Your buyers changed before your CRM did. A B2B prospect in 2026 starts in Claude or ChatGPT, asks it to compare vendors, and shows up to the first sales call with a shortlist already formed.
HubSpot built its Fall 2026 Spotlight for that buyer and released it on September 16 at UNBOUND in Boston, the conference it ran as INBOUND for fifteen years. The Spotlight is 134 updates. The headline pieces: a rebuilt Breeze Assistant that takes action instead of answering questions; a self-updating Smart CRM that logs calls, emails, and meetings on its own; Context Home; Marketing Studio with its Campaign, Content, and Nurture Agents; updated Deal Progression and Prospecting Agent; a new Revenue Agent for collections; voice and batch testing for Customer Agent; Agent Builder and Agent Hub; the HubSpot MCP Server and Agent CLI; and the first CRM integration with ChatGPT Ads, paired with an OpenAI bundle for Starter customers. Every one of them runs on the same fuel.
HubSpot published the cost of running them without it. Customers who use AI with high-quality context generate 5x more MQLs, close 4.7x more deals, resolve 6.8x more tickets, and get 7.6x more replies to marketing email than customers with low-quality context, according to HubSpot's Spotlight page. In the UNBOUND keynote, Yamini Rangan added the survey behind it: 6,000 companies, 90% using AI, 6% seeing transformational results. That 6% is four times as likely to hit its revenue targets. Her line for the other 94%: "AI with bad context is worse than no AI."
We've been saying a version of this for a year. Your company is already intelligent. It just isn't exposed yet. HubSpot now has a name for the exposed part, and a scorecard for it.
Growth Context is everything HubSpot knows about your business, your customers, and your team. Duncan Lennox broke it into three layers on the UNBOUND stage. Business context is your positioning, your company profile, and your product list. Customer context is your conversations, your ICP, and the intent signals coming off your accounts. Team context is roles, approvals, and how your people work. Context Home, new in the Spotlight, scores how complete each layer is, flags the gaps, and suggests updates as meeting notes and transcripts come in. You approve what gets used.
That maps onto our architecture without a seam. We call the whole thing your intelligence layer, and we call the system of record under it the Kernel. HubSpot is the Kernel for most of the companies we work with. Business context is your brand written as a prompt. Customer context is the CRM data every agent calls first. Team context is process-as-code: workflows, approval rules, and definitions an agent can act on without asking. Growth Context is your Kernel.
One place we read the keynote differently. HubSpot plotted 50 AI use cases by popularity against impact, and producing content and sending email landed in the popular, low-impact corner. We agree with the data and disagree with the cause. Those use cases test low because the average company runs them with nothing in Context Home. Low impact is low context. HubSpot's own 7.6x email number says the same thing from the other side, and we take that argument up in full in the Build demand post below.
Four pieces of the Spotlight sit under every demo HubSpot ran on stage and show up in all three outcome tracks. Each one has a minimum level on the Maturity Model, and each one has a single prerequisite most portals haven't met.
Context Home. L3. The prerequisite is positioning that exists in writing. Context Home can only score what you give it, and most companies keep their positioning in the founder's head and a deck from 2023. Write the ICP and the tone rules down once, and every agent in the portal reads them. That's the job of our Brand Systems as Prompts engagement, and it's the fastest way to move the Context Home score.
The self-updating CRM. L3. Every call, email, and meeting lands in the record without a rep typing it. The prerequisite is activity logging turned on for every rep and a clean pass on the data model health check. A self-updating CRM that updates the wrong properties is a faster way to make a mess. This is HubSpot implementation done as Kernel work: property hygiene and stage definitions before anything else.
Breeze Assistant. L3. The rebuild takes action now. It builds custom tools, and it can run on a schedule. The prerequisite is one prompt that has returned a correct answer on your data. Start with the Monday pipeline review or the renewals-at-risk list. When the answer matches what finance would say, schedule it. When it doesn't, the gap it exposes is your next Kernel fix.
Agent Builder. L3, on the way to L5. Describe an agent in Breeze, review the plan, approve it, and tag it into a conversation like a teammate. The prerequisite is a workflow with clean inputs and one human who owns its output. Our Breeze Agent Library is the productized version of this: agents tuned to the operational patterns we see across portals.
The developer side of the Spotlight is the part most recaps skip, and it's the part that changes what a HubSpot portal is. The HubSpot MCP Server, the Agent CLI (public beta), and Marketplace connectors for Claude, ChatGPT, Copilot, and Gemini let an agent outside HubSpot read and act on your portal under governance.
That's L4 on our model. We call the pattern CompanyMCP (the Model Context Protocol spine that connects every interface, from Claude to Slack to your customer portal, to one intelligence layer). The prerequisite is app governance: scoped access per agent and an admin who owns the permission set. Without it, an external agent is a new way to leak data. With it, your company answers questions in whatever tool your buyer or your team happens to be using. Custom MCP Servers is where we build that for clients who need their own tools exposed alongside HubSpot's.
The Maturity Model has five levels. Each one has a sign you can check this week using something that shipped at UNBOUND.
Most B2B companies live at L1 or L2. L3 is where the pain stops. It's also where HubSpot's multipliers start, because every number on the Spotlight page compares high-context customers with low-context ones, and high-context is another word for L3. We get companies from L1 to L3 in 90 days.
Mark Moffat, CEO of IFS, made the adoption point on the UNBOUND customer panel: a powerful platform nobody uses compounds bad data, and then AI has no chance. That's L2. The fix is rarely another tool.
HubSpot sorted the Spotlight by outcome on the UNBOUND stage: build demand, win deals, delight customers. Top performers in its survey run only four or five use cases per outcome. We took each outcome, kept HubSpot's count, and tagged every use case with its layer, its minimum level, and the one prerequisite that has to be true first.
Build demand. Marketing Studio, Content Agent, Nurture Agent, Campaign Agent. The post makes the case that content and email are high-impact use cases when the Kernel is loaded, and shows the before-and-after.
Win deals. Deal Progression, Prospecting Agent, buying groups, Revenue Agent. Kim Cormier of RevenueWell took her team's meeting booking rate from 22% to 40% with signals plus Prospecting Agent, per HubSpot's case study, and put the year-over-year jump higher still on the UNBOUND panel. The post covers what her Kernel had that yours may not.
Delight customers. Customer Agent, Knowledge Base Agent, the churn risk signal. A safety sequence for switching on an agent that talks to your customers: test, coach, hand off.
HubSpot's Fall 2026 Spotlight, released at UNBOUND in Boston on September 16, 2026, holds 134 updates. The headlines are a rebuilt Breeze Assistant that takes action, a self-updating Smart CRM, Context Home, Marketing Studio with Campaign, Content, and Nurture Agents, updated Deal Progression and Prospecting Agent, a new Revenue Agent, voice and batch testing for Customer Agent, Agent Builder and Agent Hub, the HubSpot MCP Server and Agent CLI, and the first CRM integration with ChatGPT Ads. HubSpot also announced an AI Growth Bundle with OpenAI. UNBOUND is the new name for INBOUND, HubSpot's annual conference.
Growth Context is HubSpot's term for the combined knowledge of your business, your customers, and your team that its AI features draw on. It has three layers (business, customer, and team context) and a home in the portal, Context Home, that scores completeness and flags gaps. In Impulse's architecture it corresponds to the intelligence layer, with HubSpot as the Kernel underneath.
Yamini Rangan used the maxing phase for the period after a new technology arrives, when everyone rushes to do more with it: more tools, more agents, more tokens. The outcomes era is what follows, when companies pick a few use cases tied to a business outcome and build the context foundation under them. HubSpot's survey found top performers run four or five use cases per outcome.
It runs, and it answers from whatever it finds. With stale or missing properties the answers are confidently wrong, which is the L2 sign in the list above. HubSpot's own finding is that AI with bad context performs worse than no AI. Clean the data model first, then schedule the prompt.
Your positioning, company profile, and product list in the business layer; your personas, ICP, and intent signals in the customer layer; your roles, approval rules, and methodologies in the team layer. Write it as instructions an agent can follow, the way you'd brief a new hire, rather than as a brand PDF.
Breeze Assistant is the interface: you type or talk, it pulls in tools and takes action. Agent Builder is where you create reusable agents, each with a defined job, that Breeze and your workflows can call. Breeze is the teammate you talk to; Agent Builder is how you hire more of them.
CompanyMCP is our term for a company whose intelligence is exposed through the Model Context Protocol, so any interface (HubSpot, Claude, Slack, a customer portal, a custom agent) plugs into the same source of truth. HubSpot's MCP Server and Agent CLI are the pieces that make a HubSpot portal part of it. It's the L4, Exposed, level on our Maturity Model.
Curious where your company sits on the Maturity Model? Book the Diagnostic. We'll whiteboard what your company already knows, what it's forgetting, and which layer to build first.