I have lost count of how many of these I have been to. At least a dozen. The first several as a customer, the rest from the partner side. Long enough to have watched this industry get something right, and then wander off from it in about twenty-four months.
Last week in Boston I barely went to sessions. I have been catching up on the recordings since. What I did instead was spend the week talking to HubSpot leadership and to customers, which means what follows is not a recap. If you want to know what shipped and what is still sitting in public beta, we wrote that up separately. This is what I actually think after a week of conversations with the people shaping where this goes.
The last two years took us off course
I want to be careful here, because this is not a story about everyone having been wrong all along. For a long stretch we were pointed at exactly the right thing. Get off the spreadsheets. Get the data in one place. Agree what a lead is. Know whether you are on track to the number. That work was correct, it was hard, and a lot of you did it.
Then, somewhere in 2024 and 2025, the business world discovered AI. We used it as a hack. Some of us got further ahead, some further behind, plenty of people are still doing generative work in a chat window and honestly that is fine.
But every conversation became the same conversation. How do we use more AI? How do we use more AI?
In about two years we got obsessed with the how. The tactics. The models. Whatever was going to get us there faster, and the thousands of use cases that came with it. What slipped in that window was the question we used to be good at: am I using this to move my business forward, and do I have the pieces in place for that to work?
My colleague, Krystina Gillenwater, gave me an analogy for this recently that I have not been able to stop thinking about.
Mario is an average guy who knows plumbing. He is trying to rescue the princess. That is the outcome. He goes along through the level, and it is not until he transforms into a bigger, more capable version of himself that he can actually clear the obstacles and get there. The mushroom is how he transforms. The mushroom is not the point. Leaping the gaps and moving between the floating platforms is the point.
AI is the mushroom. For the last couple of years we have been so focused on the mushroom that we stopped talking about the transformation.
The thing Yamini said that should have made the room uncomfortable
In Yamini's Spotlight keynote, she said something that should make us all pause.
Here is the line: bad context mixed with AI is worse than no AI at all.
Not neutral. Worse.
That should sound familiar to anybody who was around for why we all moved onto a CRM in the first place. Our data was everywhere. We were running businesses out of spreadsheets. Joe had a spreadsheet with John Doe on it and Bob had a spreadsheet with John Doe on it, and they were both calling him, and nobody could tell whether we were following up the way we said we would or whether we had now contacted the same person four times this month.
Disparate data and dirty data are the reason the CRM category exists. We knew that. We built entire operating models around it, and they worked.
Then AI turned up in everything, and we quietly stopped focusing on it. Two years, maybe a little more. We are vibe-coding CRMs. We are trying to Claude our way out of problems. More than anything, we are chasing use cases. Hey, how could we use AI better? Well, let us see if we can build this wild lifecycle nurturing sequence.
Nobody decided to abandon the fundamentals. They just stopped being the interesting part of the conversation, and eighteen months later the portal shows it.
If you can do it natively, go do it natively
I have been saying this for two years about HubSpot and I will keep saying it.
That is not an anti-AI position. I am in Claude for most of my working day. I have a Grok bot running. I am doing this stuff constantly, and I will also tell you that I have spent evenings arguing with a model about code it got wrong.
But if the thing already exists natively in the platform, use the native thing. Native features are built around outcomes rather than around one clever use case. They are tried and tested. And I do not have to maintain them, because a product team does.
The shift I would like all of us to make is from chasing use cases to chasing outcomes. HubSpot leadership said this at UNBOUND, and I am just here to say it again.
The outcome has not actually changed. The reason we all wanted a CRM was to know whether we were on track to forecast, amongst many other outcomes. That is still the reason. Not whether we can wiggle our way into a forecasting tool or find the coolest available use case. Can I see whether I am going to hit the number or not is the outcome.
It is very easy to get lost in "did we build the thing." The better question is whether the thing you built made the company better.
So: context is everything, and data is a large part of context. We have to go back to getting that part right before we spend another quarter on how to use AI better.
Implementing your current process in new software is not transformation
I want to be blunt about this one because I see it constantly.
If you bring in a partner or an internal team, and they run discovery, and they document exactly how you do things today, and then they go and build EXACTLY that in the software — you have not transformed anything. That is not transformation, and I would call bullshit on it in a second.
Transformation is when you look at every process and put a before and after next to it. Here is how we do this today. Here is how we would do it in the new world. You will not change everything, because change is hard and people can only absorb so much of it at once. But you will come out the other side knowing which changes make sense right now, which ones are too much, and what the plan is for the version of the business you are trying to become.
The analogy I keep coming back to is moving house.
You hire the movers, you pack every box exactly the way the cupboard was, and you unpack into the new place in the same order. New address, same junk drawer, same two boxes you will not open until the next move. Nothing about how you live changed. You paid to relocate it.
The move was the one moment you had permission to ask what you actually use. That is the part people skip, because packing is urgent and deciding is hard.
That is the difference between implementing HubSpot and transforming on HubSpot. One of them relocates what you already do. The other asks what is worth keeping. In CRM terms, the boxes not to repack are the lifecycle stages nobody agrees on, the properties nobody owns, and the handoff that only works because one person remembers to do it.
Where I have seen this actually pay
A plug, and I will be upfront that it is one.
We are heavy users of Revenue Hub at Growth. We run our own business on it, and we implement it for clients, because it closes a loop the industry has left open for fifteen years. We all talked about closed-loop marketing in the early inbound days. That ability to see how revenue is generated from marketing efforts. Almost nobody talks about closed-loop revenue: I close a deal, and later I can see whether we are up for renewal, or whether this customer has missed their invoice over and over and we probably should not be selling them more.
Sales and marketing driving demand got all the attention in our world. The operations and finance side of a business is where the money actually lands and were disparate.
I was on stage at UNBOUND with Sarika Garg talking about exactly this. She showed how Revenue Hub powers attribution, CPQ and the finance end of it. How to build a quote with Breeze in about a minute. How to point a revenue agent at invoice follow-up so a person does not have to do it, and frankly so it gets done better than a person would.
She brought me on to talk about the real-life version. One client came to us with invoices living as PDFs, scattered, with no unique identifier on any of them. Getting to a true revenue picture meant crawling every invoice, reconciling it, and bringing it into HubSpot under one tracking scheme with real identifiers. In the middle of that exercise we found invoicing that had never been collected at all. Not disputed. Not written off. Nobody had ever chased it, because no system existed in which it was visible.
That is the reinvention part. The technology was the occasion for it, not the thing that did it.
What I would do between now and 2027
Go back to basics. Get the data right. Get the context right. Know who your ICP actually is. Know what your handoff process actually is. Get your team aligned on all of it, in writing.
Then layer AI on top of a foundation that holds, enable your people to use it properly, and manage the change like change management is a real discipline, because it is.
We need to go back to doing what we were doing five years ago, and then do the cool stuff on top of it. Not because the last two years were wasted. The tools got real in that window and I would not give them back. But the foundation stopped getting attention while we were busy, and that is a two-year gap to close, not a decade of bad habits.
If you were in Boston, I want to know what you took away, especially if it is different from mine. And if you have already pointed AI at a portal you knew was not ready, I would love to hear how that went. Those stories are the useful ones and they are the ones nobody posts.
