Most HubSpot portals are quiet. Data goes in when someone remembers to type it, and reports come out when someone builds them. In between, the CRM just sits there and waits.
That is starting to change. HubSpot did add AI features, but the bigger change is somewhere else. AI can now work on top of the CRM. It reads the data, updates records and tells people what matters, without anyone opening a dashboard.
At Hy Digital, we run our own agency this way. More than 30 AI agents handle our morning briefings, ad alerts, client updates and reporting. This post covers what we've learned, and what it looks like when you apply the same idea to your HubSpot.
By now, most HubSpot users have heard about the built-in agents. A customer agent can answer common questions on chat, and you can connect HubSpot to ChatGPT or Claude and ask your CRM questions in plain English.
These are useful, but they're features inside the tool. Someone still has to go and use them.
The bigger opportunity is AI that works across all your tools, on its own schedule, with HubSpot as the place where everything gets recorded.
HubSpot stays your system of record. A workflow tool such as n8n or Make moves information between HubSpot and everything else your team uses: WhatsApp, LINE, Gmail, Slack, your ERP or booking system. An AI model like Claude or GPT sits in the middle. It reads what comes in, decides what it means and writes the result back to HubSpot through the API.
You don't replace your CRM. You give it people who never forget to update it.
Before the sales meeting starts, the MD and the sales lead get a short summary in Slack, LINE or email. It covers which deals moved in the last 24 hours, which ones went quiet, and who needs a call today. Nobody builds a report. The AI reads the pipeline and writes the brief.
We run a version of this for our own team every morning. It's the workflow people miss most if it ever stops.
Most problems in a pipeline are visible weeks before they hurt. The trouble is that nobody's looking at the right moment. AI watchers can check for the patterns you care about, for example:
When a watcher spots one of these, the right person gets a message with the context attached. Our own ad accounts get checked every 2 hours this way.
In Southeast Asia, a lot of the real selling happens on WhatsApp and LINE, not in email. Those conversations rarely make it into the CRM, so the pipeline report shows a cleaner story than the real one.
AI can read those conversations, along with email threads and call notes, and pick out what matters: budget, timeline, decision maker and next step. It updates the deal in HubSpot and drafts a follow up for the sales rep to approve. The rep stays in control, and the data entry disappears.
We've built WhatsApp to HubSpot flows like this for clients in real estate and lead generation using Make. For LINE, which HubSpot doesn't connect to natively, we built our own bridge. You can read how Conxsys came to life.
Most CRMs record lost deals with a dropdown that says "price" or "timing". That rarely tells the full story.
Once a month, AI can read the notes and emails on every closed lost deal and group the real reasons, in the words customers actually used. Sales leaders get a short report that's far more useful than a pie chart of dropdown values.
When AI in the CRM disappoints, the AI model itself is rarely the cause. The problems usually start in the setup:
None of these are hard to fix, but they need fixing before the AI goes live, not after.
Our advice is to start small and visible:
That's the same order we followed when we built our own setup.
Start with a 20 minute call. Within 5 working days, you get a 1 page map of the 2 HubSpot AI workflows that would make the biggest difference for your business, plus anything in your setup that needs fixing first. It's free, and the map is yours whether you work with us or not.