Unbound, Unlocked: What HubSpot's First Unbound Conference Taught Us (2026)
The biggest lesson from HubSpot's first Unbound is simple: AI is only as good as the context, data and people behind it. Leadership said it from the Partner Day stage in Boston, and our podcast guests had been saying it all summer.
This year Inbound became Unbound. HubSpot rebuilt its flagship event as an "un-conference" in Boston, September 16–18, and rebuilt much of its platform around what it now calls the outcomes era.
To get ready, we ran Unbound Unlocked, a summer series on the Sound of Inbound. We talked with speakers, Unbound Insiders and partners: George B. Thomas, the BabelQuest team (Chris Grant, Amy Brierley and Hannah Fisher), Jillian Olsen and Victoria Palacios, Josh Curcio of Protocol 80, and Tanya Wigmore and Klemen Hrovat. Earlier in the season, HubSpot's Kat Tooley walked us through how the new show came together.
Then, on Partner Day (September 15), HubSpot's leadership laid out where the platform and the partner ecosystem are headed. Put the two side by side and the overlap is striking. Here are seven lessons, and what we're taking back to our clients.
Context is the new currency, and bad context is worse than no AI
HubSpot's thesis for the outcomes era is that intelligence is necessary but not sufficient. Models keep getting smarter, but they don't know which deal is stalling, which customer is at risk, or that a product was retired last month. Context turns intelligence into outcomes.
The most surprising finding shared on stage: when HubSpot measured AI adoption across its customer base, good context improved every marketing, sales and service KPI it tracked. Bad context made every one of them worse than using no AI at all. A person naturally corrects for stale information. An AI will happily recommend a discontinued product or route an approval to a manager who changed roles.
HubSpot's answer is what it calls the growth context graph, built on three layers:
Business context: what you do — brand, voice, products, positioning.
Team context: how you work — roles, goals, processes and habits.
Customer context: who you serve — relationship history, ICPs, personas, contract and revenue details.
A new Context Home lets customers see and manage everything HubSpot knows about their business, with a score and suggestions for what to add.
Our guests were already living this. George B. Thomas said marketers skip the foundation: they get handed a powerful model and start prompting. He starts with identity documents — core values, beliefs, mission, personas — plus a story bank pulled from his own meetings and podcast episodes. Josh Curcio made the same point for sales: AI deal recommendations depend on logged calls and notes, and if reps don't log them, the AI has nothing to work with. Tanya Wigmore put it most plainly: to trust the AI, you first have to trust the data.
Start with the outcome and work backwards
A few months ago, the industry was obsessed with "token maxing" — using as much AI as possible. HubSpot's leadership says the conversation has moved to value and outcome maxing. In go-to-market, those outcomes are clear: build demand, win deals, delight customers.
The advice to partners was direct: deliver outcomes, not AI projects. Pick a clear use case, document the value the customer gets, and let that first win become the template for the next hundred.
That is exactly what Hannah Fisher and Chris Grant of BabelQuest brought to their Unbound session on turning your CRM into a decision engine. Hannah described clients who spend hours walking her through their process, step by step. She learned to stop building what was asked and start with the question the process is meant to answer. For one education client, all the tracking of leads, applications and enrollments came down to one goal: filling seats. The decision engine is the data that tells you a course isn't filling, so you can move ad budget there.
Chris framed it the same way. Stop agonizing over whether you got six MQLs or seven this week. Decide what you actually want — more sales from profitable sectors in less time — and build the pipeline view backwards from there. Hannah added a principle worth borrowing: adopt HubSpot, don't adapt it. The platform was designed by research teams; lean into it before reaching for a thousand custom objects.
Proof before purchase, adoption before AI
Buyers are changing how they buy. Jon Dick, HubSpot's Chief Customer Officer, described two groups, using Geoffrey Moore's adoption curve. The majority — roughly 80% of HubSpot's customers by the company's estimate — want to fix their go-to-market foundation and future-proof, but aren't ready to go all in on agents. They are deliberating longer, bringing larger buying committees, and they are nervous about switching: that ROI won't show up, that implementation will drag, that the landscape will shift mid-project.
The early adopters, which HubSpot calls go-to-market builders, want something different: proof before purchase. They don't buy on promise. They want to see AI work on their own data, in their own environment. CEO Yamini Rangan called this one of the biggest shifts of the year. SaaS used to mean evaluate features, implement, train, roll out. Now value comes before commitment.
For both groups, the first 90 days need to deliver undeniable value.
Our guests saw this coming from the adoption side. Amy Brierley told us adoption as an outcome was the single biggest factor in a deal BabelQuest had just won. As HubSpot moves up-market, larger companies face real financial consequences when a CRM rollout fails. Josh Curcio's Unbound session, Fix CRM Adoption: A Blueprint for Complex Sales Teams, named the two missing pieces in most deployments: measurement and accountability. Anyone, even an AI, can configure a CRM. Teams fail when managers don't measure usage or hold reps to it. It's easy to blame the software, he said, but adoption failure is usually an accountability problem.
Everyone is a builder now, and builders need guardrails
AI has flipped a decades-old rule. Businesses used to bend their processes to fit their software. Now a marketer, rep or ops lead can build an agent or automate a workflow in plain language. HubSpot is leaning in with a new agent builder, a reimagined Breeze Assistant that picks the right agent for the job, and new campaign, content, nurture and prospecting agents.
But leadership was candid about the downside. Nancy O'Dowd, who leads HubSpot's partner programs, told a story every team will recognize. An AI-forward employee builds an amazing agent and shares a celebratory Slack post. A month later, they leave for a startup. A few weeks after that, the agent stops running, and nobody knows where it lives, who has access, or how to fix it. The easier it is to build, the harder it is to trust what's been built. HubSpot calls this agent sprawl, and its answer is Agent Hub: one place to see, manage and build agents, all working from the same context.
Our guests have been building with exactly this tension in mind:
George B. Thomas went from zero case studies in nearly four years to seven in two days. A client form feeds a chain of named agents for writing, design and development. His rule: when you give AI hands, you need human eyes in the loop, so everything lands in draft first.
Klemen Hrovat builds production agents. His team shifted from building an app to building agents after clients told them they couldn't rely on AI-generated deal summaries. To run AI in production, he said, you need full control over what context goes in.
Tanya Wigmore builds guardrails right into the interface: locked-down fields, forced associations, and rules about what agents can and cannot update.
Jillian Olsen offered a simple ratio: let AI get you 80% of the way, then put your own perspective into the last 20%.
The CRM goes headless and open
HubSpot wants customers to get outcomes however they choose to work, including when no human logs in at all. Duncan Lennox, HubSpot's Chief Product & Technology Officer, described a three-layer API strategy:
| API Layer | What You Get | Example: "Which deals are at risk?" |
|---|---|---|
|
Data APIs
|
Raw access to CRM records
|
Pull every deal into another tool and figure it out yourself
|
|
Context APIs
|
Insights HubSpot has already computed
|
"Give me my ten most at-risk deals"
|
|
Work APIs
|
A HubSpot or custom agent drives the outcome end to end
|
"Find the at-risk deals and give me an action plan for each"
|
HubSpot also laid out its track record: first CRM to support MCP, first CRM connector for Claude, an Agent CLI for bulk scheduled work, and connectors for all four leading frontier models. The company says it is now the top CRM connector on both Claude and ChatGPT, and it committed to API parity with everything users can do in the app.
Our guests are already working this way. Amy Brierley uses the Claude connector for reporting and analysis, and used it to check BabelQuest's ICPs when HubSpot rebuilt its AI context settings. Josh Curcio predicted before the show that this year's story would be headless CRM, CLI and APIs connecting outside models to HubSpot data — and that's what arrived. Chris Grant raised an idea worth sitting with: maybe every content asset needs two versions, an interactive one for humans and a clean, text-based one for the LLMs.
Human creativity and brand become more valuable, not less
Duncan Lennox, HubSpot's Chief Product & Technology Officer, showed two emails side by side. One was generic AI filler — don't bother reading it, he said. The other drew on the trial product, the competitor, the rep's own voice, the prospect's price concerns and the right teammate to loop in. His point: every sentence earned its place. The difference wasn't the model. It was context, and context starts with a brand that knows who it is.
We asked every guest whether AI-generated content makes human creativity more or less valuable. Every one said more:
George B. Thomas: it depends whose hands the tool is in. Give it to someone uneducated and unsystematized and you don't get magic. His formula: human-powered, AI-assisted.
Josh Curcio: you can spot pure AI content from a mile away. Human work brings differentiation, empathy and accuracy.
Klemen Hrovat: anyone can ask the same model for the same output. What people want is your take.
Jillian Olsen: skip the company blog that exists only to feed search engines. Readers want to see how you actually think.
George went further on brand: every person and every company has one, and the time you're not spending building it, it's eroding. In a world of AI-generated sameness, brand is the moat.
Community and expert help are not going anywhere
A year ago, the question was whether AI would replace services. Nancy O'Dowd answered it this year: demand for expert help has never been higher. The bottleneck for AI transformation isn't the software. It's trusted guidance to implement it. HubSpot says it is not building a large services arm; instead, it sees partners as its forward-deployed team.
Yamini Rangan gave partners four places to focus:
Own the data and context services. In a HubSpot survey of 6,000 customers and partners, only 20% said they had data ready to start on AI.
Deliver outcomes, not AI projects.
Go deep in an industry. Industry-specific agents and workflows are now within reach.
Transform how you operate, from statements of work to migration to training.
Our guests made the same case from the audience side. Asked whether community will outperform content over the next few years, Amy Brierley, Jillian Olsen and Klemen Hrovat all said yes. When there's this much information out there, people take recommendations from brands and people they already know and trust. Even the form is becoming a conversation. George B. Thomas thinks of forms as a conversation starter, not a conversion point. Tanya Wigmore sees chat replacing the form as the interface. The data still gets collected; it just feels more human.
The un-conference itself
The rename was about more than a new logo. Kat Tooley, HubSpot's VP of Global Events and Experiential Marketing, told us in April that the change had been two years in the making. HubSpot held it until it had a clear answer to "change to what, and why?" That answer came with its move to an agentic customer platform. Inbound was the methodology that built HubSpot. Unbound is the mindset for the AI era.
Kat also described who the show is really for now. It serves the whole go-to-market team: marketing, sales, service, success and RevOps. It also speaks to the whole person, with sessions on growing as a leader as well as serving customers. And she wants events judged on value, not volume. She would rather have fewer, deeply engaged attendees than a big registration number full of people who drift through for an afternoon.
That thinking shaped the format:
Hands-on labs got their own space. Academy Labs roughly doubled and moved into the Westin. These are 90- to 120-minute sessions where attendees deploy agents in a live HubSpot portal with a professor on hand. Sparks Labs returned for collaborative problem-solving.
Networking by who you are. New zones group people by role, by industry, and by identity communities that have been part of the show for years.
What stood out:
Pick your own adventure. Jillian Olsen summed it up: pack your days with sessions or wander the floor, and you'll still have a good time.
Smaller rooms. Amy Brierley pointed to the exchanges of 10–12 people around a single topic. They're far easier than working an open expo floor.
Hands-on learning. Working sessions like George B. Thomas's 90-minute Five-Layer Marketing System let you build rather than just listen.
The hallway. Klemen Hrovat told us he attends only three or four talks across three days and spends the rest meeting people. Jillian's team has landed clients from conversations in the food truck line.
Kat's view of why this matters in an AI year stuck with us most. She sees in-person events as the other side of the AI coin. The more AI personalizes and produces for us, the more customers want to meet the humans and brands behind it. Her rule at HubSpot: every minute AI buys back should be spent with a customer. The feeling she hoped people would leave with was relief. They should leave with a playbook, new connections, and proof that everyone else is working through the same shift.
George's advice for any first-timer holds for next year too: leave room for surprise and serendipity. Some of the best moments are the ones you didn't plan.
What we're taking back to our clients
For the life sciences and B2B teams we work with, Unbound comes down to five moves:
Audit your context before you buy more AI. Brand voice, positioning, ICPs, personas and product details should be current and in one place. Bad context costs more than no AI.
Name the question before you build the process. Every property, workflow and dashboard should answer a question that drives growth.
Make adoption a deliverable. Define what "used" means, measure it, and hold people to it. Clean, logged data is what makes the new AI features work.
Give every agent an owner and a guardrail. Draft mode first, human review before publish, and a clear record of who built what.
Protect the human 20%. Let AI draft. Keep the perspective, the story and the judgment yours.
HubSpot's tagline for the week was that when partners thrive, customers win. After a summer of conversations and a week in Boston, we'd put it a little differently: when context is clean and people stay in the loop, AI actually delivers.
Listen to Inveniv’s full “Unbound Unlocked” series on the Sound of Inbound.
Sources
HubSpot Partner Day sessions, September 15, 2026: Yamini Rangan, Duncan Lennox, Jon Dick and Nancy O'Dowd (session notes)