HubSpot just named a new category
On 12 November 2026, Brian Halligan returned as CEO of HubSpot. In a post on X, he said he and co-founder Dharmesh Shah are refounding the company, and that HubSpot is now “the Customer Harness Company.”
Dharmesh followed with his own post: HubSpot will build what he called the world’s first Customer Relationship Harness, and accelerate its vision of Headless HubSpot. He framed it as the company’s third chapter. Twenty years ago it was inbound marketing. Today it’s helping businesses use AI.
There’s one gap. HubSpot hasn’t published a detailed definition of a customer harness yet, and plenty of agency owners are asking the obvious question: what is a harness?
This post answers it. We’ve spent the past year building the Knock Knock App, a signal-based AI sales layer for agencies, which works the way we think a harness should. Here’s our definition, what goes into one, and why the most important part isn’t the AI at all.
What is a customer harness?
A customer harness acts on buyer signals and puts AI to work on them, so fewer leads go unanswered.
The word comes from AI engineering. A harness is everything wrapped around an AI model that lets it do useful work: the data it can see, the tools it can use, the rules it follows and the memory it keeps. Same model, different harness, very different results.
A customer harness applies that idea to your customers. It’s the system that gives AI agents live context about who your buyers are and what they’re doing, then lets those agents act inside the guardrails your business sets. That might mean answering a question, placing a call, sending a follow-up or booking a meeting.
A CRM remembers your customers. A customer harness works them.
The four parts of a customer harness
Strip away the buzzword and a working customer harness does four things:
- Signals: notice what buyers do. Website visits, pages viewed, return visits, email clicks, form starts, chat messages and calls. Each one can be a sign of intent.
- Identity: know who is behind the signal. An anonymous visit is noise. A return visit from the operations manager who opened your proposal yesterday is an opportunity. A harness connects signals to known contacts wherever the data allows.
- Decisions: work out what to do next. Should this visitor get a chat message, a phone call, an email, or nothing at all? The harness weighs the signal against your rules and goals.
- Actions: do it while it matters. AI agents respond in the moment. They start a conversation, place a call, send a follow-up or book the meeting, and log it back to your CRM.
Miss one of these and the system breaks. Signals without identity are anonymous analytics. Identity without action is a lead list nobody calls. Action without signals is spam.

Why signals come first
Most of the conversation about AI in sales is about the agent: which model, which voice, how human it sounds. That’s the wrong place to start.
An AI agent is only as good as what it knows right now. Give the best model in the world a static contact record and it will write a polite, generic email. Give an average model a live signal, such as this person being on your pricing page for the third time this week after opening yesterday’s quote, and it has a much better idea of what to say and when.
That’s the core idea: a customer harness without signals is just a CRM with a chatbot.
Timing matters too. Buyer interest fades fast. In a classic 2011 Harvard Business Review study, firms that tried to contact web leads within an hour were nearly seven times as likely to qualify them as firms that waited even an hour longer. Often the best moment to reach a prospect is while they’re still on your website, thinking about you. A signal-based harness can act in that window. A workflow that only runs once a day can’t.
Customer harness vs CRM
They’re not the same thing, and you don’t have to rip one out to get the other.
| CRM | Customer harness | |
|---|---|---|
| Job | Stores the record | Works the relationship |
| Moves when | A person updates it | A buyer does something |
| Data | Contacts, deals, notes | Live signals and known contacts |
| Who acts | Your team | AI agents, with your team in the loop |
| Response to a live signal | Often hours or days, depending on the team | Can be within seconds, while the buyer is still there |
The CRM stays the system of record underneath. The harness becomes the layer where people and agents actually work. That may be part of what HubSpot means by “headless”: the database stays, and more of the work moves to AI.
For most businesses the practical answer is both. Keep the CRM you have, and put a harness on top of it.

What this means for agencies
HubSpot’s announcement is a signal in itself. One of the best-known names in CRM is betting its next chapter on an AI layer that works on top of customer data. Agencies can take three things from it.
- Your clients will ask about this. “Should we be using AI agents?” may soon become “Do we need a customer harness?” Have an answer ready.
- It’s a new productised offer. Many small businesses won’t want to build a harness themselves. Agencies that set one up, run it and report on it have a recurring service to sell, not a one-off project.
- You don’t need to wait. HubSpot is still sharing the details of its plans. Your clients’ buyers are visiting their websites today.
Where the Knock Knock App fits
The Knock Knock App is a signal-based AI sales layer, the kind of customer harness HubSpot is now talking about. It runs on top of the CRM your clients already use, including HubSpot and GoHighLevel, so they don’t have to choose. Here’s how it covers the four parts:
- Signals and identity: visitor recognition notices when known contacts come back to a client’s website, depending on the data available, and the Knock Knock App keeps each contact’s session history and scores.
- Decisions: NOX, the central brain inside the Knock Knock App, reads the signals, scores how warm each contact is and works out the next best move for its agents.
- Actions: AI chat, inbound and outbound AI voice calls, and Agent K for outbound prospecting respond when a buyer shows intent.
- White-label: agencies can rebrand the platform as their own and roll it out to their clients.
HubSpot has given the idea a name. Agencies can put it to work now.
Book a demo and see the signals on your own site.
FAQ
A customer harness acts on buyer signals and puts AI to work on them, so fewer leads go unanswered. It combines live signals, recognition of known contacts, decision rules and AI agents that act within the rules your business sets.
HubSpot introduced the idea when Brian Halligan returned as CEO on 12 November 2026, alongside a vision called Headless HubSpot. At the time of writing, HubSpot hasn’t published a detailed definition.
No. A CRM stores customer records. A customer harness acts on live buyer signals using AI agents. Most businesses will run a harness on top of the CRM they already have.
Signal-based selling means reaching out based on what a buyer is doing now, such as visiting your pricing page or opening a proposal, rather than following a fixed sequence or a cold list.
Yes. With a white-label platform like the Knock Knock App, agencies can run a signal-based AI sales layer under their own brand for their clients, on top of HubSpot or GoHighLevel.
References
- CMSWire. Brian Halligan Returns as HubSpot CEO, Replacing Yamini Rangan. November 2026.
- James B. Oldroyd, Kristina McElheran and David Elkington, Harvard Business Review. The Short Life of Online Sales Leads. March 2011.
Anna Hyatt
Co-Founder | Knock Knock App
AI × Human Touch = Happier Clients, More Revenue
LinkedIn Profile



