Is Your HubSpot Portal Ready for Agents?

From the October 7th, 2026, edition of How Teams Work

During his keynote at UNBOUND, Dharmesh Shah, HubSpot's co-founder and CTO, gave the room some homework:

• Pick one task you repeat every week.
• Write the job description you would hand a new hire.
• Then give that same description to an AI agent.

That is not how most of us talk about software. It’s how we talk about people.

But sooner than you think, it IS going to be how we talk about software, and in this case, HubSpot.

If the future is agentic. Are any of us really ready? More specifically: if it comes tomorrow, is your HubSpot portal ready?

HubSpot has been heading in this direction since INBOUND 2024, when it introduced its first generation of agents. Dharmesh first started teasing what this agentic future would look like when unveiling Agent.ai, a professional network for AI agents.

Agents in the HubSpot ecosystem have come a long way since then. This year was the first conference where it looked like that future was taking shape.

The conversation around HubSpot has changed. It used to be about which tools and features to use. Now it’s about what outcomes you want to see, and what you’re going to build inside your system to make those outcomes possible.

Here’s that question again: if the agentic future comes tomorrow, is your HubSpot portal ready?

What the agentic future looks like

When people think of the future, it’s common to think of flying cars and personal robot assistants. While we may not be there yet technologically, in business terms, the future we anticipate will be just as revolutionary.

First, let’s define what we mean by ‘agentic’.

An agent is a piece of AI-powered software that is designed to accomplish specific multi-step tasks autonomously or semi-autonomously. When we say the future will be agentic, we are talking about a shift from teams made up entirely of humans to hybrid teams made up of humans and agents working together.

While the capabilities of this agentic future are endless, here is a sample of what this will unlock for teams:

• Automate recurring tasks and run them on a set schedule, so the work gets done without someone remembering to do it.
• Build custom applications and automations without needing a developer. Anyone on the team can do it.
• Analyze large amounts of data, like deals and call transcripts, and surface what matters for business decisions.
• Spot patterns that used to go unnoticed, like the same objection across sales calls, and flag them to the right person.
• Pull from several tools at once and bring back one answer, then make the update once someone approves it.
• Give every individual contributor a digital team of agents to delegate to.

So what does that mean for HubSpot?

It means a shift in how people use their systems. Today, HubSpot is where your team goes to work. In the agentic future, it may not be. Agents will do more of the clicking, updating, and building, so people spend less time inside the screens. We expect much of that building, like properties and workflows, to shrink sooner than five years out.

HubSpot will become a headless CRM instead. It holds the context for your entire tech stack, including your database, your deals, and your call transcripts. Agents read from it and write back to it, inside HubSpot and in the tools around it.

That puts your business context at the core of your tech stack. Models will keep changing. Your context is what carries over to the next one.

I don't see human judgment going away, though. Agents will surface the decision, and a person still makes the call. What changes is how much time that frees up. Less time digging for the insight, more time deciding what to do with it and testing it.

Where things stand today

When HubSpot's first agents launched, they were bare bones. They were light on features and limited in what they could actually do.

That has changed.

The ones that have been around the longest, the prospecting, customer, and data agents, have had time to mature and have become genuinely useful tools.

During her UNBOUND keynote, HubSpot's CEO Yamini Rangan shared results from a recent survey of 6,000 companies. Out of those surveyed, 90% reported using AI, but only 6% said they had seen transformational impact for their business.

These numbers match what we see with our customers. Most of them are using AI, but few are using more advanced features like agents, and even fewer have seen positive ROI from these investments.

Most companies we speak with have licenses for tools like Claude, ChatGPT, or Gemini. But ask five people within the organization how they use them, and you get five answers. Each person's work stays in their own silo.

Here is what the gap looks like up close. One company we work with put a customer agent on its product pages. Routing buyers to sales or to checkout worked on day one. Easy. Answering questions about the products did not. Once they turned it on, the agent didn't know the products or how to answer questions about them.

The team had to build out the product details and the questions buyers actually ask. Now every question the agent can't answer gets flagged as a gap to fill, so it improves over time. Once they saw it work, they wanted it everywhere.

The lesson isn't about that one agent. Being ready for the agentic future comes down to process and people as much as data. When we are assessing whether a company's HubSpot has the proper foundations to run AI effectively, we look at these three pillars:

1. Data: Do you have the right data structure in place and a database that is clean and consistent?
2. Process: Do you have processes in place that are consistent and align with how your business actually operates?
3. People: Do you have the proper training and governance in place for your team?

During the keynote, Yamini framed this as a list of three questions:

1. What outcomes do you want to drive?
2. Do you have the right context foundation?
3. Do you have the right people?

Different wording. Same idea.

Four questions to ask about your portal

When a team asks if their portal is ready, we start by asking them these four questions:

• Do you know how your system is built? Start with what exists in your portal today. Which properties, workflows, and reports have you built, and which ones does your team actually use? Without that, it's hard to give an agent the right context about your business.
• Does it match how the business actually runs? This goes back to your processes. Do the processes and automations you've built match how the business really operates? Do your pipelines line up with how reps actually sell? Are you collecting the right information at the right times? Are reps skipping steps, and are the right guardrails in place? An agent works from what the system says, so the two need to match.
• Does everyone use the same definitions? What counts as a qualified lead? When does a deal move to the next stage? If sales types company size as a number and marketing picks a range from a dropdown, an agent that groups or scores records will give you muddy answers.
• Do your integrations bring in what your processes need? Calls and meetings that aren't logged look like no activity at all. Billing and subscription data sits in another system. Call transcripts never reach the record. An agent can only work with what it can see.

Here is what it looks like when the answers are no. We built a white space agent for one of our clients to highlight cross-sell opportunities in their existing customer base. It looks at what each customer already has and suggests what else they could buy.

This process will only work properly, though, if the agent has access to the right data, which in this case, it did not. Many of those customers had no line items, or no information about products in use. This data was never included in a previous migration. The agent had little to work from.

In some cases, it pulled product data from an old structure. That gave us false positives. The team was pointed at customers for cross-sell opportunities that weren't real. It also gave us false negatives. Good opportunities got missed.

The agent wasn't broken. It read what was in the records, and the records were out of date. That problem was there long before the agent was.

If the data is wrong in your CRM today, it will be wrong when an agent uses it. The difference is that the agent acts on it.

Where to start

When you're trying to level up AI-driven processes within your org, it can be intimidating to know where to start.

Our recommendation: Don't try to automate everything right out of the gate. Start with one goal. Then pick one or two processes that support it.

Take a look at your week and pick something repeatable that you do all the time. Then think about how AI can help you automate it. Here are a few good places to start:

• Sales call prep and takeaways
• Sales prospecting
• Pipeline reporting and analysis
• Lead management
• Marketing campaigns and content

Any of them works. The point is that if it's something you do repeatedly, it's a prime candidate for automation.

It doesn't even need to be a huge, complex process. Select a single use case and try to integrate AI into how you work today. For example, a data agent step inside a workflow can replace custom code.

Keep a person reviewing the output at first. Feed the corrections back into the agent to keep improving the process.

Are you ready?

Go back to Dharmesh's homework. Picture the job description for an agent working in your CRM. Could it do that job with the data you have today?

That is the work underneath all of this. Get clear on your processes. Get your data structure right. Then let agents take the repetitive work, so your people can spend their time on the decisions only they can make.

So here is that question again. If the agentic future was tomorrow, is your HubSpot portal ready?

Thinking about how you can use agents in your HubSpot portal? I’d love to take a look at what you’re building.

Want to get the "How Teams Work" newsletter sent to your inbox? Subscribe Now.