How do I choose a niche for an AI consulting business?

"Texas business owners" isn't a niche. I made that call about my own AI concierge offer: Texas is a location, not a vertical. A niche tells an owner you already understand their problems.

Should I niche by industry or by location?

Industry. Even if a Texas offer did well, you'd get a landscaper, a roofer, an accountant and an insurance agent, and every single thing you do for them is different. A friend of mine has eight or ten AI clients, all in different industries, and he's already frustrated, because nothing he builds can be reused.

In one vertical, the third, fourth and fifth client can use what you built for the first. Two people I talked to in the same week were doing AI for any company, and both said it sucks, because everything's different all the time.

How narrow is narrow enough?

One level deeper than the category you first think of. "Home services" is too broad. Pest control is a niche. Pool service is a niche. The test: can you template 90% or more of the solution so each new customer is close to copy-and-paste?

A pest control owner I talked to said the same thing from the buyer's side. Templatize one vertical to around 80% reusability and sell it by peer referral, because owners in the same trade talk to each other.

Narrower than feels comfortable is usually right. If you try to be the best AI person, you'll struggle. If you're the AI person for local media in towns under 12,000 people, you can become the number one person in the world at it quickly. That's how I became the expert in my own space.

"The best thing to do if you could, would be like, I'm the AI guy for pest control. Because then you can easily own that freaking niche and make, you know, good money. Now your TAM is going to be small. You're not going to make, you know, a hundred million dollars, but you could probably make five."

What makes an industry a good AI niche?

Look for	Why it matters
Fragmented	Lots of small owners you can sell the same thing to
Light regulation	Projects don't stall on compliance and carrier limits
Seasonal	Harder for private equity to roll up the software layer
Recurring revenue	The customer can carry a monthly retainer
Obvious ROI	Faster lead follow-up pays for itself
Money, but overlooked	Big enough to pay well, small enough that nobody markets to it, like public adjusters

Pool service checks most of those boxes. That's why I told a generalist AI automation founder to go all in on it, ideally with dedicated branding and its own YouTube channel.

Franchise-style niches are the extreme version. I know an operator who sells the same GoHighLevel setup to Montessori schools. Every location needs nearly the same marketing, so he has thousands of clients and one graphic designer. I'd rather build something so custom to a weird industry that clients pay $500 a month than sell a $97 tool that kind of works for everybody.

What does a niche business look like when it works?

The best examples I know come from insurance, where I spent nearly twenty years.

• One software, one carrier. A former agent set up one complicated agency CRM for herself and started posting about what she fixed and automated. So many agents asked her to do it for them that she sold her agency. Now she implements that one software almost entirely for one carrier's agents and can't keep up. My guess is she makes at least half a million a year.
• One skill, then everything around it. Another agent started doing automations for other agents, hired an overseas team as the work grew, and added a virtual assistant business when clients asked to hire assistants too. Separately, a former agent who learned Zapier automations before AI recently posted that he has 375 employees, five or six years in.

Both grew through referrals. Industries like insurance all talk, and there are Facebook groups with 5,000 to 50,000 agents in them. Probably 50 or 100 people in those groups have real influence. Win over one or two, offer them 20% of everyone they send, and they'll spread the word for you.

Which industries look good but aren't?

Roofing. I spent eight years as an insurance adjuster working with roofers. Big ticket per job, effectively no recurring revenue, and a huge number of tiny operators, a lot like realtors. It only works if you bundle AI with marketing, a website and lead generation.

Real estate. It's big, so too many people go after it, and breaking through is hard. Smaller niches like pest control, accounting or estate planning have far fewer specialists.

Dentists and lawyers. Everyone targets them, so they get hounded and are hard to reach unless you have an in. I wouldn't pay $6,000 for a dentist convention to promote a $50-a-month side project.

Insurance, with a caveat. I have the background and I know how much money is there. I think AI for insurance agencies is maybe the biggest play out there, like automating the manual quote-rating process. I also don't love it. It's regulated, big companies are piling into AI for insurance, and I expect fewer agencies over the next few years. Captive agents can't run much because their carrier controls the systems. Independent agencies are different: numerous, wealthy, still manual, and not locked into one carrier's stack. I'd have to go learn insurance again, which I don't want to do. More in how insurance agents can use AI.

"Home services" as a label. It's big in Texas and faster follow-up has an obvious payoff, but it's a category. Pick one trade inside it and learn from real conversations with those owners.

Should I pick an industry I already know?

Yes, if you can. The edge is knowing the problems. I'd been doing newsletters for two years, I knew the problems, I built something to solve them, and I already had an audience of newsletter people to sell it to.

Depth also brings opportunities you'd never get as a generalist. At a local media conference I met the host, who runs the biggest agency in the country. Now I help his clients. If I'd just been a newsletter guy, he wouldn't have cared.

Start as narrow as your existing audience allows, then widen. My plan for AI services starts with legacy local media, because that's where I already am. Once we have clients and proof, we broaden into all media.

If you don't have that, borrow it. Find people who are deep in a niche, especially ones with an audience, and ask them directly:

"What problems do you have? If you could build anything, what would you want?"

For an industry you don't know, skip the ads and go human to human. Call people, admit you're new, offer to do the work free, and ask for introductions to the Facebook groups and the big names. Another angle: take something proven in one industry to an industry that isn't using it. That's how I started. Local media was anti-Meta-ads, so people like us came in with Meta ads.

Most local business owners won't touch a complex AI agent platform. It isn't for them. A simple tool shaped around their trade, sold through someone they already follow, is.

What should I bring besides AI?

Marketing. Meta ads knowledge on top of AI is rare among AI-only operators. It lets you sell the whole back office instead of one automation at a time, which is how you justify $3,000 to $5,000 a month.

A media brand for your vertical is the other piece. Very few trades have a brand that teaches its owners AI: a newsletter, a YouTube channel, a $99 digital product, a $200-a-month community with calls. The smaller the audience, the stronger their attachment to you. Be the AI expert for your vertical, not the AI expert in everything. See why generic AI consulting doesn't work for the other half of this argument.