Field Notes / AEO & AI
Deep Dive

What is a customer harness? HubSpot's refounding, explained

Chris Tveter
Chris Tveter
November 16, 2026
30 min read
The Short Answer

A customer harness brings together everything about your customers in one place: your CRM data, emails and meeting transcripts. Whichever AI you use, Claude, ChatGPT or HubSpot's own Breeze, plugs into it, works from that information, takes action on it and stays within the rules you set. HubSpot is now positioning itself as that harness. The brains will keep changing. The harness is what stays.

HubSpot's refounding hinges on one word almost nobody understands: harness. Brian Halligan is back as CEO, and on his first day he described HubSpot as "the Customer Harness Company." Most HubSpot users I talk to have no idea what that means for them. This is the plain-English version, written for the average HubSpot user, with how I already work this way every day.

Listen to this post

Narrated by Chris Tveter, CEO of AIRops Agency · 6 min

What did HubSpot announce in the refounding?

On November 12, 2026, Brian Halligan returned as CEO, and he and Dharmesh Shah declared a refounding of HubSpot built around a new idea: the customer harness.

Brian put it in one line on day one: HubSpot is now the Customer Harness Company. In the same announcement, Dharmesh wrote that HubSpot will be creating the world's first Customer Relationship Harness and accelerating its vision of Headless HubSpot.

In other words, your CRM is becoming a CRH. Customer Relationship Management is turning into a Customer Relationship Harness.

"We will be creating the world's first Customer Relationship Harness and accelerating our vision of Headless HubSpot."
Dharmesh Shah, HubSpot co-founder and CTO November 2026

What is a customer harness?

A customer harness brings together everything about your customers in one place, so that whichever AI you use can work from it, take action on it, and stay within the rules you set.

That includes your CRM data, your emails and your meeting transcripts. The AI model, the "brain," can change. Claude, ChatGPT, HubSpot's own Breeze, or whatever comes next. The harness is what stays.

The simplest way I can put it: the brains will keep changing. The harness is what stays: your customers, your history, your rules.

Why does an AI need a harness at all?

Because on its own, your AI is smart but disconnected: it doesn't know your customers, it can't act in your CRM, and it has no rules for what it may touch.

Most people use AI this way today. To draft one follow-up, they copy the deal details out of HubSpot, paste in the email thread, paste in a meeting transcript, type their own notes, and ask for a draft. It works, but you gathered everything yourself, for one deal, and then you paste the result back into HubSpot.

A harness removes that legwork. The AI already has the context, already has the tools, and already knows its limits.

How is a harness different from Loop Marketing?

They answer different questions: Loop is about how you market, and a harness is about how you work.

HubSpot introduced Loop Marketing a little over a year ago as its playbook for growth in the AI era: Express, Tailor, Amplify and Evolve. That covers what you say, who you say it to, and how you keep improving. The harness covers where your HubSpot work actually happens, and who (or what) does it. You can run Loop through a harness. They are not competing ideas.

What are the three parts of a harness?

A harness comes down to what the AI knows, what it can do, and what it's allowed to do.

What it knows

Your contacts, companies, deals, emails, calls, meetings and transcripts. This is the customer history the AI works from, instead of whatever you remembered to paste in.

What it can do

Update records, run workflows, put HubSpot's agents to work and build quotes. This is the difference between an AI that writes a draft and an AI that does the work in your CRM.

What it's allowed to do

Who can access what, and where a person signs off before anything reaches a customer. In practice, I recommend starting read-only, expanding what the AI can change as you build trust, and keeping a human approval step on anything customer-facing.

How do Claude, ChatGPT and Breeze connect to the harness?

Claude and ChatGPT plug into HubSpot through MCP, and Breeze works from inside HubSpot.

MCP, the Model Context Protocol, is the cable. Think of it as USB for AI: a standard plug that lets an outside AI tool connect to HubSpot. In his Day 1 notes on the refounding, Dharmesh described Headless HubSpot as capabilities that will be consumable by AI agents and AI harnesses like Claude and ChatGPT via MCP.

Breeze, HubSpot's own AI, already lives inside the platform. Dharmesh has said HubSpot is doubling down on both Breeze and the MCP connectors, so this is not one or the other.

A note for technical teams: MCP vs. the CLI. MCP is the connection built for AI tools like the Claude app and ChatGPT, and it is how most HubSpot users will work with the harness. The HubSpot CLI is the command-line tool developers use, and coding tools can use it too. If you are not building on HubSpot, MCP is the one that matters.

What does "headless HubSpot" mean?

Headless HubSpot means you won't always need to open HubSpot to use it.

For most of its history, the only way to work in HubSpot was to log in and click through its screens. Headless means the same data and actions are available wherever you already work, including inside Claude or ChatGPT. The interface is not going away. It is just no longer the only way in. As Dharmesh put it, sometimes the interface just needs to fade into the background.

Why is the harness a smart bet for HubSpot?

Because HubSpot doesn't have to win the AI model race; it has to be the place every AI works from.

Models will keep leapfrogging each other. Your customer records, your history with each account and the rules for how your team operates do not change every time a new model ships. Whichever brain wins, it still needs a harness.

What does working through a harness look like day to day?

For me, it means much of my HubSpot work now starts with a sentence in Claude instead of a login.

Earlier this year, I sat on panels at HubSpot's Sales Hub Summit and Commerce Hub Summit (now Revenue Hub). The product teams asked what I wanted to be able to do. My answer was to let me do my HubSpot work without opening HubSpot, since I was already doing much of it in Claude. Every one of those asks is now available, and this is how I work today:

  1. A daily scheduled task. Every morning, a scheduled task in Claude runs against my HubSpot portal and tells me which deals need follow-up and which emails I missed. I don't log in to find out.

  2. Quotes in Claude. I build quotes directly in Claude, which is exactly what I asked the product team for.

  3. Less clicking. Most of what used to mean clicking through screens now starts with a sentence I type to Claude.

One honest limitation: the harness is still being built. Dharmesh's own wording is "will be creating" and "stay tuned." Partners testing the connector have reported that some context, such as Context Home, is readable by Breeze but not yet by outside AI tools through MCP. Check what your connector can actually see before you rely on it.

What this supersedes.

The old approach

The CRM is a place you go to do the work. You log in, search, dig through activity and type the follow-up yourself. When you use AI, you copy and paste your customer context into it one deal at a time.

The current reality

HubSpot becomes the harness: what your AI knows, what it can do and what it's allowed to do. Your AI of choice plugs in through MCP, works from your real customer history, and brings the work to you for approval.

How should a HubSpot user get started?

  1. Connect HubSpot to the AI you already use. Add the HubSpot connector in Claude or ChatGPT.

  2. Start read-only. Ask questions before you let it change anything: which deals went quiet, which emails need a reply, what was promised in recent meetings.

  3. Check what it can see. The connector works through your own HubSpot access, so confirm what it can read and edit.

  4. Automate one recurring job. Turn the question you ask every morning into a scheduled task.

  5. Keep a person in the loop. Approve anything that reaches a customer, then expand what the AI can do as you build trust.

The CRM stored your customer relationships. The CRH lets AI act on them, safely. I've been working that way all year, and now everyone gets to.

Is your HubSpot portal ready to be a harness?

An AI only works as well as the customer data, context and permissions it plugs into. The AI Readiness Audit shows what your AI can see in your portal, what it can act on, and what to fix first.

Chris Tveter, CEO of AIRops Agency
Chris Tveter
Chris Tveter is the CEO of AIRops Agency, an AI Revenue Operations firm and HubSpot Diamond Partner. He has worked across 285+ HubSpot portals and does much of his own HubSpot work through Claude.