We built an AI agent inside HubSpot. Here's what it caught.

We built a custom AI agent inside HubSpot that checks the health of every active client project and flags risk before a client ever raises it. Here's the problem it solves, how it works, and what it caught in its first real test.

What problem were we solving?

Unlike many other businesses implementing AI, our biggest problem wasn't messy data; it was time. We run a lot of client delivery projects at once, and nobody on the team had the hours each week to properly cross-reference every active project's health, budget burn, milestones, tasks and recent client conversations. Risk was often caught only once it was too late.

Why build it in HubSpot instead of a separate AI tool?

Because our delivery data already lives there. Companies, projects, milestones, tasks, timelogs, notes, emails and calls are all connected records in our own portal already. Building this as a Custom Agent in Breeze Studio meant it could reason directly across those real, live relationships. A standalone tool would've meant exporting all of that somewhere else and stitching it back together ourselves, which is exactly the kind of problem we didn't need to invent.

What did we actually build?

A Client Delivery Health Analyst Agent. Give it a client company and it reviews:

• Our own custom-designed health scores for that client's active projects, budget pacing, timeline pacing, milestones and scope creep
• The real budget numbers sitting behind those scores
• Every overdue or at-risk task, grouped by workstream
• The last month of emails, calls and meetings, for how the client is actually feeling

It writes a plain-English note straight back onto the company record, a clear red, yellow or green call, what's actually driving that result, and one thing the account manager should do next.

What results have we seen so far?

Only last week, it caught a project sitting 178% over its fixed-fee budget with two overdue milestones, and no signs of discontent from the client in recent email threads. What would have quietly become a profitability problem turned into an alarm bell nobody could ignore. That's the real value of an agent like this, auditing a volume of data no person has time to tackle themselves. It doesn't replace our account managers; it empowers them.

What's next?

We're working through running it across our whole client list, automatically, every week, alongside building more agentic automation and custom agents to push the needle on other internal pain points.

Frequently asked questions

What is a HubSpot Custom Agent?

A Custom Agent is an AI agent you build inside HubSpot's Breeze Studio. You give it instructions, decide what CRM data it can read or write, and choose whether it runs on demand or on a schedule.

What's the difference between a Custom Agent and HubSpot's Agentic Automation Builder?

A Custom Agent is the agent itself, its instructions, its knowledge, what it's allowed to do. Agentic Automation Builder is the workflow canvas that runs that agent at scale, chained together with regular automation steps.

Can I get a custom agent built for my business?

If you're running HubSpot and there's a process like this quietly eating hours nobody has spare, that's exactly what a custom agent is for. Get in touch and we'll help you build your own.