From the May 12th, 2026, edition of How Teams Work
Every year, HubSpot drops its Spring Spotlight and the internet lights up with recaps, hot takes, and feature breakdowns. If you've been following along, you've probably seen more than a few posts covering what was announced.
We're not going to do that.
Instead, we've been paying close attention to how Dharmesh Shah, HubSpot's CTO and co-founder, has been framing the bigger shift happening underneath all of these announcements.
If you're not already following his newsletter at simple.ai, it's worth subscribing. He's one of the clearest voices on where this is all heading, because what he's describing isn't really about features. It's about a fundamental change in how CRM platforms, and the teams that run on them, are going to operate over the next 6 to 12 months.
So here's our translation on six things that are actually changing and what that means for the rest of us.
Today: Your HubSpot instance was built for humans. The navigation, the property descriptions, the workflow logic — all of it assumes a person is clicking around, reading labels, making decisions.
What's next: That's changing. Dharmesh has been writing about what he calls AUX, an agent user experience, where agents are treated as first-class users with their own workflows and their own context, not just endpoints bolted onto a human interface.
You can already see this inside HubSpot.
Smart Deal Progression is a live example of what's happening under the hood. On the surface, it looks like a time-saver. AI suggests post-meeting CRM updates, emails, and next-step recommendations so your reps aren't manually filling things in after every call. But what's happening behind the scenes is more interesting than the feature itself.
For this to work, each deal property needs a second layer of context. Not the description your reps read, but a prompt written specifically for the AI. What is this field for? What should it look for in a transcript, and how should it decide what to populate? It's a mini instruction set, written not for a human but for the agent doing the updating.
We've been seeing this pattern show up across tools for the last year and a half. Smart Deal Progression is one of the clearest examples of where it's already arrived inside HubSpot.
The question for your team isn't whether this is coming. It's whether your CRM is set up in a way that an agent could actually do something useful with it or whether it's still built purely for the human clicking around on a Tuesday afternoon.
2. The people who automate their own work are the ones who get promoted
Today: Think about what happens around a sales call at most companies.
What's next: Dharmesh's take on this is worth sitting with: the people who automate repetitive parts of their role don't get replaced. They get handed bigger problems. The skill that matters now isn't your job title or years of experience; it's whether you're curious enough to look at a broken or boring process and determined enough to fix it.
At Process Pro, we automated both ends of the sales call process.
Before every intro call, we pull in the prospect's background, their form submission responses, and any prior conversation context so that we're walking in with the full picture rather than generic talking points.
After the call, we aggregate what was discussed and generate a draft follow-up email in a predefined format, built around what that particular customer shared.
The result isn't just time saved. Our team spends the time they used to spend on documentation actually thinking about how to deliver the right solution for that customer.
The output is better. The client experience is better. And the people doing the work are focused on the part of their job that requires their expertise and judgment, not the part that was always just overhead.
That's the pattern. Not "AI replaces the person." It's "AI handles the task so the person can do the job they were actually hired to do."
Today: If you have a CRM admin or a RevOps person on your team, their day probably looks something like this: building workflows, writing property documentation, cleaning data, managing integrations, and keeping everything running.
It's specialized work that requires someone who knows the system deeply.
What's next: Agents are going to take over the intelligence-heavy repetitive work, i.e., building automations, detecting data issues, cleaning records, and updating fields.
What's left for the human is judgment. Deciding what should be built. Setting up the guardrails so agents can run reliably. Reviewing the output and knowing when something's off.
The job isn't going away. The job description is changing. And the people who thrive in that shift are the ones who stop thinking of themselves as the ones doing the CRM work and start thinking of themselves as the ones directing it.
Today: For the last couple of years, HubSpot's AI capabilities have been largely contained within the platform, limited to certain record types, read-only in many cases, and built around what HubSpot's own product team decided to prioritize.
What's next: That's opening up. HubSpot is expanding its MCP and APIs so that outside agents, not just the ones HubSpot builds, can work inside the platform. Workflow creation just went to beta. More features are becoming accessible to agents by the week.
Dharmesh has even built a separate company, agent.ai, that entirely focuses on agents working on top of platforms like HubSpot.
That's not a coincidence.
The platforms that win aren't going to be the ones that try to build every agent in-house. They're going to be the ones that give their users, and outside builders, the keys to build what the product team never could have anticipated. More people building means more use cases, more value, and more reasons to stay on the platform.
For your team, this means the ceiling on what you can automate inside HubSpot is getting higher, and it’s happening faster than most people realize.
Today: You pay for HubSpot by the seat. X number of users, X cost per month. It's predictable, it's familiar, and it's the model most software has run on for the last two decades.
What's next: That math stops working when agents are doing work that used to require human seats. If an agent is cleaning your data, maintaining your property documentation, and building automations, work that previously justified headcount, paying per person starts to feel like the wrong unit entirely. We can see examples that HubSpot is evaluating a shift toward consumption and usage-based pricing with Prospecting Agent and Customer Agent. You pay for what you actually use, not just who has a login.
For teams getting real value out of the platform, that's a good thing. For teams that have been adding seats without asking hard questions about what those seats are actually producing, it's a different conversation.
The question stops being "how many seats do we need?" It becomes "what are we actually using this for, and is it delivering?" The teams that start asking that now will be in a much better position when the pricing structure shifts under them.
Today: Most teams using HubSpot haven't had to think about building anything. You use what HubSpot ships. You connect a few integrations. The platform does what it does, and you work within it.
What's next: That's changing. As HubSpot opens up its MCP and APIs, outside agents can now work inside your instance, ones that have nothing to do with what HubSpot's own product team built or prioritized. Which means the question isn't just "what does HubSpot offer?" anymore. It's "what do we want to build on top of it?"
While this sounds like a developer problem, it's actually a strategy problem.
The teams that get the most out of the platform aren't going to be the ones waiting for HubSpot to ship the right feature for their specific process. They're going to be the ones who look at where their workflows were breaking down and build something to fix it. The ceiling on what's possible inside HubSpot is rising fast. What you do with that is still up to you.
We've been doing the same thing since day one: coming into a business, learning how it actually works, and building custom solutions that fit the way that business operates.
We don't do cookie-cutter. We don't put clients in a box.
What's changing is the toolkit. The discover, design, and build process we run with every client now includes things it didn't two years ago: AI-assisted workflows, agentic actions, agents, and humans working in tandem.
The outcomes we're after for our clients are exactly the same. We just have more and better ways to get there.
The same is true for your team. The goal isn't to chase every new feature or rebuild everything from scratch. It's to get clear on your processes first, then figure out where agents can take work off your team's plate, so the people you have can focus on the work that actually requires them.
That's always been the job. It just looks a little different now.
Have a question about how any of this applies to your team? Throw some time on my calendar.
Want to get the "How Teams Work" newsletter sent to your inbox? Subscribe Now.
