AI agents inside Jobber

An AI agent for Jobber is a small program, connected through Jobber's official API, that watches your account for events (a new request, a quote that went quiet, a job closed with no invoice) and drafts the next step for someone on your team to approve. Jobber already automates some of this on its own. An agent earns its keep where Jobber's built-in rules stop: deciding which quote deserves a call, connecting Jobber to your phones and QuickBooks, and reporting across all of them.

This guide covers what Jobber already does, where the gaps are, exactly how an agent connects, and what should stay with a person.

What Jobber already automates

Start here, because some owners pay for an "AI agent" that duplicates a setting they already have.

Quote and invoice follow-ups. On the Connect plan and up, Jobber can send two automatic quote follow-ups for quotes still awaiting a response, up to 90 days after sending, to clients who have follow-ups turned on. Invoice follow-ups work the same way for past-due invoices. The same Automations page can auto-draft quotes from new requests and notify you about high-value quotes.

AI Receptionist. Jobber's own receptionist answers calls and texts, books visits and takes requests, texts back callers who hang up, and escalates urgent calls. Jobber's pricing page lists it as a $29 a month add-on, or included in the Plus plan. Connect starts at $99 a month billed annually.

Reports and Zapier. Plus-plan admins can describe a custom report and have Jobber AI build it. Jobber also connects to Zapier, which offers triggers such as Quote Sent, Quote Approved and Job Closed, and actions such as Create Client and Create Job.

If two fixed follow-ups and an answering service cover your problem, turn those on first. They're cheaper than anything we'd build.

Where the built-in tools stop

Jobber's automations are rules: same message, same timing, every quote. That's fine for a gutter cleaning. It's not how a good salesperson treats a full patio and retaining wall quote from a client who asked three questions about drainage.

The gaps we see most often in Jobber accounts:

• No judgment on follow-up. Two automated nudges, then nothing. Nobody decides which quiet quotes deserve a personal call this week.
• Phones live outside Jobber. If you keep your own phone system instead of the AI Receptionist, missed calls never become requests unless someone types them in.
• Money lives in QuickBooks. A job closed in Jobber and an unpaid balance in QuickBooks are two screens. Nobody compares them daily.
• One view for the owner. Jobber reports on Jobber. Ad spend, call volume and collections are somewhere else.

Small companies feel this most because the same two people answer phones, schedule and bill. The Census Bureau reported that fewer than 20% of firms with four or fewer employees use AI, and a Federal Reserve analysis put AI adoption at about 18% of firms as of year-end 2025. Most Jobber users are not behind their competitors yet.

What an AI agent does inside Jobber

Here is what three common agents watch, read, draft and write back. These are illustrations of the pattern, not a client story.

Agent	Starts when	Reads from Jobber	Drafts	Person approves	Writes back
Quote follow-up	A quote has sat in "awaiting response" for a set number of days	Quote lines, client notes, past jobs, any replies	A personal follow-up or a call reminder, ranked by value	Salesperson sends or edits	A note on the quote
Missed call to request	Your phone system reports an unanswered call	Client lookup by phone number, open requests	A callback text and a draft request	CSR sends, confirms request	New request or client note
Closed job, no payment	A job is closed and the balance is still open in Jobber or QuickBooks	Job, invoice, visit notes	Invoice reminder or a flag for the office	Office manager	Note on the invoice

A fourth agent sends the owner one morning message: new requests, quotes sent and won, jobs completed, money still open. That one reads but never writes.

How an agent connects to Jobber

There is no hidden back door. A legitimate agent uses the same public API as every app in Jobber's App Marketplace.

1. Authorization. Jobber's API uses OAuth 2.0, not API keys, and a company admin authorizes the app for that account. The app asks for scopes (what it can read or write). Ask for the scope list and refuse anything the job doesn't need.
2. The API. It's a GraphQL API with a single endpoint. Every request must name an API version; Jobber supports old versions for at least 12 months, so someone has to update the agent when versions retire.
3. Webhooks. The agent subscribes to events such as client created or invoice updated. Jobber's webhook rules matter in practice: the app must answer within 1 second, the same event can arrive more than once, and the message carries only an ID, so the agent has to fetch the record afterward. A well-built agent checks the signature on every message and ignores duplicates.
4. Rate limits. Jobber allows 2,500 requests per 5 minutes per app and account, plus a cost budget per query. An agent that re-reads your whole client list every hour will get throttled. Good ones read what changed.
5. Private vs published apps. A custom app built for your company can stay unpublished. Jobber's docs say a draft app is blocked if it connects to more than 5 paying Jobber accounts, so an agency using one unpublished app across many clients is a warning sign.

Connecting ChatGPT to Jobber

We didn't find a native ChatGPT or Claude connector in Jobber's documentation. The working paths are Zapier (fixed triggers into a ChatGPT step) or a custom app on Jobber's API. Zapier is fine for one step, like summarizing new requests into Slack. It gets fragile once you need memory, judgment across systems or an approval queue.

Setup steps and what you need

• A Jobber admin login to authorize the app.
• A list of the two or three jobs you want handled, in order. Usually quiet quotes first.
• Access to the other systems involved: phone provider, QuickBooks, ad accounts if you want them in the morning report.
• One person per agent who approves drafts for the first few weeks.

Expect the first agent to run with a person approving every draft. Loosen it only after you've read enough drafts to trust it. We explain that model in human-in-the-loop AI.

Limits and risks

• Duplicates. Because Jobber can send the same event twice, a careless agent texts a customer twice. Ask how they prevent it.
• Two systems texting. If Jobber's own follow-ups are on and an agent also follows up, the customer hears from you four times. Pick one per quote stage.
• Token and version drift. Access tokens expire and API versions retire. Someone has to own that, which is the reason for managed AI operations.
• Mistakes. Jobber's own help pages note that Jobber AI can make mistakes. So can ours. Pricing, discounts and promises to customers stay with a person.

Where Cortex fits

Cortex builds AI agents that work inside Jobber and the other software you run, with a person approving anything that leaves the building. We are not a Jobber partner or a certified app, and we'll tell you when Jobber's own automations are enough. See which systems we connect on integrations and how this works for home service companies. For the bigger picture of agents that read across your whole stack, read the AI operations layer.

The free AI Operations Audit looks at your Jobber account, phones and books in 30 minutes and tells you which job to hand an agent first.