ServiceTitan integration AI receptionist: what actually works

Four out of five shops who ask us about a ServiceTitan integration AI receptionist assume we plug into the FSM the same way their dispatch board does. We do not. Almost nobody in this category does yet, and the vendors who claim “native” usually mean one of four patterns underneath.

This post is the honest map. If you run ServiceTitan or Housecall Pro and you are shopping an AI receptionist, here is what “integration” means in practice, which pattern fits which shop, and where the gaps still sit. No vaporware, no “coming soon” slides — just what you can actually turn on this week.

Updated July 2026. Two things changed since we first published this in April — ServiceTitan’s webhook platform and ServiceTitan’s own AI voice agent — and both change the answer:

ServiceTitan’s V1 webhook platform is closed to new subscriptions, with V2 still in development. The pattern most third-party vendors quietly relied on isn’t available to new builds.
ServiceTitan now sells its own AI voice agent that answers calls and books jobs from inside the platform. If you’re on ServiceTitan, that’s now a real option you should price against anything third-party — including us.

We’ve also added a walkthrough of the app-approval chain, because it’s the step that catches owners off guard: your admin holds the credentials, not the vendor.

What “servicetitan integration ai receptionist” really means

When a voice AI vendor says “we integrate with ServiceTitan,” they’re describing one of four architectures underneath. Knowing which one matters because they fail in different ways.

Direct API integration — the AI reads and writes to the FSM via its public API (ServiceTitan’s Tenant API, Housecall Pro’s developer API). Jobs, customers, and dispatch slots update in near real time. Cleanest UX, longest build time, most constraints on what the API will let you do.
Zapier / Make middleware — the AI dumps structured call data (name, address, job type, preferred time) into a Zap, which creates the job or customer in the FSM. Fast to stand up, cheap, breaks quietly when fields change.
Webhook + FSM-side automation — the AI fires a webhook on call completion; the FSM (or a thin glue service on your side) picks it up and creates the booking. Works with any FSM that supports inbound webhooks or has a rules engine.
No integration (manual dispatch) — the AI captures the call, sends a summary to the dispatcher’s phone and email, and the dispatcher keys the job into the FSM. Zero integration work, zero breakage, but it assumes a human is watching.

Every shop we onboard lands in one of these four buckets. The pattern you want is usually not the one the sales deck shows.

ServiceTitan vs. Housecall Pro: what integrates, how, and the known limits

ServiceTitan and Housecall Pro cover maybe 70% of the residential HVAC, plumbing, and electrical shops we talk to. ServiceTitan alone closed its fiscal 2026 with roughly 10,800 active customers, per its FY26 results, and pushed $21.7 billion of gross transaction volume through the platform in the single quarter ended April 30, 2026 — up 23% year over year, on $268.8 million of revenue and net dollar retention above 110%, per its Q1 FY27 release. That’s the installed base any voice AI vendor has to plug into. Their API surface areas are different enough that you need to pick the pattern per FSM, not per vendor.

Capability	ServiceTitan	Housecall Pro
Create a customer record via API	Yes (Tenant API)	Yes (developer API, partner-approved)
Create a job / booking via API	Yes, with location + job type	Yes, with customer + schedule slot
Read real-time dispatch availability	Limited — you query open slots, not live capacity	Limited — scheduling endpoint exists, no live GPS
Zapier connector	Available, partner-published	Available, native
Inbound webhooks into AI	V1 closed to new subscriptions; V2 in development	Yes — but MAX plan only
Change-tracking if webhooks are unavailable	Poll with `modifiedOnOrAfter` filters	Poll the jobs/customers endpoints
Typical build time for a native API plug-in	4–8 weeks	2–4 weeks
Known limitation	Dispatch logic still lives in the FSM — third-party AI cannot route techs	Webhooks and most API scopes gated behind the top plan tier

Two rows there changed since spring, and they matter more than the rest of the table.

ServiceTitan’s V1 webhook platform is closed to new subscriptions while V2 is still being built, and ServiceTitan’s own guidance is to migrate off V1 as soon as V2 lands. Until then the documented interim pattern is polling: hit the relevant endpoints with a modifiedOnOrAfter filter and diff what came back. That’s slower and chattier than a push, and it’s the reason a vendor demo recorded in 2025 may not reflect what a new tenant can actually turn on today. Ask any vendor pitching you a “ServiceTitan webhook integration” whether their subscription predates the freeze.

On Housecall Pro, webhooks require the MAX plan. Not Essentials, not Pro. If you’re on a lower tier, budget the plan upgrade as part of the integration cost or plan on polling — the public API docs spell out which events fire once you’re eligible.

The big caveat on both platforms: a third-party AI does not replace the dispatcher’s routing logic. No public API in either product exposes “which truck is closest and free in the next 90 minutes.” You can create a job, attach it to a customer, and hand it to the FSM’s rules engine — but the rules engine (or the human dispatcher) still decides which tech takes it.

For the InstaNexus AI receptionist specifically, our current production pattern on both platforms is webhook + FSM-side automation where the tenant already has webhook access, polling where it doesn’t. Direct API integration is on the roadmap, not shipped. If you need the AI to read live dispatch availability before quoting an arrival window, we will tell you straight that today we quote a window the dispatcher can override rather than pretending to book a live slot.

The approval chain nobody mentions on the sales call

Here’s the part that surprises owners: a vendor can’t connect to your ServiceTitan account on their own. You turn them on, by hand, from inside your own tenant.

ServiceTitan’s developer docs spell out the chain. Third-party developers get only the integration environment — a sandbox seeded with sample demo data. Production access is customers-only. Which means every vendor demo you watch is running against fake data unless a real shop already granted them access to a live tenant.

Getting from that sandbox into your live tenant takes four steps, in order:

The vendor registers an app in the Developer Portal, picks Read and/or Write API scopes, and adds your tenant ID so the app becomes visible to you. Every ServiceTitan customer has at least one tenant ID; your admin can pull it.
ServiceTitan approves it. If the vendor registered a Public App — one built for many customers rather than just you — tenants only see it after ServiceTitan’s review, which the docs put at a few business days after registration. A Private App skips that queue but works for exactly one customer.
Your admin generates the credentials. This is the load-bearing step. Your ServiceTitan admin reviews the scopes the app is asking for, then generates and shares the client ID and client secret. The vendor can’t mint their own. Say no and nothing connects.
The vendor switches to production. Going live means swapping the integration client ID and secret for the production pair and calling api.servicetitan.io instead of api-integration.servicetitan.io. The app key itself works in both environments.

Two consequences worth knowing before you sign.

If the vendor later adds a scope — say they start writing jobs when they used to only read customers — that generates a new app version, and the new scopes stay dormant until your admin upgrades the app version inside ServiceTitan. That’s good for you. Permissions can’t quietly expand behind your back. It also means an integration can sit half-broken for a week because nobody clicked upgrade.

ServiceTitan also tells developers not to ask customers to create users inside their own environments for development work. So if a vendor asks you to spin up a ServiceTitan login for their engineers, ask why. Not automatically fatal. Worth a straight answer.

Three questions to put on your next demo call:

Is your app registered Public or Private, and has ServiceTitan approved it?
Which exact API scopes do you request — read, write, or both?
What does View Connections show for tenants my size right now?

That third one does the most work. View Connections shows a developer which tenants are on their app, on which version, with which scopes live. A vendor with real production customers answers it in ten seconds. A vendor with a sandbox and a slide deck changes the subject.

ServiceTitan now sells its own AI voice agent. Should you just buy that?

Sometimes, yes. This is the section a vendor blog isn’t supposed to write, so here it is.

ServiceTitan’s AI voice agent answers inbound calls, books jobs against real ServiceTitan data, confirms and reschedules existing appointments, recognizes members and repeat customers by their record, states your dispatch fee, and transfers to a human when it hits something it shouldn’t handle. It ships as part of Max, ServiceTitan’s AI layer, and ServiceTitan says it works alongside your existing phone system or inside Contact Center Pro.

The structural advantage is the one no third party can match: it lives inside the dispatch board. Everything above about “no public API exposes live tech availability” applies to outside vendors. ServiceTitan’s own agent isn’t outside. Its capacity-matching reads the same scheduling data the dispatcher sees, which means it can genuinely offer a slot rather than quote a window and hope.

Two things to weigh against that.

First, the reported results are ServiceTitan’s own numbers. The company says shops on Max show call booking rates up over 500 basis points and close rates up over 1,000 basis points versus peers, and that Max locations more than doubled in the quarter ended April 30, 2026, as reported in Plumbing & Mechanical. On the June 4, 2026 results call, the company added that it’s automating Max onboarding and expects to “again double the number of locations on Max during Q2,” in its own words. Read that as evidence the queue is moving, not that the product is proven. Apply the same skepticism you’d apply to a number on our site. Vendor-aggregate lift figures rarely control for the fact that the shops who adopt new tooling first are already the better-run shops.

Second, it’s still in limited release. ServiceTitan customers apply for early access and onboard as capacity opens — and the company’s own framing is that it’s still working to widen that funnel. If you need call coverage in the next two weeks because your CSR gave notice, “apply for early access” is not an answer.

Here’s the decision, compressed. If you’re a ServiceTitan tenant, you’re happy staying single-vendor, and you can wait for the queue — evaluate their agent first. It has an integration advantage you can’t buy around. Go third-party when you run more than one FSM across locations, when you want the phone layer independent of the software that runs your P&L, when you need to be live now, or when your call volume doesn’t justify the ServiceTitan tier the agent requires. Those are the four cases where the shops we onboard land outside the FSM.

Which pattern fits which shop

We pick the integration pattern based on call volume, staffing, and how tolerant the shop is of a 2-minute delay between call and FSM entry.

Under 40 calls a day, owner-dispatched. Use pattern 4 (manual). The owner gets an SMS summary and a link to the call recording. Total time to key into ServiceTitan is 60 seconds per booking. You save the integration spend for year two.
40–120 calls a day, one dedicated dispatcher. Use pattern 2 or 3 (Zapier or webhook). Bookings land in the FSM within 30 seconds of call end. The dispatcher confirms the tech assignment and the customer gets the “on the way” SMS from the FSM, not the AI.
120+ calls a day, multi-location, dispatch team. Use pattern 1 or 3. You need either a direct API integration or a monitored webhook-or-polling pipeline, plus a defined fallback when the integration is down. At this scale, a 5-minute outage of the Zap during a heat wave is a real revenue event.
Any shop that answers 24/7 for emergencies. Prioritize the summary SMS pathway regardless of which primary pattern you pick. When a pipe is flooding a basement at 2am, you want a human seeing the call within 60 seconds. FSM plumbing can wait until morning.

If you are specifically running HVAC, we walk through the exact call-to-booking flow on the AI receptionist for HVAC page and in our guide to booking HVAC jobs automatically. Plumbing and roofing shops follow the same decision tree with different job-type fields.

Honest limitations (the part vendors skip)

Five things we wish every vendor in this category was upfront about. We will be.

1. Live tech availability is not exposed to third parties. Neither ServiceTitan nor Housecall Pro publishes a “tech A is free at 2:15pm with a 12-minute drive time” endpoint to outside developers. An external AI that offers you a firm arrival slot is either guessing or has built a side system that duplicates your dispatch board. Expect a 2-to-4-hour window, not a precise ETA, unless your dispatcher approves the slot inside the FSM. The exception is the FSM’s own agent — ServiceTitan’s reads the scheduling data directly, which is exactly why the build-or-buy math changed this year.

2. Price quoting is bounded. Both FSMs carry price books, and both expose them via API — but only for flat-rate line items. A custom quote (“replace the evaporator coil on a 15-year-old Carrier”) still needs a human. The AI can quote a diagnostic fee and an emergency surcharge and nothing more.

3. Custom fields break integrations quietly. If your dispatcher adds a new required custom field in ServiceTitan on Tuesday, every AI-created job on Wednesday fails silently until somebody notices the FSM is missing bookings. This is the single most common post-launch incident. Fix: lock the job-creation schema, alert on API errors, audit weekly.

4. Call recording and consent live outside the FSM. Both FSMs will store a link to the recording, not the audio itself. If a state requires two-party consent and you turn on recording inside the AI, make sure the consent prompt fires before any caller data hits the FSM. This is not legal advice — talk to counsel.

5. The “integration” is only as good as the service menu. The AI cannot create a job for a service you forgot to tell it you offer. Every integration project we have run starts with a service-menu audit, not an API call. If you want a pre-onboarding view of what that looks like, our AI receptionist vs. answering service pillar walks through the shape of a real service menu.

For the curious, the primary sources we keep bookmarked are the ServiceTitan developer portal and the Housecall Pro API reference. Both are public. Both will tell you more honestly than any vendor deck what the API actually supports — and both moved this year, which is why we rewrote half this post. Read the docs before you read the sales page.

Build or buy: when a direct integration is worth it

Direct API integrations cost money to build and money to maintain. The question every shop asks — should we pay the vendor extra for a native integration, or run a Zap? — has a simple answer.

Pay for native when all three of these are true:

You are past 100 calls a day and a 30-second lag actually costs jobs.
You have at least one staff member who owns the integration when it breaks.
You need the AI to read FSM data (customer history, open estimates, warranty status) to handle the call, not just write a new booking.

Otherwise, a Zapier or polling flow is the right answer. You can always graduate later without retraining the AI on your service menu. The service menu is the moat; the integration is plumbing.

And there’s now a third door. If all three conditions above are true and you’re a ServiceTitan tenant, the cheapest path to “the AI reads FSM data” may be buying it from ServiceTitan rather than paying anyone to rebuild it from outside. We’d rather tell you that on a demo call than have you find out in month four.

Frequently asked

Q: Can an AI receptionist integrate with ServiceTitan and Housecall Pro today? A: Yes, via Zapier, webhook, polling, or direct API. Which pattern fits you depends on call volume and whether you need the AI to read FSM data or just write new bookings.

Q: Do ServiceTitan webhooks still work for new integrations? A: Not for new subscriptions. ServiceTitan’s V1 webhook platform is closed to new subscribers and V2 is still in development, so new builds poll the API with a modifiedOnOrAfter filter instead. Existing V1 subscriptions keep running for now, but ServiceTitan advises migrating to V2 once it’s available. If a vendor demos a live webhook flow, ask whether their subscription predates the freeze.

Q: Should I just use ServiceTitan’s own AI voice agent instead? A: If you’re a ServiceTitan tenant on a single FSM and you can wait for the limited-release queue, evaluate theirs first — it reads live scheduling data, which no outside vendor can. Look elsewhere if you run multiple FSMs, want the phone layer independent of your FSM, or need coverage live now.

Q: Can a vendor connect to my ServiceTitan account without me? A: No. On their own, a third-party developer only gets ServiceTitan’s integration sandbox. Your ServiceTitan admin has to review the app’s requested scopes and generate the client ID and client secret before anything touches live data. If a vendor implies they can flip a switch on their side, they’re describing a connection you still have to authorize.

Q: How do I check whether a vendor’s ServiceTitan integration is real? A: Ask three things — is the app registered Public or Private and has ServiceTitan approved it, which read and write scopes does it request, and what does View Connections show for tenants your size. View Connections lists every tenant on the app with its version and live scopes. Anyone shipping in production answers all three quickly.

Q: Does InstaNexus have a native ServiceTitan or Housecall Pro integration? A: Direct API integration is on the roadmap. Today we run webhook + FSM-side automation where the tenant has webhook access and polling where they don’t. We will tell you straight what is shipped vs. planned on the demo call.

Q: Will the AI book an exact arrival time? A: No, and no honest vendor in this category does yet. Neither FSM exposes live tech availability via API. The AI quotes a 2-to-4-hour window and the dispatcher (or the FSM’s rules engine) confirms the slot.

Q: Do I need a specific Housecall Pro plan for webhooks? A: Yes — webhooks are gated to the MAX plan. On lower tiers you either upgrade or poll the API. Price the upgrade into the integration before you sign anything.

Q: What breaks most often in these integrations? A: Custom field changes on the FSM side. If a dispatcher adds a required field and nobody updates the integration, every new job fails silently. Audit the job-creation schema weekly for the first 90 days.

Q: How long does a direct API integration take to build? A: 2–4 weeks on Housecall Pro, 4–8 weeks on ServiceTitan, plus UAT. Webhook patterns go live the same week onboarding finishes.

Talk through the right integration pattern for your shop

If you run ServiceTitan or Housecall Pro and you want a straight answer on which integration pattern fits your call volume, dispatch team, and FSM setup — not a slide that says “native integrations” — book a 15-minute demo. We will walk through your current call flow, show where a voice AI plugs in, and tell you honestly which pattern we would run for a shop your size.

Book a free 15-minute demo →