure it for two weeks, then decide

Track the time the task took before and after. Be honest. If it saves you three hours a week, keep it and pick the next task. If it doesn’t, try one adjustment, then drop it.

If it stalls, take the test again

Stalling isn’t failure. It’s information. If the tool stalls, go back to the five questions. Usually one of them has quietly become a yes. The task turned out to have a hand-off in it, or the knowledge it needed was in your head. That’s the point to look at outside help, and you’ll be a much better buyer for having tried it yourself first. For a sense of what comes after single tools, AI agents for small business show what changes once several tasks are linked up.

Conclusion

“Do I need an AI consultant?” is a fair question, but there’s a more useful one underneath it: what kind of problem do I actually have? A task needs a tool. A decision needs a plan. A business that runs out of your head needs someone to build it a brain and a workforce, then keep both working as you grow.

Run the five questions honestly. If you score low, save your money and spend a Friday afternoon setting up a tool properly. If you score high, stop adding tools to a system that has no centre. The real fix is getting what’s in your head into a system the whole business can use: an AI brain that knows how you work and an AI workforce that does the repetitive jobs without you. That’s what gives you your evenings back. Another subscription won’t.

Not Sure Where You Landed?

If you scored two or more and want a second opinion, book a free 30-minute Discovery Call. Tell me about the problem you ran through the test, and I’ll tell you straight whether it’s a tool job, a plan job or a build job. If a tool will do, I’ll say so and point you to the right one.

Frequently Asked Questions

About Octavius