How to Configure Core AI Tools in HubSpot

HubSpot now ships with a serious layer of AI: an assistant embedded in every record, properties that research and fill themselves in, and agents that can work autonomously across your account. The problem isn't capability — it's that almost no team has been shown how to use any of it well. "Turning on AI" ends up meaning a few people trying the assistant once and forgetting it exists.

Configuring your core AI tools well is what turns that unused capability into a CRM that works for you between the moments your team touches it. Here's what it actually involves, what to decide before you start, and where teams get it wrong.

What "core AI tools" actually means

There are three layers, and each does a different job.

The everyday assistant is an AI copilot built into HubSpot. It summarizes an account before a call, drafts and replies to emails, writes content in context, and answers "what's going on with this account?" — all without leaving the record. It's the easiest, fastest win, and it's the one your whole team should be fluent in.

Smart properties are CRM fields that fill themselves in. Instead of a person researching whether an account is expanding, whether they mentioned a competitor, or what the short version of a long account history is, you give the AI a plain-language instruction and point it at your data — your CRM records, your call and meeting transcripts, your emails, and the live web — and the property populates itself. This is what turns your CRM from a passive database into something that surfaces the insight you care about, automatically, across every record.

Agents and agentic workflows are AI that does real, repetitive work on its own — researching, summarizing, triaging, routing. HubSpot offers ready-made agents you can customize, and (on the right plan) a builder for creating your own agents and multi-step automations for jobs no off-the-shelf feature covers.

Why it's worth doing deliberately

Done well, this is one of the highest-leverage things you can set up in HubSpot. The payoff shows up as time your team stops spending on work the platform can now do:

Insight surfaces on its own — expansion signals, risk signals, fit — without anyone reading every transcript
Repetitive internal work gets handled by agents instead of people
Your team drafts, summarizes, and researches in seconds with the assistant
The AI understands your business, so its output sounds like you rather than generic marketing copy
Your AI usage stays affordable, because it's budgeted and capped rather than left wide open

That last point matters more than it sounds. HubSpot's AI runs on credits — a usage-based allowance — so the difference between a great AI setup and an expensive surprise is entirely in how it's scoped and governed.

Where teams go wrong

A few patterns we see over and over:

Treating "set up AI" as unlimited. The most important decision isn't technical — it's how far to take this. A focused set of high-value wins is a very different project from an ever-expanding library of agents. Deciding that up front is what keeps the effort (and the cost) sane.

Ignoring the credit model until the bill arrives. Every agent run and every smart-property run consumes credits — and smart properties can bill even when they don't find a value to fill. Estimating usage, setting run limits, and scoping which records the AI runs against is the whole discipline. Skip it and AI becomes something people are afraid to use.

Automating on AI output no one checked. AI-generated values are close-but-wrong often enough that building automation on top of unreviewed output erodes trust quickly. The right move is to run it, read it, correct the instruction, and then wire it into a workflow.

Letting the AI guess about your business. Agents and drafting tools inherit whatever context you give them. Grounded in your real brand voice, customers, and positioning, the output is genuinely useful; grounded in nothing, it reads generic — which is especially costly in trust-heavy fields like payroll and HCM.

How we approach it

We treat this as a scoping decision first and a build second. Before touching anything, we size your actual appetite for AI and confirm your credit plan — so the setup matches what you need and stays affordable. Then we work in order: enable the underlying AI settings, review your existing properties to see which the AI can fill and build the smart properties worth having, configure the agents and automations that fit your real work, and train your team on the everyday assistant so it becomes a habit.

We also keep this milestone deliberately "core." The specialized agents — the one that researches and prospects for sales, and the one that resolves customer support conversations — are set up separately, because each deserves its own focused build. What this milestone gives you is the foundation everything else stands on: the AI turned on, grounded in your business, governed on cost, and actually used.

The goal isn't "AI is enabled." It's a CRM that quietly does more of the work for you, and a team that trusts it enough to rely on it.

If you'd like to learn more about how to configure core AI tools in HubSpot, contact The Gist.