Most agencies get this backwards. Niche down on the selling and stay general on the building, because fulfillment is no longer what forces the choice. Here are five filters for picking a trade, what the AI adoption data says about which ones are open, and when staying a generalist is the right call.
Lumina9 min read
Niche down on the selling, stay general on the building. Pick one trade for your outbound, your demo and your proof, and keep the ability to deliver for anyone who arrives by referral. The old reason to commit an entire company to one vertical was that each new industry meant learning a new business from scratch and rebuilding the system. When fulfillment is the same work no matter whose logo is on it, niching stops being an operations decision and becomes a marketing one, which is a much cheaper decision to get wrong.
That is the short answer. Most people asking the question are really asking something narrower: which trade, and how do I know before I burn three months finding out.
What niching actually buys you
Not credibility in the abstract. Four things you can measure.
Your opening line stops being generic. "We build AI systems for small businesses" gets you nothing. "We put an AI receptionist on the phones at roofing companies so storm-season calls stop going to voicemail" gets you a conversation, because the prospect can picture the exact failure you named.
Your objection handling gets reps. Every trade has three or four objections and they repeat. A dentist asks what happens when a patient asks about insurance coverage. A plumber asks whether it can tell an emergency from a water heater quote. By your tenth call in one trade you answer those before anyone raises them. Across ten trades you improvise every time.
Your pricing gets a spine. You can only hold a number when you know what a job is worth to the buyer. If a roof is $12,000 and they lose two a month to voicemail, $2,400 a month is an easy sentence to say out loud. Without their economics, you discount.
And your demo gets reused. Our plan publishes one demo at a time with unlimited drafts, which bites if you are pitching eight industries in a week. Pitch one trade and the published demo is almost always the right one to send.
What the adoption data says about which trades to pick
Do not pick a trade on vibes. There is public data on who uses AI and for what, and it points somewhere specific.
Nationally, adoption is still small. The Census Bureau's Business Trends and Outlook Survey found overall AI use among US businesses hovering between 17% and 20% from December 2025 through May 2026, and under 20% among firms with four or fewer employees. The spread by sector is wide: information at 39.7%, finance and insurance at 33.9%, retail trade near 14%. A Census working paper on the microstructure of AI diffusion puts the overall firm rate at 18%, with sales and marketing the most common function at 52%, and 57% of adopting firms using it in three or fewer functions at all.
Read that as a generalist and you conclude the trades are cold. Read it as someone who sells into them and you notice the national average hides the thing you care about.
Inside home services the picture is different. Jobber surveyed 1,050 home service business owners in December 2025 and found 52% already using AI day to day, with 27% more planning to start within the year. What they use it for is the interesting part. 54% for estimates, quotes and contracts. 52% for invoices and billing. 51% for writing emails and proposals. Only 35% for customer communication or follow-ups.
A separate survey of 1,017 residential and commercial contractors by Thrive Analytics for ServiceTitan, reported by Contractor magazine, found active engagement rising from 46% in December 2025 to 52% by September 2026. 34% named off-hours customer support as something they want. 44% said their barrier is understanding where AI applies, and only 11% reported high or complete trust in the AI tools they already use.
Three things follow, and they should shape your niche choice more than any guru's list.
The back office is taken. Quoting, invoicing and email drafting are jobs owners already do themselves with ChatGPT. Selling those means selling against a free tool they are comfortable with. We wrote up how to answer the ChatGPT question, because it will come up.
The phone is not taken. Customer communication sits at 35% in the Jobber numbers, lowest of the top five, and a third of contractors say they want off-hours coverage. Answering a call at 9pm is the gap.
Trust is low, so demos beat decks. 11% high trust means your prospect has probably tried something that did not work. Nothing you say fixes that. Them talking to the thing fixes it, which is the whole argument for putting a live demo in their hands instead of a recorded walkthrough.
How to actually choose the trade
Five filters, in order of how much they matter.
Job value. Your fee has to look small against one saved job. Roofing, HVAC replacement, personal injury, med spa packages and dental implants clear this easily. House cleaning and lawn care do not, which does not mean you cannot sell them, only that you sell on volume and your price ceiling is lower.
Calls they already lose. The pitch writes itself when they miss calls. Trades with emergency demand, seasonal spikes, or owners working with their hands somewhere loud are the natural fit, because the phone rings while nobody can answer it.
One decision maker. An owner-operator with six employees signs on the call. A 40-person company with an office manager, a marketing contractor and a part owner takes six weeks.
Density where you are. Count them. If your metro has 180 of a given business and you need 10 clients, that is under 6% penetration and it is fine. If it has 25, you have a hobby, not a niche.
An unfair in. A brother-in-law who runs commercial HVAC beats every other criterion here. Your first clients come from people who already trust you, so weight this heavily in the first ninety days. We go through that sequence in getting your first ten AI agency clients.
Run those five against the trade you are considering before you build anything. Then check whether that trade buys at all, which is a different question we covered in how to tell which local businesses actually buy AI.
Where niching does not help, and who should skip it
This is the part most advice leaves out.
Niching does not raise your price. A roofing buildout and a dental buildout cost us the same to deliver and are worth about the same to buyers of similar size. Niching improves your close rate and your sales cycle, not your invoice. If you expect to charge 40% more because your website says roofing, you will be disappointed.
Niching does not fix a bad offer. When the thing does not work on the phone, selling it to one trade means one trade knows.
Concentration is a real risk. A single-trade book is one bad season from trouble. Roofers in a metro with no storms have a hard year, and they all have it at once. Two or three adjacent trades with the same buyer profile is safer than one, and costs almost nothing extra when the packs are already written.
The obvious trades are crowded. HVAC, plumbing and roofing are what every AI agency picks, because they are what every list recommends. The work is still there, but you are the fourth call that month. Quieter ones get picked up less: septic service, foundation repair, auto glass, veterinary, independent pharmacy, commercial elevator maintenance. Fewer competitors, same missed calls.
Stay general if you have fewer than five clients. Your first handful come from your network, and your network is not one trade. Take them, learn from them, and let the niche show itself. The trade you enjoy selling to that also pays on time is usually the one to double down on, and no spreadsheet tells you which that is.
The structure that works: narrow sales, wide fulfillment
The old advice said pick one and commit because delivery capacity was the bottleneck. Every new vertical meant new research, new prompts, new integrations, a new way of talking.
That part is packaged now. We maintain over 100 vertical packs carrying the services, price bands, objections and booking language for each trade, plus a generic pack written to read as deliberately industry-neutral for anything appointment-based that is not on the list. Switching a demo from a dental practice to a tree service is a dropdown, not a project.
So run two layers. Your outbound, your positioning and your published demo all point at one trade. Your fulfillment stays open, because when a referral arrives from a different industry you should take the money. Serving that one odd client costs you a different pack and a different intake form, not a new business line.
In practice, you pick the trade, name it on your site, build the list, run the calls and send the one demo. Every other pack stays in drafts. Revisit in ninety days with real close-rate numbers instead of a feeling.
FAQ
How long before I know a niche is working?
Ninety days and roughly 150 real conversations, whichever comes later. Less than that and you are reading noise. Look for whether the second and third calls got easier, and whether anyone referred you without being asked.
Can I change niches later?
Yes, and it costs less than people fear. You lose the site copy, the list and some reputation inside that trade. You keep the fulfillment process, the pricing, the demo mechanics and everything you learned about running the call. Most agencies switch once or twice before settling.
Should my niche be a trade or a geography?
A trade, then geography inside it. Selling every type of business in one town keeps your referral engine working but your pitch never sharpens. Selling one trade across three metros sharpens the pitch and the referrals still travel, because owners in the same trade talk to each other more than neighbors on the same street do.
What if my prospect's trade is not in the pack list?
Use the generic pack, written to read as industry-neutral rather than as a template with the trade left blank. If you are committing to that trade, ask for a pack. We write niche packs by hand and build subscriber requests first.
Where to start this week
Pick two trades you could plausibly reach and run the five filters on both: job value, calls they already lose, one decision maker, how many exist near you, and whether you know anyone in it. One wins on paper. Build the demo for that trade, seed it with a real local business name, and call twenty of its competitors before changing anything.
To see what that demo looks like first, talk to the one on our homepage and switch the trade selector to the industry you are weighing. The platform is $49 a month with the demo site, client portal and fulfillment access included, and client builds get priced on a 30-minute call so the markup is yours to set. Start here once you have picked the trade.
