What Is an AI Consultant, and Does a $5M Company Need One?

What is an AI consultant? Somebody you pay to figure out where AI actually makes you money, then go make it happen and prove it worked. That's the job.

Most of what gets sold under that title is a slide deck or a software subscription. Neither one decides anything, and deciding is the whole point.

What is an AI consultant?

An AI consultant helps a company decide where artificial intelligence belongs in its operations, implements the tools, trains the people who'll use them, and measures what came back. At $5M to $50M the job is business judgment first, technology second. You're looking for the one or two places AI pays, not installing it everywhere.

Four different animals wear this badge

The enterprise firm. Accenture and everybody chasing Accenture. Machine learning deployment, data platforms, a steering committee, six figures to start. They're good at what they do. What they do starts at a company size you're not.

The technologist. Builds custom models. Can tell you exactly what's possible. Can't tell you what's worth doing, because that means knowing your margins, your crew capacity, and which customer takes 90 days to pay.

The tool reseller. Sells you a subscription with the words "AI consulting" stapled to it. The consulting is a demo. The deliverable is an invoice, every month, whether anything changed or not.

The operator. Has run a company. Decides, builds it, trains your people, measures it, then gets out of the way. Rarest of the four, and the only one that fits a company your size.

You probably just figured out which one pitched you last month.

What the job actually looks like

Four steps. Skip one and the whole thing dies quietly, which is what happens to 95% of corporate AI pilots according to MIT's 2025 study. Ninety-five percent.

Find where AI pays. Full read of the business. Finance, operations, marketing, sales, HR. Not "where could we use AI" but "where do people sit around waiting, or retype something that already exists somewhere else." It's usually two spots, and it's usually not the one the owner guessed.

Build it into what you already run. The CRM. The inbox. The forms. The books. Not some new tool with its own login that nobody opens after week three. I went through the sequencing in AI for contractors, and it holds for every trade.

Train the people, not the org chart. At the manufacturer, content production time dropped 38%. That number came from prompt libraries written for the exact job each person did, sitting inside the software they already opened every morning. Not a training seminar. Nobody has ever learned anything from a training seminar. Full story here.

Prove it before you scale it. Hours saved. Response time. Cost per output. Before and after. If the first workflow doesn't show a number in 30 days, you don't get to build the second one.

Those four are what's on the AI advisor page. Not complicated. Just rarely done in that order by somebody who has to live with the result afterward.

What it costs, and what the alternative costs

An operator engagement runs about the same as senior fractional marketing help. Five to fifteen thousand a month, scoped to hours, no long contract. Enterprise firms start at six figures and send you a team you'll never meet in person.

But here's the expensive option, and it isn't either of those.

It's the pilot that fails. A subscription nobody cancels, six months of your team quietly working around a tool they didn't want, and an owner who now believes AI doesn't work for a company like his. That one costs more than any consultant, because it shuts the door for two or three years.

So does a $5M company need one?

Yes, if any of this sounds familiar. AI is showing up in pockets, a chatbot here, some automation there, and nobody's connecting it to a plan. Leadership feels pressure to do something with AI and has no idea where to point it. Or a pilot already died and nobody can tell you why.

No, if you don't have real operations yet to apply it to. Also no if what you actually want is custom software built. That's a developer, and a good one will tell you that in the first ten minutes.

Honest middle ground: most owners between $5M and $50M don't need forty hours a week of AI thinking. They need eight to fifteen, from somebody who's carried a number, until two workflows are earning their keep.

Five questions that sort the four types fast

Have you run a company? Show me one rollout and the number it produced. Who does the actual implementation, you or somebody I'll never meet? What are you going to tell me to stop doing? How do we measure this, and by when?

Answer all five in plain words and you're worth a second call. The general version of that list, for any marketing consultant, is on the consultant page.

Where to start this week

Find the one process where your people wait around or retype things. That's your first meeting, right there.

Before anybody touches a tool, set the account rule. Company accounts for company data, training switched off. The AI usage policy template takes twenty minutes and prevents the most common way this goes sideways.

Then write down the three most repetitive jobs in your company and bring that list to the call. An operator will pick one and tell you what it should return in 30 days. A reseller will open a slide deck.

• An AI consultant decides where AI pays, builds it into what you already run, trains your people, and proves the return. Judgment first, technology second.
• Four kinds use the title: enterprise firm, technologist, tool reseller, and operator. Only the operator fits a $5M company.
• The expensive option isn't the consultant. It's the failed pilot that convinces an owner AI doesn't work for a company his size.
• Ask five questions: have you run a company, show me a rollout with a number, who implements, what do I stop doing, how do we measure it.
• Start with the process where people wait or retype the most, and set the company-account rule before anyone opens a tool.