I'm explaining it to someone else · 5 min read
The words that survive the room — explaining HubSpot AI to whoever decides
You're a HubSpot admin, a manager or an ops lead carrying HubSpot's AI to the person who signs off on it, and you need the words that hold up when they push back.
If you have five minutes
- Whoever decides is asking three questions, and none of them is "what does Breeze do now."
- They're asking what it costs to run, what happens if it's wrong, and whether the name you're using is even the current one.
1 of 4
The room you're walking into isn't asking what the feature does
Whoever decides is asking three questions, and none of them is "what does Breeze do now." They're asking what it costs to run, what happens if it's wrong, and whether the name you're using is even the current one. Get any of the three wrong and the rest of the pitch doesn't land.
2 of 4
Say this, not that
- Say: "HubSpot's AI" or "HubSpot's AI tools," not "Breeze."
- HubSpot's own terminology guidance, effective July 23, 2026, retired Breeze as the umbrella term. Breeze Assistant is still Breeze Assistant — that name didn't move. But "Breeze" as a catch-all for the whole AI layer is now the old term, and using it in a room where someone read the same memo you didn't reads as not current.
- Say: it inherits your permissions, it doesn't grant new ones.
- The honest framing for anything that sounds alarming — Breeze Assistant writing a payment link from a sentence, for instance — is the guardrail, not the capability: "But Breeze can only do what the user already has permission to do, so this does not hand anyone new power, it removes the manual clicking from power they already had." That's the sentence that gets a security-minded stakeholder to relax.
- Do not say: "just ask Breeze."
- Not without naming what changed to make that true. Trust in an AI answer breaks the moment sourcing disappears: "Hard to trust the responses if you're not telling me where they came from." If the surface you're pitching doesn't cite where its answer came from, say so, plainly, rather than asking someone to trust it on faith.
- Ask for the test, not the rollout.
- The honest pattern across nearly every take in this archive is the same: treat a new AI capability as a test before it's a rollout. On Breeze Assistant's quote approval actions, the caution is specific — "I think it loses value really quickly if you still have to click into a different area and go and research context to try and get the thing done." Ask it once, see if it actually holds up, and only then retire whatever manual process it's meant to replace.
- Name the credit budget before someone else finds it.
- Custom agents stopped being free on Thursday, July 23, 2026, and started drawing down a shared credit budget. The read on record is direct: "A capability expansion and a pricing change landing in the same week is not a coincidence." If you're proposing an agent that runs at volume — inside a workflow, for instance — say up front that it now has a running cost attached, not just a running risk.
3 of 4
Old term, current term — the vocabulary card
| HubSpot called it (before Jul 23, 2026) | HubSpot calls it now | What survives unchanged |
|---|---|---|
| Breeze (as the umbrella) | HubSpot's AI, HubSpot's AI tools, or HubSpot's collection of AI tools | — |
| Breeze Agents | HubSpot's agents, HubSpot's AI agents, AI agents, or simply "agents" | — |
| Prospecting Agent, Customer Agent (capitalized) | prospecting agent, customer agent, closing agent, etc. (lowercase) | — |
| Breeze Assistant | — | Breeze Assistant, unchanged, capitalized |
| Agent Hub | — | Agent Hub, unchanged, capitalized — "the home base in HubSpot for managing, building, and tracking agents" |
4 of 4
Five questions they will ask, and the honest answers
- "Isn't this just Breeze with a new coat of paint?"
- No — the umbrella term changed because HubSpot decided "Breeze" was doing double duty as both a product name and a category name, and that was confusing its own customers. Breeze Assistant is exactly what it was. The agents underneath it are the part that's actually multiplying.
- "What does this cost us to run?"
- Credits, as of July 23, 2026 — before that date, custom agents were free to run. If you're proposing anything that runs an agent at volume, budget for it the way you'd budget for any metered API, not as a flat license fee.
- "Who's actually using this for something real?"
- Be honest about where adoption is thin. On AI-assisted invoice corrections, the hosts said it plainly on air: "We would happily invent a prize for the first person who even hears of somebody using Breeze Assistant to manage invoices in any real way." Some of what's shipped is capability without adoption yet, and pretending otherwise costs credibility the first time someone checks.
- "What happens when it gets something wrong?"
- Depends what "it" is. An agent-builder draft that's wrong gets reviewed and rebuilt before anyone acts on it — that's the "plan first, generate second" discipline the hosts asked HubSpot to build directly into the product. A customer agent that surfaces a coaching gap ("Action Gap") isn't reporting a bug in the agent; the read on record is that "if there's a gap to be found like this, it often relates to there being a real process gap" — the fix is usually a human process, not an AI patch.
- "Do we need to train people on this?"
- Yes, and it isn't a document or a video. The take on Prompt Sharing in Breeze Assistant is specific about what enablement actually looks like: "As a HubSpot admin or manager, your job needs to be get people comfortable having conversations with Breeze." The example on record: someone new at a company had never seen Breeze Assistant at all because the person who used to own HubSpot there had left. The fix wasn't more documentation — it was writing a few good starting prompts and handing them over directly.
The line to hold onto.
None of the wins described in this archive came free. Every one of them required a human to fix trust, cost, or setup before the AI did anything worth keeping. That's not a knock on HubSpot's AI — it's the honest shape of what "AI works here" actually means in 2026, and it's a better pitch than pretending otherwise.
Where to next
Where this comes from
- HubSpot's own terminology guidance, effective July 23, 2026. the vocabulary card is HubSpot's own old-term / current-term table, quoted
- Breeze Assistant Support for Payment Links Creation, covered on air Jul 21, 2026. live; the permission-inheritance sentence this door recommends saying out loud
- Breeze in Slack: A better first impression, clearer help, and cleaner citations, covered on air Sep 3, 2026. live; why sourcing is the trust question, in Chris Carolan's own words
- Breeze Assistant: Quote Approval Actions, covered on air Aug 27, 2026. live; the caution about a capability that still makes you leave to find context
- Use Webhooks in custom agents, covered on air Jul 20, 2026. private beta; the capability-and-pricing-in-one-week read behind the credit-budget answer
- AI-assisted invoice corrections with Breeze Assistant, covered on air Aug 14, 2026. private beta; the honest adoption answer, said on air
- New Customer Agent Coaching Opportunity: Action Gap, covered on air Aug 28, 2026. live; why a coaching gap is usually a process gap a human owns
- Prompt Sharing in Breeze Assistant, covered on air Aug 20, 2026. public beta; what enablement actually looks like, and the example on record
What we do not know yet
- Whether the credit-cost framing above lands the same way for a business already deep into Revenue Hub or Service Hub agent usage hasn't been tested in this archive — the take on webhooks and credits speaks in general terms, not against a specific portal's usage pattern.
- The archive has no on-record moment of a controller, a security lead, or a compliance officer pushing back on any of this in the room — every take quoted here comes from the hosts anticipating the objection, not from someone who actually raised it live. Treat the "five questions" section as well-informed anticipation, not a transcript of a real pushback.
Mid-build on an agent, or running one of the connectors that is going away?
Come walk through it on the show and we will work it out on air, weekday mornings. The record's own advice on the connectors leaving is exactly that — a conversation is more useful than a changelog entry.