Show recap

Voice in Breeze, Categorize Record, Agents on Mobile, and a Lead Score Built Live with Breeze

Thursday, November 19, 2026 · with Chris Carolan, Casey Hawkins, and Trisha Merriam

From the show

Chapters

  1. 02:04Good morning, HubSpot Nation: Thursday, November nineteenthTopic
  2. 02:29Trisha is back from sunny CaliforniaTopic
  3. 02:46Last week's big news: the day the CEO changedTopic
  4. 02:59Caught in the ground stop, rebooking flightsTopic
  5. 03:13A peek at today's updates: there are a lot of themTopic
  6. 03:35Today's events at anotherorangemorning.com/eventsTopic
  7. 03:46Office Hours today: the Unified Revenue View with Ryan GinsbergTopic
  8. 04:12Do your lifecycle stages mean anything?Question
  9. 04:22The answer: probably notTopic
  10. 04:37The demo portal's value stages, and some beefTopic
  11. 05:02Hover over a lifecycle stage to see what it meansTopic
  12. 05:23Tooltips on properties from the HubSpot documentation field?Question
  13. 05:44Move the documentation where the team can see itTopic
  14. 05:56An open invitation to talk value path stagesTopic
  15. 06:08Some big ones on the board todayTopic
  16. 06:30Voice in Breeze Assistant: isn't that just the microphone?Question
  17. 06:57What's new: talk to Breeze Assistant on desktop, and it talks backTopic
  18. 07:23Voice typing, or a two-way conversation?Question
  19. 08:10Why does it matter? Say what you need and keep movingQuestion
  20. 08:39Resistant to voice AI: my fingers workTopic
  21. 08:58Talking the Hubsessed newsletter through with ClaudeTopic
  22. 09:23The best AI-assisted writing comes from back-and-forth conversationTopic
  23. 09:37Whose technology runs the conversation?Question
  24. 09:55When voice AI interrupts you constantlyTopic
  25. 10:12Breeze Assistant desktop?Question
  26. 10:36It adds two-way calling to Breeze: HubSpot on desktop, not a desktop appTopic
  27. 11:17Letting another AI tell Breeze to do stuffTopic
  28. 11:26Three-way calls between multiple agentsTopic
  29. 11:58Natural language with AI: it gets to know you betterTopic
  30. 12:16From "Have you asked Breeze yet?" to "Have you talked to Breeze yet?"Topic
  31. 12:29Categorize Record workflow action, in private betaTopic
  32. 12:43What is it? An AI action that sorts CRM records into categoriesQuestion
  33. 13:13Why does it matter? Classify and route records without manual effortQuestion
  34. 13:26How does it work? Can you explain it to us?Question
  35. 13:38All of those ridiculously branched workflows: goneTopic
  36. 13:43Choose the record, set the categorization rulesTopic
  37. 14:06Assignment by industry and state, without the branches?Question
  38. 14:28Route by why someone contacted youTopic
  39. 14:48Anything you've ever branched a workflow forTopic
  40. 15:10Have a conversation with Breeze about the rulesTopic
  41. 15:39Where custom coded workflow actions came to the rescueTopic
  42. 16:13The agent step can already build the branching. So how is this different?Question
  43. 16:36You don't have to build the branch?Question
  44. 17:03Do you have to use branches anymore?Question
  45. 17:21It moves the complexity upTopic
  46. 17:37Ticket triage is the obvious one; assignment is the hurdleTopic
  47. 17:55When the territories keep shiftingTopic
  48. 18:08Reviewing AI magic when there are no branches to checkTopic
  49. 18:53A workflow becomes one boxTopic
  50. 19:10How do you vet that?Question
  51. 19:17Ask Breeze to explain the underlying logic and confirm itTopic
  52. 19:28Why the humans involved shouldn't opt outTopic
  53. 19:53Except for this situation: where the rule doesn't applyTopic
  54. 20:09Seventeen edge cases that aren't really edge casesTopic
  55. 20:20Turn the transcript around: does this look right?Question
  56. 20:44You need a human making a call somewhereTopic
  57. 21:09At some point the human brings in judgmentTopic
  58. 21:18Moving faster from existing workflowsTopic
  59. 21:42New product word alert: Platform Professional and EnterpriseTopic
  60. 22:01Introducing agents in Breeze Assistant on mobileTopic
  61. 22:11What's new? Delegate to agents from the HubSpot mobile appQuestion
  62. 22:57Why does it matter? Agents need answers and approvals away from the deskQuestion
  63. 23:41True story: the top sales rep who is always in her carTopic
  64. 24:22"Don't ask her to do anything on her phone"Topic
  65. 24:41Now she can be hands-freeTopic
  66. 24:55Pull up the prospecting agent, out loudTopic
  67. 25:20Are you not giving Casey a turn today?Question
  68. 25:27I wasn't even supposed to be on todayTopic
  69. 25:35Use Breeze Assistant for lead scoringTopic
  70. 25:57What is it? Breeze can create, edit and analyze lead scoresQuestion
  71. 26:32Why does it matter? Breeze and the lead scoring app weren't connectedQuestion
  72. 26:52Now Breeze is fully connected to lead scoringTopic
  73. 27:11On vacation, and less than twenty-four hours with itTopic
  74. 27:24The advice: go in with an open mindTopic
  75. 27:33Have a conversation with Breeze instead of dictatingTopic
  76. 28:13It uses your HubSpot context, so fill that out firstTopic
  77. 28:43Ask: where did you get that from?Question
  78. 29:03Describe your ideal output: we want more people like thisTopic
  79. 29:15Good sales conversations can be easier to isolate than customersTopic
  80. 29:48A client asked about this the same dayTopic
  81. 30:14Doesn't it already do this?Question
  82. 30:46The big difference: it will now build it for youTopic
  83. 31:00Why did this contact score?Question
  84. 31:32The last thing they did that added to the scoreTopic
  85. 31:53Taking notes on what Casey's sayingTopic
  86. 32:05Speak now or forever hold your peaceTopic
  87. 32:23Will It Breeze: the update, pasted with no other contextTopic
  88. 32:30Build a new score or refine an existing one?Question
  89. 32:37Contacts or companies?Question
  90. 32:57Breeze checks the customer profileTopic
  91. 33:07Which team should the score focus on? Operations teamsQuestion
  92. 33:30Which outcome should the score aim at? A customerQuestion
  93. 34:07Breeze proposes a score capped at a hundred pointsTopic
  94. 34:16Breeze: each signal shows up on only three contactsTopic
  95. 34:29Ask the queen of scoring, Casey HawkinsQuestion
  96. 34:40What data was used? What were the filters? Who were the examples?Question
  97. 34:59Too quick to an out-of-the-box answerTopic
  98. 35:42How did you come to this suggested draft?Question
  99. 35:52We're on a live show: no customer names or sensitive dataTopic
  100. 36:10Breeze: nothing's been created or activatedTopic
  101. 36:27Breeze: one form and one page visit, three contacts each, not strong evidenceTopic
  102. 36:39Breeze: the weights were starting assumptions, not learned from your dataTopic
  103. 36:46Breeze: buying intent, or people who are already customers?Topic
  104. 36:56Now that we're thinking critically, what are the next steps?Question
  105. 37:12Breeze: validate the signals, then test against another sampleTopic
  106. 38:17Breeze checks its assumptions around buying intentTopic
  107. 38:30Breeze revises the idea to start with fit onlyTopic
  108. 38:44Breeze: a score preview by itself wouldn't prove anythingTopic
  109. 39:13Operations Teams Fit Baseline: an unpublished draft, up to sixty points for fitTopic
  110. 39:30Can you suggest some records for testing?Question
  111. 40:04Breeze picks five anonymous recordsTopic
  112. 40:25Breeze: good for checking the rules, not predictive accuracyTopic
  113. 40:34Can you link me to the unpublished draft?Question
  114. 41:02Breeze keeps the results anonymous for the showTopic
  115. 41:13Breeze: we're missing some title variationsTopic
  116. 41:20A rules check, not proof it predicts anythingTopic
  117. 41:26Thanks, Breeze: a great starting pointTopic
  118. 41:43Breeze ended the call on her ownTopic
  119. 41:55Start talking to Breeze, pleaseTopic
  120. 41:57If you need examples, you'll find this clip hereTopic
  121. 42:07New: HubSpot community questions you can turn into promptsTopic
  122. 42:20Start that conversationTopic
  123. 42:25Have a great rest of your ThursdayTopic

Another Orange Morning Recap: Thursday, November 19, 2026

Chris Carolan, Trisha Merriam and I were back together this morning, and it was a good one to come back for. Trisha missed last week while she was in sunny California, and she got caught up in the ground stop on the way home. I was not even supposed to be on today. I am technically on vacation. Then I saw what was on the board.

Here is what we got into: voice in Breeze Assistant, the Categorize Record workflow action, agents in Breeze Assistant on mobile, and the one that made me glad I showed up, Breeze Assistant for Lead Scoring. Then Chris had Breeze build a lead score with us, out loud, on the show.

Office Hours: The Unified Revenue View

Today's Hubsessed Office Hours is Chris and Ryan Ginsberg on the Unified Revenue View, the thing you need to happen whether you are using AI or not, and how easy it is to do in HubSpot when you can ask Breeze to do it for you. You can find it on the events calendar.

Trisha and I also talked about our session where we asked whether your lifecycle stages mean anything. We used the demo portal, which has Chris's value stages set up, and Trisha had some beef with them. The upside was a great on-screen demo of the new documentation functionality: when you are confused about what a lifecycle stage means, you mouse over it and the documentation is right there. If your documentation is filled out, people do not have to be confused anymore. Chris has an open offer to Trisha to talk value path stages any time she wants.

Four Updates, Mostly Breeze

Voice in Breeze Assistant

Trisha picked this one, because she uses voice for everything, and the first question was fair: isn't that already just the microphone? It is not. This is a two-way voice conversation with Breeze Assistant on desktop, already on mobile. Breeze pulls the CRM context, answers out loud, asks questions when it needs more detail, and confirms CRM changes before it makes them. It is in public beta, expected November 22.

I will admit I was resistant to voice AI for a while. My fingers work, I'm good. But over the last few months I switched, and the switch was writing the Hubsessed newsletter. I just talk to Claude, say my piece, and Claude makes it coherent. Some of my best AI-assisted writing has come out of those back-and-forth conversations.

Trisha's question was whose voice technology HubSpot is using, because the last time she tried talking to Claude it interrupted her constantly and she could not get a thought out of her head. And no, as far as I know there is no HubSpot desktop app. "Desktop" here just means HubSpot in your browser versus the mobile app. Chris is already letting his own GPT agent tell Breeze what to do, and he is hoping for three-way calls between agents soon, for the same reason I like talking things through out loud.

Why it matters: CRM work means switching screens, hunting for details and typing updates. Prep for a meeting, log the notes and create the follow-up tasks in one conversation instead. And "have you asked Breeze yet?" is about to become "have you talked to Breeze yet?"

For more details: the voice in Breeze Assistant update.

Categorize Record Workflow Action

Chris was excited about this one. It is an AI-powered workflow action, in private beta, that sorts records into categories based on rules you write: ticket triage, lead classification for sales, spotting upsell opportunities from product usage. You can follow it with a branch, or with an edit record action that sets a property. His summary: all of those ridiculously branched workflows, gone.

My first thought was lead assignment. You know the workflow: if this industry and this state, go to this person, unless it is that other state. Instead you explain the rules once and let the action do the routing. Trisha added the service version: route everyone who picks "I have an issue with the platform" straight to customer service. I have built lead assignment by territory and watched the territories keep shifting until we just did it by state, because that was easier to adjust.

Here is the part that makes me uncomfortable, and it is something Chris and I talked about yesterday on The Road to UNBOUND. My review process is built for a workflow where I can see every branch and check that each one is built right. When I'm relying on AI magic, all I can do is run test scenarios and check what came back. That is an adjustment.

Trisha's head went to the same place: we are headed toward a workflow that is one box with a lot of information in it, so how do you vet that? Probably by asking Breeze to explain the underlying logic and confirm it for you. Chris's answer was that the humans involved should never opt out. Trisha named why: people "forget about the seventeen edge cases that are not really edge cases because they're happening a good amount of the time." Somewhere a human still has to make a call.

Why it matters: It moves the complexity up front, so you describe what you care about instead of building every branch first. Chris also caught a new product word on this one, Platform Professional and Enterprise, which made me a little uncomfortable.

For more details: the Categorize Record update.

Agents in Breeze Assistant on Mobile

Not to be outdone by desktop, agents are coming to Breeze Assistant in the HubSpot mobile app. You can bring in first-party agents like the prospecting agent or your own custom agents, assign them work in a normal conversation, answer their follow-up questions, approve actions and review the results from your phone. Push notifications tell you when an agent needs you.

Trisha had the perfect story: a client's top sales rep who is always in the car between facilities and meetings, where the ops lead's answer to "should she use the mobile app?" was please, no, not while she is driving. Hands-free changes that.

Why it matters: Agents can do a lot, but they stall when they need an answer or an approval and you are away from your desk. Now you can keep them moving between meetings. Creating and managing agents still happens on desktop.

For more details: the agents on mobile update.

Breeze Assistant for Lead Scoring

So this is why I was here. Breeze can now create, edit and analyze lead scores. Tell it you want to build a score and it reviews what is already in your portal, including your ideal customer profile and past conversion data, and drafts a score you refine in plain language or from an uploaded file. Before this, Breeze could only answer questions about lead scoring, and Create Score with AI was Marketing Hub Enterprise only, with lifecycle stage and time frame as its only inputs.

I have had less than 24 hours with it, but my advice, with a spoiler alert: go in with an open mind and have a conversation with Breeze about the score you want, instead of just dictating what you have already decided. It uses your HubSpot context, so fill that out first if you have not. Ask where it got things from. Talk about what a good outcome looks like for you.

One tip: tell it who you want more of. Sometimes the people who had good sales conversations with you are easier to isolate than your customers, because the people marked as customers often include billing contacts and other folks who are not your target.

The thing I am most excited to test is why a contact scored. Sales teams want to know what put this lead in front of them right now, and especially the last thing the person did that moved the score. Before this, Breeze could not see lead scoring, so that was hard, if not impossible, to get. Trisha had a client ask her about lead scoring yesterday on an inherited portal, and his first question was the one I hear all the time: didn't it already do this? It did not build it for you. Now it will.

For more details: the Breeze lead scoring update.

Will It Breeze: Build Us a Lead Score

Chris copied the update, gave it to Breeze with no other context, and started a voice conversation. Breeze asked whether to build a new score or refine one (new), whether to prioritize contacts or companies (contacts), which part of our customer profile to focus on (operations teams), and which conversion mattered (becoming a customer). It came back fast with a proposed score capped at 100 points.

Then Chris asked the queen of scoring, which is apparently me, whether to create it. My answer: not yet. Ask what data it used, what the filters were and who the examples were. Trisha was not a fan of how quickly it jumped to an out-of-the-box answer when she expected something customized to the portal, and I had seen the same thing in my own testing.

Pushed, Breeze was honest. The two activities it picked showed up on three converted contacts each, and in its words, "That's not strong evidence and it doesn't prove they cause conversion," and its starting weights were "just my starting assumptions, not something learned from your data." It revised to a fit-only baseline, left out the signals it could not vouch for, created an unpublished draft, and tested it on five anonymous records it picked to cover a match, some title variations and missing data. Its own read: a rules check, not proof that it predicts anything. Then it ended the call on its own.

That is the conversation I was talking about. You can watch the whole run on its Will It Breeze page.

The Rest of Today's Updates

A few more shipped that we did not get to on air: a Data Sync API for AI agents, a sunset notice that AI enrichment of the ticket category property is moving to Topics by January 10, a User mappings page for the Slack integration, filters in the "Add existing" association panel, and owner references that now survive a copy into a sandbox. Everything is in the HubSpot updates feed.

The Sign-Off

Go start talking to Breeze. We will continue this one tomorrow.

Produced by Value-First Media