Home services and trades
What ServiceTitan's Max and Atlas launches mean for contractors' data
By SourceX Editorial · Updated
Short answer
ServiceTitan's Max and Atlas launches mean a contractor's own job, call and customer records will increasingly drive AI agents inside the platform. For owners, four data questions matter most: which features are live or still in pilot, which records the agents read, whether your data is used beyond your account, and how you keep a copy you control.
Key takeaways
- AI agents inside a field service platform work from the contractor's own records, so gaps in those records show up in front of customers.
- Generally available, early access, pilot and roadmap mean different things; confirm each feature's status in writing before planning around it.
- Call summaries, drafted estimates and AI-written notes are new records, and your contract should say who owns them and whether they export.
- Using a vendor's AI and licensing your own records are separate decisions governed by separate terms.
What ServiceTitan announced, and what to verify yourself#
ServiceTitan presented Max and Atlas at its 2026 Pantheon conference as part of a broader push to put AI agents to work inside the platform contractors already run on. Launch coverage describes Max as a group of agents drawing on each customer's own data, an approach vendors often call a company brain.
This analysis deliberately does not restate feature lists, agent counts, prices or release dates. Those details come from launch-week materials, and names, packaging and availability tend to shift after a conference, with some capabilities shown on stage limited to pilots in certain trades. Before you plan around anything, check ServiceTitan's release notes, ask your account manager, and read the order form or AI addendum that would govern it. The table below is a way to sort what you hear.
| Status label | What it usually means | What to ask |
|---|---|---|
| Generally available | Released to customers on a qualifying plan or add-on | Which plan or add-on, at what cost, and under which terms? |
| Early access or beta | Open to selected customers and likely to change | What happens to data and outputs created during the beta? |
| Pilot | Tested with a small group, often one trade or region | Are we eligible, and are pilot terms different from the main agreement? |
| Announced or roadmap | Planned, not yet in customers' hands | Is there a committed date in writing, or only an intention? |
Why agents raise the stakes on records you already keep#
Agents raise the stakes because they act on your records in front of customers instead of sitting in a report a manager reads later. A stale pricebook item that once only skewed a margin report can become a price quoted on the phone, and a lapsed membership still marked active can become a promise of priority service.
The table lists records an agent is likely to touch and the check worth running before it does. None of it requires new software; it is housekeeping that also improves your own reporting.
| Record | What an agent might do with it | Check before switching it on |
|---|---|---|
| Pricebook and price rules | Quote or suggest prices on calls and estimates | Retire items you no longer sell and confirm current prices |
| Membership status | Offer priority booking or member discounts | Lapsed and canceled plans are marked correctly |
| Customer and location notes | Summarize a home's history for a CSR or technician | Gate codes and personal remarks are not read back aloud |
| Call recordings and transcripts | Learn how your team books, triages and escalates | Your greeting covers AI analysis, not only recording |
| Equipment history | Flag aging systems and suggest replacements | Install records are followed by later service visits |
| Technician notes | Draft visit summaries and follow-up tasks | Free text does not contain remarks you would not want repeated |
- 01Your work records
- 02Personal details removed
- 03You approve the release
- 04Licensed to AI developers
Five questions to ask before turning agents on#
Five written questions decide who controls your data once agents are working on it. Send them to your account manager, ask for answers in writing, and file the replies with your contract so the next person who reviews the account can find them.
- Which of our records does each agent read, and can we see or limit that list agent by agent?
- Is our data used only to serve our account, or also to train or improve models used for other customers, and can we opt out?
- Who at the vendor can see agent outputs and logs, such as call summaries, generated notes and recommendations?
- Are agent-generated records part of our data, and are they included in a full export if we leave?
- Does switching agents on change what we must tell customers, for example on recorded or AI-handled calls?
Agent outputs are new records in your job history#
Agent outputs are new records that did not exist before: a call summary, an automatically drafted estimate, a suggested dispatch or a note written from a technician's voice memo. Over time they will make up a growing share of your job history, which is why ownership and export of those outputs belong in the contract, not in a support ticket.
They also change what your history means. A record written by a person shows a human decision; a record drafted by an agent and accepted with one click shows something different. Ask whether outputs carry a label or audit field that marks them as machine-generated, and whether that label survives an export. Anyone who later relies on the archive, including your own managers, an acquirer in diligence or an outside AI developer, will want to tell the two apart.
Who controls the records: the contract, not the keynote#
Control over your records is set by your subscription agreement and any AI-specific terms, not by what is said on stage. Field service agreements commonly state that the contractor owns its customer data and that the vendor processes it to deliver the service. AI features may arrive with an addendum or updated terms that add new permissions, such as using aggregated or de-identified data to improve the vendor's products.
Read any click-through acceptance before an admin approves it on the company's behalf. Look for the definitions of customer data and outputs, any product improvement or model training clause, how to opt out, and whether opting out switches the feature off. If the language is unclear, ask for a written clarification rather than relying on a sales conversation.
Keep a copy of your history that you control#
A regular export of your ServiceTitan history is the simplest protection while AI features change around you. It preserves your records if you switch platforms, gives you a baseline to compare agent results against and keeps options open for uses you decide on later.
Ask your account manager which export routes your plan includes for full history and for attachments such as call recordings, photos and forms, since standard reports often leave those out.
- Keep native record IDs so jobs, invoices, estimates, calls and equipment can be joined again.
- Include agent outputs, and the field that marks them as machine-generated if one exists.
- Note the export date and which AI features were switched on at the time.
- Store the export in storage your company controls, with access limited to named people.
- Test the export on one month of data before relying on it, then set a quarterly schedule.
Illustrative: an HVAC owner reviews an AI addendum#
Illustrative: the fictional owner of a residential HVAC and plumbing company on ServiceTitan wants AI help with inbound calls before the heating season. An admin is about to accept updated AI terms in the account, and the owner asks the operations manager to review the data side first.
The review turns up three issues. The terms allow aggregated data to be used for product improvement, and no one knows whether an opt-out exists. The phone greeting says calls are recorded for quality only, with no mention of AI analysis. And nobody can say whether agent-written call summaries would come out in an export.
The owner sends the five questions to the account manager, has counsel review a new greeting and holds acceptance until the written answers arrive. The company then starts with internal call summaries for managers, keeps customer-facing agents for a later phase and adds agent outputs to its quarterly export.
Where SourceX fits, and where it does not#
SourceX does not build or sell field service AI. Its role is separate: helping established contractors license de-identified job, call and equipment histories to AI developers, under terms the contractor approves.
Using a platform's agents and licensing your own records can coexist, but check your agreement for any restriction on third-party use or any AI addendum that affects exported data. Licensing runs through the SourceX five-step transaction, Supply, Rights, Preparation, Approval and Delivery, and the SourceX Evidence Packet records which systems and terms the records came from and which records were machine-generated. The data is licensed, not sold.
Frequently asked questions
Does using ServiceTitan's AI mean ServiceTitan owns my data?
Not by default. Ownership is set by your agreement, and field service agreements commonly keep customer data with the contractor. AI features can add permissions for the vendor, such as product improvement, so read any AI addendum or updated terms before accepting them, and ask how to opt out.
Should I wait until pilots finish before turning anything on?
That depends on how much risk you accept on customer-facing work. Many owners start with internal uses, such as summaries for managers, before letting agents talk to customers. Whatever you choose, confirm each feature's status in writing and what happens to data created during a pilot.
Will AI agents make my historical records more or less valuable?
Clean, linked history becomes more useful once agents rely on it, both inside your business and to outside developers. Records created by agents need clear labels, though, so anyone using the archive later can tell human decisions from machine suggestions.
Do I have to tell customers when an AI agent handles their call?
Disclosure rules for recorded and AI-handled calls vary by state and are still developing. Review your greeting, privacy notice and booking forms with counsel before an agent answers or listens to customer calls, and keep a dated record of each disclosure version.
What should we clean up first before agents rely on our data?
Start with the records agents touch on every call: the pricebook, membership status, call dispositions and customer notes. Archive items you no longer sell, fix lapsed memberships, make booking dispositions required and remove access codes from free-text notes. Those fixes improve your own reporting even if you never turn an agent on.
Related resources
See if your company qualifies
A short company assessment. No data uploads are needed.