Universal CRM Connector for an AI Calling Agent

The challenge

Call Agent AI builds AI agents that answer and make phone calls for businesses. To be truly useful, the agent has to work with the customer records a business already keeps: who the caller is, their history, and what should happen next.

The problem was that every business uses a different CRM. Each one (HubSpot, Zoho and others) needed its own integration, written from scratch and running to hundreds of lines of code. Every new CRM meant more work, more code to maintain and a longer wait before a new customer could go live.

What I delivered

• A universal CRM connector. One shared approach to connecting CRMs, so adding a new one takes a small amount of CRM-specific setup instead of a whole new integration. It's built to support several types of CRM, including HubSpot, Zoho and Jobber.
• A complete Jobber integration. Jobber is a popular CRM for home-service and field-service businesses. I designed the architecture, built it and delivered it end to end, so the AI calling agent now works directly with a business's Jobber data. Just by talking with the caller, the agent can do the work a receptionist would: find or create the contact, offer available time slots and book the meeting, log the call and its notes, and handle invoices.
• Documentation for the team, so the engineering team can add and extend CRM integrations using the same pattern.

How it works

Most of what a CRM integration does is the same everywhere: sign in securely, find a contact, create or update a record, and keep the data in step. The universal connector handles all of that once, in a shared core. Each CRM then only needs a small, specific layer that explains how its data is shaped.

For the business, that means the AI agent behaves the same way whichever CRM they use. For Call Agent AI, it means a new CRM is a small, contained piece of work rather than a new project.

Key decisions

• Build the pattern, not just the integration. Writing a Jobber integration on its own would have solved one customer's problem. Building it on a universal connector solves the next ten.
• Keep the AI agent CRM-agnostic. The calling agent talks to the connector, not to each CRM directly, so it doesn't need to change when a new CRM is added.
• Deliver end to end. I owned the Jobber work from architecture to a working connection with the live agent, not just a prototype.

Built for trust

Businesses are trusting the connector with their customer records, so it only accesses the data the agent needs for the call, signs in through each CRM's own OAuth flow, so no CRM passwords are ever shared or stored, and keeps each business's connection separate.

The results

• A new CRM no longer means a new integration project written from scratch.
• Call Agent AI's AI agent can now work with Jobber, opening the product up to home-service and field-service businesses.
• The engineering team has one clear, documented pattern for every future CRM.