It said 6. The real answer was 96.
A leadership team asked HubSpot's AI how many deals were committed this quarter. It said 6. The real number was 96. Luckily, someone checked.
That's where most HubSpot customers are right now, and it's a good place to start from. The first job isn't really AI at all. It's making sure HubSpot is clean and trusted enough for AI to give you the right answer.

Tick what's true today. Nothing is sent anywhere. It's just a quick way to check you're reading the right page.
Ticked three or more? You're at the start, which is the cheapest place to get this right.
Only one or two? Have a look at the stage before this one too.
Ask HubSpot's assistant about your pipeline and it reads your deals, contacts, notes and calls. If half your deals have no contacts and nobody records calls, it is summarising a relationship it has never seen. The answer comes back confident and wrong.
The other blocker is trust. Leadership worries about where the data goes. The team worries about what AI means for them. Both are fair, and both are solved with the same thing: clear settings, a human signing off, and a first win that obviously makes someone's week easier.
Fix what would mislead it, then switch it on.
Checked against HubSpot's knowledge base, November 2026.
From our own client work. Names left out on purpose.
A leadership team asked HubSpot's AI how many deals were committed this quarter. It said 6. The real number was 96. Luckily, someone checked.
One firm stopped all AI work because the CTO wasn't comfortable with where customer data went. What unblocked it was showing the setting that keeps their data out of model training, and exactly who could see what.
About half of one firm's deals had no contacts attached and calls weren't recorded. The AI could see the deal value and stage and nothing else, so every summary was thin.
We change nothing in your portal during the audit.
We connect with read-only access and pull your portal structure, data quality, AI settings and permissions.
Your portal scored out of 100 across the areas that decide whether AI gives the right answer.
A prioritised fix list, which AI features to switch on now, and which to leave alone for a while.
Two or three things your team will feel this month, fixed by your team or by us.
Everything is documented and owned by you. It keeps working whether or not we're still involved.
Published prices. Fixed before we start.
Your portal scored out of 100, read-only, with a prioritised fix list. The right first step for most teams.
New to HubSpot? Start clean, with AI in mind from day one.
Already know what needs fixing? We work through it in priority order.
You decide. HubSpot has an admin setting that controls whether your data can be used to improve its AI. We show you where it is and switch it off in our own portal. AI features work either way.
Usually not. Breeze Assistant and Context home are available on every HubSpot plan. Credits only come in when agents start doing work for you, which is the next stage.
Not in anything we build. The point is to take the copy-paste and the chasing off people's plates. A person approves anything that reaches a customer or changes a record.
It can help find the mess and suggest fixes, quickly. It shouldn't fix things silently. We set it up to flag problems and ask for confirmation, because bad data fixed confidently is worse than bad data left alone.
ChatGPT doesn't know your customers or your pipeline. HubSpot's AI does, as long as the data is right. Most teams end up using both, and that's fine.
Thirty minutes. Tell us how your team uses AI today and we'll tell you honestly what to fix first, even if it isn't with us.

We use cookies to see how people use the site and to make it better. Accept for full analytics, or decline and we will keep only anonymous, cookie-free basics. See our privacy policy.