HubSpot has released a public beta that I think could be far more significant than it sounds at first glance: your HubSpot conversation history is now searchable in Breeze Assistant.
That includes emails, live chats, and messages associated with your customer conversations.
The important part isn’t simply that Breeze can search them. It’s what this makes possible for HubSpot users who have historically had very limited ways to analyze the actual content of their conversations.
Conversation Data Has Been Surprisingly Difficult to Analyze in HubSpot
HubSpot has long offered reporting around tickets and customer service activity. You can report on things like ticket volume, response times, resolution times, ticket names and descriptions, categories, and other structured CRM data.
But that’s different from analyzing what people are actually saying.
Until now, if you wanted to perform meaningful analysis of conversation data itself, you generally needed to access that data through HubSpot’s API, which meant involving a developer.
For most HubSpot teams, that made questions like these much harder to answer than they should have been:
- What are customers asking us about most often?
- What questions keep coming up in website chat?
- Which conversations are turning into support tickets?
- What issues are causing customers the most confusion?
- Which knowledge base articles are actually being recommended in chatflows?
- What information should we teach a Customer Agent to handle?
With this update, you can start asking those questions directly in Breeze Assistant using plain language.
And you can get surprisingly specific.
How I Put the Breeze’s Conversation Search to the Test
I tested this with one of my clients because we have a very practical reason for wanting better visibility into their chat conversations.
We’ve been considering implementing HubSpot’s Customer Agent, but I haven’t wanted to turn it on simply because the technology is available.
Before handing more customer conversations over to AI, I want to understand what website visitors are asking about most, how well the existing content addresses those questions, and where we still have gaps.
This new Breeze capability gave us a much easier way to do that.
I first asked Breeze to analyze the visitors’ conversations with our rules-based chatflow year to date.
It was able to categorize those conversations, identify the types of questions visitors were asking, provide a thorough analysis, and make useful recommendations based on what it found.
That alone was valuable.
But then I kept going.
From a Simple Question to Hundreds of Rows of Usable Data
After reviewing Breeze’s initial analysis, I asked it to give me the individual chat questions behind the findings.
It started with a small sample of about 20 chats out of the 1,100 year-to-date, but I kept prodding for more. Eventually, I settled on a list of approximately 300 customer questions pulled from those conversations.
Then I wanted to know more about what happened within those chats.
When a knowledge base article was recommended during a conversation, which article was it, and why was it recommended?
Breeze was able to identify that information and add it to the spreadsheet.
But just as importantly, I wanted to know when the opposite was true: Was there no relevant knowledge base article available to answer the customer’s question at all?
Breeze could identify those gaps as well, and generate a newer version of the spreadsheet.
That distinction is incredibly useful. We’re not just looking at what customers ask. We can see where our existing knowledge base supports those conversations, what content is surfacing for particular questions, and where customers are asking questions our knowledge base doesn’t adequately address yet.
What started as a fairly simple request to analyze our chat history became a detailed dataset showing what customers are asking, how those questions can be categorized, which knowledge base content was being recommended in those conversations and why, and where there may be opportunities to create new content.
And I did it by having a conversation with Breeze Assistant.
No API. No developer. No custom integration required.
This Is Where Breeze Becomes Much More Useful
There has been a lot of attention on what AI can create inside HubSpot.
I’m equally interested in what it can help us understand.
Companies already have enormous amounts of valuable customer language sitting inside their HubSpot portals. It’s in live chats, support conversations, emails, and tickets.
The problem has been getting useful information out of it at scale.
Making conversation history available to Breeze Assistant changes that.
Instead of relying entirely on fields and properties that someone decided to structure in advance, we can start asking questions of the unstructured conversations themselves.
For a customer service team, that might mean identifying recurring support issues or topics that frequently lead to escalation.
For marketing, it could mean finding the exact language customers use when describing their needs.
For sales, it could mean analyzing common objections or understanding how different customer segments talk about a particular problem.
And for a HubSpot admin or consultant, it opens up an entirely new source of information for deciding what should be automated, documented, measured, or improved.
It Could Also Make Customer Agent Implementations Much Smarter
This is probably the part I’m most excited about.
I’ve been deliberately cautious about implementing Customer Agent for this particular client because I don’t think the right approach is simply to turn on an AI agent and see what happens.
I want to know what customers are asking first.
Now we can use actual conversation history to identify the questions a Customer Agent is most likely to encounter, understand which knowledge base articles have historically been useful in answering those questions, and, perhaps most importantly, identify questions for which we don’t currently have appropriate knowledge base content at all.
That gives us a much clearer roadmap for what needs to happen before we put Customer Agent in front of customers.
We can strengthen the knowledge base where necessary, identify the topics the agent should be prepared to handle, and establish a baseline based on real customer conversations rather than assumptions.
Once we begin testing Customer Agent, we’ll also have real historical data against which we can evaluate its performance and continue improving the content that supports it.
In other words, conversation search doesn’t just tell us what’s happened in the past. It can help us build a much more informed AI strategy going forward.
How to Search Your HubSpot Conversations With Breeze
The feature itself is simple to use.
Open Breeze Assistant and ask a question about your customer conversations in plain language. Breeze searches the conversation history available in HubSpot and synthesizes an answer based on what it finds.
I wouldn’t stop with the first answer.
Ask it to break the results down further. Ask for the underlying questions. Ask it to categorize them. Ask for a spreadsheet. Ask which knowledge base articles were recommended in those conversations and why. Ask it to identify questions where no relevant article was available.
That’s where I’ve found this update becomes especially useful.
Available Now in Public Beta
HubSpot says conversation history search in Breeze Assistant is available as a public beta across all HubSpot hubs and tiers.
For me, this is one of those HubSpot updates that removes a technical barrier that probably shouldn’t have existed in the first place.
The data was already there. We just couldn’t easily interrogate it.
Now we can.
If you’re using live chat, tickets, or other customer conversations in HubSpot, I’d start asking Breeze some questions. You may find that the conversations already sitting in your portal can tell you a lot about what your customers need next.