Ask a room of HVAC, plumbing or roofing owners whether AI matters and most hands go up. Ask who has done anything about it and most hands come down. That gap is not laziness. It is a real problem: the pitch decks are everywhere, the vendors all sound alike, and nobody has told you which of the fifty possible projects to do first.
Here is the short answer. Start where money is already leaking out of your business, in the places you can count: calls nobody answered, leads that waited too long, estimates that went quiet, and customers who stopped hearing from you after the job. Put AI there first. Chatbots, content generators and "AI strategy" decks can wait.
The gap is real, and it is documented
ServiceTitan's 2026 Residential State of the Trades report, published April 7, 2026, surveyed 1,000 residential owners, executives, general managers and directors. The research was run by Thrive Analytics on ServiceTitan's behalf. The headline numbers:
- 74% of contractors view AI as an efficiency engine.
- About 25% are currently using AI.
- More than half report uncertainty about where to start.
- Among adopters, 48% report increased productivity.
- 73% believe starting early creates a competitive advantage.
Read those together. Three out of four think it matters, one in four does anything, and over half are stuck at step one. ServiceTitan's release also says the research is for informational purposes only and that it makes no assurances about accuracy, and it is a survey of opinions and self-reported use, not a measurement of results. Treat it as a map of where the industry is, not proof that AI pays. The proof has to come from your own numbers, which is why the rest of this guide is about measuring.
Why "where to start" goes wrong
Owners tend to pick a starting point in one of three ways, and all three are weak.
They start with what is visible. A chatbot on the website, an AI tool that writes social posts, a logo generator. These are easy to buy and easy to show off. They also rarely touch a booked job. A chatbot that answers questions about your service area does not put a technician in a truck.
They start with what a vendor sells. Every answering-service company, texting platform and CRM now has an AI feature. Each one is the answer to a different question. The vendor's demo shows the feature working, not whether that feature sits on your biggest leak.
They start with everything. A "digital transformation" plan with twelve workstreams and no first deliverable. Nothing ships, the team gets tired, and the owner decides AI is hype.
The way out is to pick the starting point by the money, not by the technology.
Find the leaks first
A contracting business has a short list of places where revenue that was already earned, meaning you paid for the lead or the customer already knows you, slips away. Look at these five in this order.
1. Missed calls
Every call you do not answer is a lead you paid to generate, or a customer who tried to reach you, that went somewhere else. The size of the leak depends on your missed-call rate, booking rate and ticket size, and you can work it out in about an hour. We wrote out the formula, with sourced benchmarks, in what a missed call really costs a contractor.
Why it goes first: it is the easiest leak to count (your phone system already reports it), and the fix is well understood. If you want to know what an AI receptionist can and cannot do here, read our buyer's guide to AI receptionists for contractors before you sign anything.
2. Response time on new leads
Not every lead comes by phone. Web forms, ad leads and marketplace leads sit in an inbox or a CRM queue until someone looks. The longer they wait, the colder they get, and the next company on the customer's list is usually one tap away. Our guide to speed to lead in home services lays out what the research does and does not show, including which popular statistics have no traceable source.
3. Follow-up after first contact
A caller leaves a message, a form gets filled out, an appointment gets booked and then cancelled. Most shops follow up once, maybe twice, then stop. A sequence of texts, calls and emails that keeps going for a few weeks is cheap to run and almost impossible to maintain by hand when your CSRs are also answering phones. See why contractors lose booked jobs after the lead comes in for where jobs drop out between first contact and the schedule.
4. Unsold estimates
You did the site visit. You wrote the quote. The customer said they would think about it. That is the most expensive kind of lost lead, because you already spent a technician's or estimator's time on it. Most shops have no systematic follow-up on estimates that have not closed. The mechanics and the math are in our guide to following up on unsold estimates.
5. Old customers
Your database holds people who bought a system five years ago, people who lapsed from a maintenance agreement, and people who asked for a quote in 2023. They know your name. Re-contacting them costs far less than buying a new lead. It is also the leak most owners have never measured, because nothing alerts you when a customer quietly goes away.
How to size each leak in one afternoon
You do not need a consultant to rank these. Pull five numbers from your phone system, your field service software and your books, and estimate each leak in dollars per month.
| Leak | What to pull | Rough sizing |
|---|---|---|
| Missed calls | Missed calls last 30 days, split by hour and branch | Missed new-job calls x booking rate x average ticket |
| Slow response | Web and ad leads last 30 days, with time to first contact | Leads not contacted within an hour x booking rate x average ticket |
| Weak follow-up | Leads that never booked and had fewer than three touches | Count x a conservative recovery rate x average ticket |
| Unsold estimates | Open estimates older than 14 days | Total quoted value x a conservative close rate on re-contact |
| Old customers | Customers with no job in 24 months or a lapsed agreement | Count x a conservative reactivation rate x average ticket |
Two rules keep this honest. First, use your own data wherever you have it, and label every guess as a guess. Second, use conservative recovery rates. If you cannot defend a number in front of your own accountant, it does not go in the sheet.
When you are done, you will have a ranked list with dollar amounts. The top one or two items are your starting point. For most contractors we would expect missed calls and unsold estimates to land near the top, but that is a hypothesis, not a finding. Your list may differ, and that is the point of doing the work.
Where not to start
A few places where we would not begin, and why.
- A website chatbot. It handles questions. Your revenue problem is bookings. Add it later if your data says visitors are leaving with unanswered questions.
- AI-written content and social posts. Useful for marketing efficiency, but it does not recover revenue you have already earned. It is a good second project, not a first one.
- Replacing dispatch or scheduling logic. High stakes, high complexity, and one wrong job assignment in January costs real money. Earn trust with lower-risk projects first.
- Anything that cannot be measured in booked jobs. If the vendor's success metric is "conversations handled" or "tasks automated," ask what that number has to do with your schedule.
- A multi-quarter roadmap before a first deliverable. Ship one thing in 30 days, measure it, then decide what is next.
Pick the first project by four tests
Once you have a ranked list, run your top candidates through these tests.
- Is the leak counted in dollars, from your own data? If not, measure for 30 days first, using the baseline steps below.
- Does the fix end in a booked job or a booked estimate? Not a lead, not a conversation, not an email open.
- Can it connect to the software you already run? An AI tool that cannot write back to ServiceTitan, Housecall Pro or Jobber creates a second inbox to check. Our guide to AI integration with ServiceTitan, Housecall Pro and Jobber covers what to verify.
- Does it respect the texting and calling rules? Automated follow-up touches consent, registration and recording laws. Read TCPA texting rules for contractors before you turn on any outbound sequence. That page is general information, not legal advice, and you should confirm your setup with counsel.
Baseline before you buy anything
The most common mistake after "starting with the wrong thing" is starting without a baseline. If you do not know last month's booked jobs from phone and web leads, you cannot tell whether anything you buy worked.
For the 30 days before you launch, record weekly:
- Inbound calls, answered calls and missed calls
- New leads by source (phone, web, ads, marketplaces)
- Time to first contact on new leads
- Appointments booked and estimates delivered
- Booked jobs and revenue from those leads
That is a spreadsheet, not a project. It also makes the decision for you: if the numbers show you answer 95% of calls and follow up on everything, your leak is somewhere else.
What a sensible first project looks like
A good first project has a narrow scope, a fixed price and a short clock. It touches one or two of the leaks above, connects to your field service software, and reports in booked jobs every week. It does not need a six-month discovery phase.
This is how we built the Booked-Jobs Recovery System: a fixed-scope engagement, live in 30 days and measured in booked jobs. It covers missed-call recovery, web and form capture, multi-touch follow-up, after-hours AI qualification, dormant database reactivation, CRM and field service software integration, lead scoring and weekly booked-job reports. Pricing is scoped on a call. Di-Hy does not sell a software product, so the recommendation follows your stack and your leaks, not a vendor's roadmap.
If you are not sure which leak is biggest, the AI Readiness Audit is built for that step. If you want to talk through strategy before you commit to a project, see our AI strategy consulting service.
A realistic sequence
An order we would use, and adjust to your numbers:
- Weeks 1 to 4: Baseline. Count calls, leads, response times and booked jobs. Rank the leaks.
- Month 2: Fix the biggest leak with the smallest tool that closes it, wired into your field service software.
- Months 3 and 4: Review weekly booked-job numbers against the baseline. Add the second leak only if the first one paid for itself.
- After that: Consider the visible stuff (chatbots, content) once the revenue leaks are plugged.
This is slower to describe than "buy an AI tool" and faster to pay back.
Frequently asked questions
Where should a contractor start with AI?
Start with the places where revenue is already leaking and you can count it: missed calls, slow response to new leads, thin follow-up, unsold estimates and dormant customers. Size each in dollars from your own data and begin with the biggest.
How many contractors are using AI?
ServiceTitan's 2026 Residential State of the Trades report, a survey of 1,000 residential contractors run by Thrive Analytics, found about 25% currently using AI, 74% seeing it as key to efficiency, and more than half unsure where to start. It is a self-reported survey, so read it as a snapshot of sentiment and adoption, not a measure of results.
Do I need to replace my field service software to use AI?
No. A good first project connects to the software you already run and writes results back into it. If a tool cannot do that, treat it as a red flag.
How do I know if an AI project worked?
Compare booked jobs and revenue from phone and web leads against a 30-day baseline taken before launch. Do not accept "conversations handled" as proof.