AI Consulting as a Side Business: Rates and Scope

AI consulting as a side business means selling decisions, not builds. A business pays you to work out where AI belongs in their operation, where it doesn't, what has to be true about their data and their people before it's safe, and which tool to pick. You deliver a written verdict they can act on without you. The rate follows from how many days a month you can actually sell, and for a side business that number is small enough that the day rate has to be high or the whole thing pays less than the day job it sits beside.

Every page ranking for this search is either an enterprise firm's services page or a list of enterprise firms written by one of them. None of it is for a person with a job, a skill, and eight spare days a month. This is.

What An AI Consultant Actually Does

A consulting firm's glossary that ranks for this phrase defines it well enough. AI consulting helps organizations determine how AI can address business priorities, improve operations and support growth, and it lists the work as readiness assessments, vendor evaluation, implementation planning, workforce preparation, governance, data quality and security, responsible-use policy, and human oversight.

That's eight things, and at enterprise scale each one is a team. For a small business the same eight collapse into three questions the owner can't answer alone.

Which of our repeated tasks should a model touch, and which should it never touch. What are we allowed to put into these tools, given what's in our files. And what do we tell staff and customers about it.

A consultant answers those in writing, with reasons, in a form the owner can hand to whoever does the work. The moment you start doing the work yourself, you've become something else, and that something else is priced differently and covered in the piece on starting an AI agency from zero.

The Scope Ladder

The boutique offer page in the top 10 lists five engagement types, executive alignment through to an offsite, with no prices and no durations. I'd rather show the ladder with the time box and the acceptance test attached, because that's what makes a side business survivable. Every rung has an end.

The audit day itself is mostly listening. You sit with the owner and whoever runs the back office, you ask them to walk you through last Tuesday, and you write down every task that happened more than once, who did it, what it touched, and what went wrong the last time it went wrong. By the end you have a list that the owner has never seen written down, and the verdicts almost write themselves, because a task that touches customer payment details and has no review step is a "never" before any tool gets mentioned. The second day is turning that list into a document the owner would be embarrassed to ignore.

The audit is the front door and the only rung I'd sell cold. One small agency in the results for the neighboring search sells exactly this, a flat $1,000 audit credited toward the first build, and the structure is right even if the price is theirs. The credit turns a cautious client into a paid discovery call, and a client who doesn't go ahead still paid for your day.

The workshop is the rung most people underrate. It's the only one where the client watches AI work on their own invoices or their own support tickets, and I'd guess it's where most of the retainer conversations start, though that's a guess and not something I've sold.

The Day-Rate Worksheet

The rate question has to start from days, not from what consultants "usually charge," because a side business doesn't have consultant days. It has evenings and a weekend, and half of those go to finding the next client.

The inputs here are mine and labeled. Eight sellable days a month, which is a lot for someone with a job. Half of them billable, because selling, preparing and writing up eat the rest. That leaves four billed days.

Two things fall out of that table.

The first is that a $500 day rate, which sounds respectable, is a $2,000 month at a 50% billable share and a $62.50 hour. If the billable share slips to a quarter, which it will in any month you're mostly prospecting, it's a $1,000 month. Side-business consulting pays well per hour or it doesn't pay at all, and the audit priced at a thousand dollars for two days of work is sitting right at the bottom of that table.

The second is that the number to fight for isn't the rate. It's the billable share. A referral costs no selling hours, a repeat client costs no discovery hours, and a client on the oversight rung pays for a capped set of hours you didn't have to find. Every rung in the ladder is designed to make the next one cheaper to sell, and that's the whole point of the credit on the audit.

I couldn't load a primary wage source for employed analysts while writing this, so I'm not quoting a salary. I'd rather give you the arithmetic than a figure I can't trace. The worksheet is more useful anyway, because your number depends on your days.

The Production Trap

The one Reddit thread in the top 10 is consultants describing what went wrong when the gold rush hit. The snippet says it plainly, that when consultants frame value around quick outputs like research, decks and synthesis, they train clients to see consulting as production. I couldn't open the thread, so that's all I'll quote, and it's enough.

The trap works like this. A client asks for "a few prompts we can use." You write them, because it's easy and they're grateful. Next month they want the prompts updated, then a draft of the policy, then "could you just run the report." Each request is small, each one is production, and none of them is the decision they hired you for. Your day rate is now a freelancer's hourly rate with extra meetings.

I'd write the boundary into the first proposal. Deliverables are verdicts, maps, shortlists, policies and sessions. Drafts, prompts, documents and reports are examples inside those deliverables, never standalone items. If a client wants the standalone items, that's fine, and it's a different service with a different price, which is what the AI freelancing guide is about.

A rule I'd hold myself to. If more than a third of the hours in an engagement went to producing artifacts rather than deciding things, the next proposal to that client is priced as production or declined.

What To Refuse

A consultant's authority comes from being willing to say no, and there are four things I'd refuse regardless of the fee.

Decisions about individuals. Any workflow where a model scores, ranks, screens or flags a specific person, for hiring, firing, lending, tenancy or discipline. Whatever the law says in the client's jurisdiction, the reputational and human cost of getting it wrong lands on the client and the referral chain leads back to you.

Regulated advice at the output layer. A model drafting a medical, legal or financial answer that goes to a customer without a qualified human reading it. Draft assistance for the professional, yes. Replacement of the professional, no.

Anything the client wants to attribute to the machine. "The AI decided" is not a sentence a business gets to say to a customer, a regulator or an employee, and a consultant who helps them set it up has built the excuse rather than the system.

Data you can't see the boundary of. If the client can't tell you what's in the files they want to feed a tool, the first engagement is finding out, and it's not a workshop.

Straight Answers To The Questions Google Suggests

What exactly does an AI consultant do? Works out which repeated tasks a model should handle, which it should assist, and which it should never touch, then writes down what has to be true about data, staff and disclosure before that's safe. The deliverable is a decision with reasons. The build is somebody else's job, sometimes the same person wearing a different price tag.

What is AI consulting's salary? I'm not going to quote a figure, because the primary source I tried wouldn't load and I won't repeat a secondary figure I can't trace. For a side business the honest calculation is days times rate times billable share, and the worksheet above shows why $500 a day can be a $1,000 month.

What are the top 10 AI consulting companies in the USA? The list ranking for this question is published by a firm that ranks itself second. If you're a business looking to hire, I'd ignore rankings and ask three things. Will they put the verdict in writing, who in their team has run the tools on data like ours, and what does it cost when we want to stop.

Which jobs won't survive AI, what's gone by 2030, what disappears in five years? Google bundles three versions of the same question under this search, and I can't forecast any of them from this chair. Neither can the consultants who answer with a numbered list. What I'd say to a client asking is narrower and more useful. Look at your own repeated tasks, mark the ones a model does adequately today, and plan for those. A workflow-level answer is honest, and I don't have an occupation-level one to sell you.

The Small Bet I'd Run

Pick one business type you understand from the inside, because the audit is only worth a thousand dollars to someone whose workflows you already recognize. Then sell one audit, and only one, at a fixed fee with a credit toward the next rung.

I haven't sold consulting, so the sample I'd bring is the one I have. I run local image and video generation on an M4 Pro for my own projects, which means I can talk about what a local setup costs, what breaks, and what it can't do, from receipts rather than from vendor pages. Whatever your equivalent is, the thing you've actually run, that's the sample. A consultant with a sample beats a consultant with a deck.

Write the production boundary into the first proposal before you're tempted to skip it. Track billable share, not revenue, for the first three months. And if the audit turns into "could you just build it," take a breath, because that's a real business too, and it's a different one. The tool-by-job breakdown in the best AI tools to make money is the shortlist I'd start the selection rung from, and the AI agency piece is where the building version of this gets its own worksheet.

Stop when three audits in a row produce no second rung. That's the market telling you the verdicts aren't worth a day rate to that business type, and it's cheaper to hear it after three than after thirty.