Approval at the Start and the End, and a Promise You Can't Keep

Another Orange Morning Recap: Wednesday, November 4, 2026

Zach Hussion joined Chris Carolan and me for a Wednesday with two HubSpot User Groups on the calendar and four updates on the board that all came back to one question: how much do you trust the tool, and what have you done to earn that trust? Bulk actions, planning before building, and a social content agent that promises a lot all got the same treatment.

Here is what we got into: Zach's monthly updates HUG, my Women of HubSpot HUG and a LinkedIn event I fixed on air, then the redesigned product updates page, rejecting suggested duplicates in bulk, Breeze Research and Chris's AI TED talk on planning, and creating social posts with Content Agent in Breeze Assistant.

Before the Updates: Two HUGs and a Calendar Fix

Zach used to run his updates HUG quarterly. Q1 and Q2 happened, then the business and UNBOUND got busy, and a whole quarter of updates turned out to be too much for one session. So he is moving to monthly: the first week of each month he looks back at the previous month and picks the ones he thinks are the coolest. Today was the October edition, and his team cuts each session into short clips afterward.

Chris called it one of the easiest ways to start creating content about what you know, and Zach's version of why is the one I would steal: "you just take your experiences plus HubSpot and make a video and no one else will make that video ever."

Then I had a confession. Looking at the events page this morning, "I set up my LinkedIn event wrong and didn't notice until I looked at this page today." I fixed it while we were talking. At noon Central is my Women of HubSpot HUG. "We're specifically talking about the difference between mentorship and sponsorship and how to get both of those things from others, but also how to give both of those things to others." I planned that whole quarter before UNBOUND, but in last month's session one of the women mentioned she was struggling to get leadership buy-in for herself, and the topic picked itself.

Chris shared a post he made this morning, an apology to his agent team. He had been frustrated with the state of our data model documentation, and when he looked closer, he was the one moving the goal posts. His point was bigger than AI: "it's not a new problem related to AI. It's do people have the authority they need to do what you ask them to do, the expertise they need, the scope, the access, all of those things." Hand someone a title and then gate everything they need, and you have not set them up to succeed. Blame all around, potentially, or opportunity all around. I'll take opportunity.

Four Updates, One Thread: How Much Do You Trust It?

Redesigned Product Updates Page

This is the most important category we cover on this show, and it matters to us more than almost anyone. The product updates page has been redesigned to highlight what's most relevant to your team, and as of October 21 you can catch up on all new product updates since your last visit in the Discover tab. I had mixed feelings. The old page had a New to You section that was very good, unless you accidentally opened it before the show and lost your place in the list. We ended up taking screenshots to keep track.

The new view helped with something I do close to the first of each month. Every so often an older update shows up in the list late, so yesterday I pulled every product update from October to catch any we missed. Those all live on the Another Orange Morning site under HubSpot updates, with our take and the clip when we covered one. Zach asked what a page looks like when we didn't cover it, and I'll be straight about it: "It's kind of weak sauce right now." Finding an example took a minute, and Zach's read was that "it's getting harder and harder to find things that we don't cover."

Chris pushed on the idea that there are too many update posts out there. In his words, "the misunderstanding is that there's too much content about product updates." Copy and paste posts, maybe. A video clip of a real person giving their take on an update is how the education gap closes, especially once AI is pulling from it to answer questions about someone's specific business. Zach added that every audience is different, the work keeps you current and it feeds AEO and your own second brain. But "it's just not a field of dreams… you have to keep consistently doing it and be purposeful."

Chris shared a sign it's working. George B. Thomas, who did the previous version of this show with us, now makes a short podcast for each update, and Chris said one episode "cited our show as another HubSpot news show" and referenced us in the notes. His takeaway: "when you make it easy for AI to share your stuff now, when AI is creating content based on that stuff, it's going to include you in that content."

Why it matters: if you follow updates closely, the Discover tab gives you a way to catch up since your last visit. If you write about updates, your own experience on top of the release note is the part nobody else can make.

For more details: Redesigned product updates page.

Reject 100+ Suggested Duplicates in Bulk

Chris doesn't have Data Hub, so we pulled this one up on my screen. You can now reject more than 50 suggested duplicate pairs at a time from the duplicate management tool. By default HubSpot surfaces contacts with a similarity score of 30 percent or more, and as you work through them you may decide that pairs below a certain threshold are not duplicates at all. Filter to that group and reject it in one action, so you can focus on the pairs you care about.

My first question was about the name: "So does the plus mean all?" My screen showed all 9,278 pairs selected. Zach had both reactions at once: "It's scary to be like, oh, I got to do this one at a time. I'm never going to do this." And then: "it's also scary to be like, oh, let's just do all nine thousand of these." It helped to remember it's rejecting, not merging. Even so, "I personally have seen more situations where I want to accept in bulk."

Zach connected it to the finer duplicate filters that came out in October. Dial those in so the suggestions are better, and "I have a lot more confidence in doing a mass reject or accepting things in bulk."

Why it matters: a bulk action is only as good as the filter in front of it. Tune what counts as a duplicate first, then clear the noise in one pass instead of one pair at a time.

For more details: Reject 100+ suggested duplicates in bulk.

Breeze Research: From Question to Action Plan

Breeze Research is a deeper investigation mode in Breeze Assistant. Instead of answering in one shot, Breeze builds a plan, works through the relevant data step by step and returns a structured report with findings and recommended next actions. It is built for questions that need data from several sources and some real analysis before you can decide what to do.

Chris has wanted this for a while. At last week's Agent Hub HUG he asked whether "all of these Breeze conversations that are to build agents and workflows just start in plan mode," and the answer was yes. It reminded me of something Sarah Lane-Hawn recommended when we talked about when AI helps and when it hurts: tell it to show you first, before it does anything. I need that, because "I'm moving a million miles an hour, so I'm kind of letting AI just do things."

Zach asked Chris whether that should live in your standing instructions, so that in general "we should be talking about these things and reviewing a plan before we just jump into doing stuff." Chris's answer: "This is a part of the AI learning curve for sure. And it works until it doesn't." He builds plans with a five-section command based on Trust Insights' 5P framework and reviews each section himself. But he has learned where the approval belongs: "these agents already have the expertise. The approval needs to be at the start and at the end."

What set him off was a prompt full of do-nots: don't send invoices, don't publish, don't do this or that. The team spent its effort avoiding those things and reported back on every one. People work the same way: "if you throw instructions like that at me all the time, I'm not going to have any confidence." Many of the gates in our own system were built when the models were weaker, and he is now taking them down in an orderly way. His close: "the focus is on what not to do, instead of what to do. And that's my AI TED talk for today." Zach's review: "That's a good clip right there."

Why it matters: a plan you can read before Breeze acts is where trust starts. Put your review at the beginning and the end, and write instructions about what to do rather than a list of what to avoid.

For more details: Breeze Research: from question to action plan.

Create Social Posts Using Content Agent in Breeze Assistant

Content Agent launched at UNBOUND, and its social post capabilities have now moved into Breeze Assistant in beta. New with the move: Facebook posts alongside LinkedIn, bulk creation, planning a series of posts from a piece of starter content or a recommendation, and a conversational way to edit posts before you save them as drafts. It creates posts in your brand style and generates images from your company's own templates.

Zach brought a timely frame: Dharmesh Shah's recent post on AI foam, the content smart people send that is clearly AI generated and missing their usual point of view. "I call it AI foam because it lacks shape and substance," Zach read. Avoiding it takes the work of giving the tool real context, not just clicking the button.

Then Chris got to the line in the update that says HubSpot ensures every post avoids common and uncommon AI tropes and follows industry best practices. His response: "some of those industry best practices become tropes very, very quickly," so "don't ensure, come on, that's a promise you can't keep." The bigger worry is how many people will see that sentence and ship a whole month at once. I read the next line, that it nails your voice on day one with no setup required, and Chris and I answered it together, talking over each other: "Sounds like a no brainer. Open and shut case."

Credits came up next. Chris found a credit cost listed in the catalog and read it on air; we did not check it against a live portal, so look it up in yours. Zach's working rule: "let's just assume that everything is going to use a credit at some point." Chris: "great power, great responsibility, folks. Before you hit that bulk button, double check the credits." Zach wondered "how many phone calls or chats HubSpot gets on a daily just being like, I don't know what happened, but we just used fifty thousand credits."

Chris tied it back to his TED talk with one question: "Are you even clear on your own voice?" If the first post doesn't sound like you, the fix is the context the agent can see, and Breeze can help you build that.

Why it matters: the agent can only sound like you if you have written down what you sound like. Try one post before a bulk run, read it as a stranger would, and check what it costs.

For more details: Create social posts using Content Agent in Breeze Assistant.

The Rest of the Updates

We wrapped with a lot left on the board. Also new: the Run records on a schedule trigger in the Agent Hub workflow builder, a Sandbox help panel for in-app guidance, Voice on Breeze Assistant on mobile for calls, notes and follow-ups, updates to the HubSpot connector for Copilot, an Azure Synapse connection to Data Studio, re-evaluation of price book auto-assignment rules and a change to how the customer service unsubscribe link behaves. A few more carried over from Tuesday's list: Premium Engines in the AEO tool (Claude, Google AI Mode and AI Overviews), email templates for manual email tasks in workflows, and property comparisons in segments and workflows. Everything else that shipped is in the HubSpot updates feed.

The Sign-Off

You probably already own the value you are looking for in HubSpot. You just need to wake up to it. We will see you tomorrow morning.