The short version
Pick one industry you already understand. Learn its most expensive boring process in detail. Offer a small paid diagnostic rather than a project. Talk to people who already know you, because the first client nearly always comes from a warm connection rather than a cold campaign.
Why doesn't "AI consultant" work as positioning?
Because it describes what you use rather than what you fix, and it puts you in a category the buyer cannot evaluate. Faced with three people who all say “AI consultant”, a business owner has no basis to choose except price and confidence, which selects for the wrong things.
It also fails the recognition test. Nobody wakes up wanting an AI consultant. They wake up knowing that three people spend two days a week matching invoices and the month closes four days late. Someone who leads with that sentence has already demonstrated more than a credential can.
How do I choose a niche without limiting myself?
Start from what you already know rather than from market size. If you spent six years in logistics, you understand freight documentation, the systems involved, and why the obvious fix does not work — and that knowledge is worth more than any technical advantage, because it cannot be acquired by reading.
The fear of narrowing is misplaced at this stage. You are not turning down work you have; you are making it possible for one specific group to recognise you. Nothing prevents you taking an interesting project outside it, and after two or three engagements the niche is an asset rather than a constraint — it is the reason your second project takes half the time and can be priced higher.
Where does the first client actually come from?
Overwhelmingly from someone who already knows you: a former colleague, a previous employer, a client from a different line of work, someone you worked with years ago. Not because cold outreach never works, but because a first engagement is a purchase of trust, and trust is what a warm connection already supplies.
Which makes the practical first step unglamorous: write down twenty people who know your work and would take your call. Then talk to them about their problems rather than your services. A useful opening is some version of “what is the most tedious, expensive thing your team still does by hand?” — most people answer that question immediately and with feeling.
What should I offer them?
Not a project. A small, fixed-price diagnostic: you look at one process, sample real data, measure how much of it is genuinely routine, categorise the exceptions, and produce a written recommendation with a scope and a price. Days rather than weeks.
This works for three reasons. It is an easy yes, because the cost is low and the deliverable is concrete. It pays you for thinking rather than giving it away in a “free consultation”. And it means you never quote a build blind — which is how new independents lose money on their first fixed-price project.
Being willing to conclude “don’t build this” is what makes it credible. Say so up front, and mean it.
What if I have no portfolio yet?
Build one on your own problem. Every independent has processes of their own — invoicing, proposals, research, follow-up — and automating one properly gives you something real to describe: what you built, what broke, what it saved, what you would do differently.
A specific, honest account of a small system you built for yourself is far more persuasive to a sceptical buyer than a certificate and a capabilities list. It demonstrates the thing they are actually trying to assess, which is whether you finish things.
The other route is a deliberately discounted first engagement in your chosen niche, in exchange for specific things agreed in writing: a named reference, permission to publish the work, and an introduction. Not free — discounted, with terms.
How long should this take?
Longer than most people budget for, which is the main reason independents give up early. A first engagement arriving within a few months of starting deliberate outreach is a reasonable expectation; one arriving in the first fortnight is luck rather than process.
This matters if you hold a certification, because it runs for twelve months. Starting the pipeline work at month nine and hoping is how people end up renewing a credential that never earned anything. The conversations should start the week you pass, not the week you finish preparing.
Does being certified actually help?
More than practitioners expect and less than they hope. It rarely wins the work on its own. What it does is answer the question a buyer cannot otherwise answer about a stranger — is this person credible at all — and it does so with something independently verifiable rather than assertion.
Two things follow. Make it checkable: a public, verifiable profile someone can look at without asking you for anything. And pair it with evidence of delivery, because the credential opens the conversation and the story about something that broke is what closes it.
What do I do when they say the price is too high?
Ask what they were expecting, and listen for whether the objection is about the number or about the risk. Most of the time it is risk — they cannot tell whether this will work, so any price feels speculative.
The answer is not a discount, it is a smaller first step. A diagnostic they can afford to be wrong about de-risks the decision without devaluing your work, and a client who buys a small thing and is pleased buys the larger thing afterwards at full price. Discounting the build to win it teaches them what your work costs, permanently.
Should I write publicly?
If you can sustain it, yes — but write about the problem rather than the technology. A piece titled after the process you fix reaches the person who has that process; a piece about model capabilities reaches other practitioners, who are not your buyers.
The most effective format for a new independent is a specific account of something you built: what the process looked like before, what you changed, what broke, what it saved. That is simultaneously marketing, a portfolio piece and proof you finish things. Two of those a year beats weekly commentary nobody remembers.
What about the directories and marketplaces?
Worth being on where listing is free and the audience is right, and worth almost nothing as a strategy on its own. Buyers arriving through a listing are comparing you against everyone else on it, which is the least favourable context available.
The exception is anywhere your credential is independently verified, because that solves a real problem for a buyer assessing a stranger. Treat it as somewhere for a prospect to land and check you out after a conversation, rather than as a source of conversations.