There was no shortage of AI at UNBOUND 2026!
It ran through the keynotes, product announcements, developer sessions, Partner Day and plenty of conversations in between.
But after a few days on the ground, the bigger story wasn’t simply that HubSpot is investing heavily in AI. That part isn’t particularly surprising anymore. What became much clearer was how HubSpot sees AI fitting into the platform, how it expects businesses to use it, and what that means for the people building on HubSpot.
There was a noticeable shift away from talking about AI as a tool you open when you need help with something. The conversation is moving towards AI as part of the infrastructure of a business: connected to the CRM, working from real customer and company context, identifying signals and increasingly taking responsibility for defined pieces of work.
At the same time, HubSpot is becoming more extensible, search and buyer behaviour are changing, customer success is moving closer to the centre of growth, and the role of HubSpot partners is evolving alongside all of it.
Individually, many of the announcements are interesting. Taken together, they point towards something bigger: the CRM is changing from a system that records what your business has done into a system that increasingly participates in what your business does next.
That distinction ran through almost everything we saw at UNBOUND.
One word came up repeatedly at UNBOUND: context.
The major AI models are available to everyone. Access to AI itself is no longer much of a differentiator. The difference is increasingly what that AI knows when you ask it to do something.
HubSpot describes this as Growth Context, bringing together what a business knows about its customers, company, team and operations so AI can work from something far more useful than a generic prompt. Context Home is designed to give businesses greater visibility and control over what HubSpot’s AI understands about them.
Think about the difference in practice. A generic AI model can understand how customer retention works. AI operating within the right CRM context could potentially know that a particular customer hasn’t engaged with the business for three months, has two unresolved service issues, previously bought a certain product, opened a recent campaign and has a renewal approaching.
The intelligence of the model matters, but so does the information it has available when making a decision. Most companies will have access to similar models and increasingly similar agents. What competitors don’t have is your accumulated customer context: years of conversations, buying behaviour, service history, preferences, relationships, outcomes and institutional knowledge.
For years, businesses have treated CRM data largely as something required for reporting. In an agentic environment, that same data becomes the raw material from which AI works. The competitive advantage may not be the AI model at all. It may be the quality of the context you can give it.
For the past few years, most business use of AI has started with a prompt. Write this email. Summarise this meeting. Research this company. Analyse this spreadsheet.
UNBOUND showed how quickly that model is changing.
With Agent Hub and Agent Builder, HubSpot is moving towards businesses creating agents with defined responsibilities that can use CRM context and participate directly in workflows. Instead of asking AI to help with an individual task each time, the goal is to give an agent a job.
That could involve researching accounts, monitoring a pipeline, identifying customer signals, completing repetitive operational work or dealing with routine service enquiries before escalating situations that actually require a person.
For businesses, that changes the starting question. Rather than asking, “Where can we add AI?”, it becomes much more useful to ask where people are losing time, where important things are being missed and which pieces of work could be handed to an agent.
There is a deeper distinction here. The useful unit of AI transformation isn’t the prompt. It’s the process.
Saving someone three minutes writing an email is useful. Redesigning a process so that an agent continuously watches for a particular signal, gathers the relevant context, takes an approved action and only involves a person when judgement is required is something else entirely. That is where AI starts changing the economics of how a business operates rather than simply making individual employees slightly faster.
CRMs have traditionally relied on people doing a lot of administration to keep them useful. Calls need to be logged, notes added, properties updated and records maintained.
The problem is obvious to anyone who has managed a CRM: people are busy, and CRM administration is rarely the most important thing on their list.
HubSpot’s move towards a self-updating CRM is intended to change that relationship. AI can increasingly capture context from calls, emails, meetings and other interactions, reducing the amount of information people need to manually enter.
This isn’t only an efficiency improvement. Better captured context improves what the person looking at the record understands, but it also improves what AI can understand and act on later.
There is an interesting flywheel hiding inside that. Better automatic capture creates better context. Better context makes agents more useful. More useful agents create more activity and information, which improves the context again.
The CRM starts becoming less like a database employees maintain and more like a living representation of the customer relationship. If HubSpot can get that flywheel right, the old complaint that “our CRM is only as good as the data people put into it” starts to become considerably less true.
One of the more interesting ideas from Dharmesh Shah’s UNBOUND keynote was the Age of the Builder.
AI is reducing the distance between having an idea and being able to create something from it. Developers can prototype and iterate more quickly, while people without traditional development backgrounds can build increasingly capable tools of their own.
That doesn’t make technical skill or experience irrelevant. It changes where some of the difficult work sits. When building becomes cheaper, deciding what deserves to be built becomes more valuable.
Understanding the problem, knowing the user, recognising edge cases, designing the right workflow and exercising judgement become the constraints. That has implications well beyond software development. We may see significantly more internal tools, niche products and highly specific solutions become commercially viable simply because the cost of creating them has fallen.
The scarce resource moves from execution towards understanding.
App Objects, UI Extensions, APIs, agents and AI-assisted development were all significant parts of the developer conversation at UNBOUND.
Together, they point towards a HubSpot that can increasingly be extended around the way a particular business or industry actually operates, rather than forcing every process into the same standard CRM structure.
The timing matters. AI is lowering the effort required to build at exactly the same time HubSpot is expanding what can be built on its platform.
For years, businesses have assembled technology stacks by buying another specialised SaaS product every time they encountered a specialised problem: a CRM, portal, quoting system, membership system, event platform, customer success platform, reporting platform, then another integration connecting each one.
As core platforms become more extensible and building becomes cheaper, the pendulum may start moving back from buying another platform towards extending the platform you already have.
Not in every situation. Specialist software will continue to make sense where the complexity warrants it. But the threshold at which another entire system becomes necessary is changing.
That leads directly into one of the areas most relevant to us.
HubSpot provides a horizontal customer platform, but businesses don’t operate horizontally. Associations have memberships, renewals, committees, events and CPD. Manufacturers have distributors, products, quotes and complex sales processes. Construction companies have projects, estimators, subcontractors and long commercial cycles.
Historically, those requirements have often led organisations towards specialised systems that then sit alongside the CRM. But there is another model emerging: industry software as a layer on the customer platform rather than an entirely separate destination.
It’s exactly the thinking behind NativelyAMS. Associations can use HubSpot as their core platform while NativelyAMS provides the association-specific functionality HubSpot doesn’t provide natively.
That becomes particularly interesting in an AI environment. If membership, engagement, event attendance, communications, service interactions and organisational relationships exist within the same underlying customer context, an agent has a much richer understanding of the member.
The architectural decision about where data lives therefore becomes an AI decision too. Every disconnected system potentially removes context from the intelligence trying to understand the customer.
Answer Engine Optimisation, or AEO, had a significant presence at UNBOUND.
People increasingly ask ChatGPT, Gemini, Perplexity and other AI systems questions that would previously have begun with Google. HubSpot presented research based on more than one million AI responses to understand what influences whether brands and sources appear in those answers.
That introduces another audience for digital content. A website still needs to make sense to the person reading it, but its information also needs to be clear enough for AI systems to understand, retrieve and potentially cite.
There is a deeper change happening here than “SEO for ChatGPT”. For twenty years, businesses designed websites partly around persuading somebody to click. AI answers increasingly remove the click altogether.
That means being understood may become as important as being visited.
A company could influence a buying decision without the buyer ever visiting its website. Equally, a competitor could become part of the buyer’s consideration set because an AI system understands its offering better. That changes what visibility means.
The changes in search are part of a broader shift in how buyers research.
The familiar journey of search engine to website to form submission to salesperson was already becoming less reliable. AI accelerates that change.
Someone might research a problem through ChatGPT, read discussions on Reddit, look at a company’s LinkedIn content, compare options through an AI assistant and speak to colleagues before the business they eventually contact has any idea they exist.
That creates an uncomfortable reality for marketers: the buyer may know far more about you than you know about the buyer.
By the time someone becomes visible in your CRM, a substantial part of the buying journey may already have happened elsewhere. Attribution will become less perfect at exactly the moment businesses want it to become more precise.
The response can’t simply be better tracking. Businesses will need to become more comfortable understanding influence rather than attempting to attribute every decision to a measurable click.
UNBOUND took that idea another step further.
OpenAI spoke about AI as an emerging marketing channel and the changing consumer journey, while HubSpot’s work around ChatGPT Ads points towards commercial activity happening inside AI experiences themselves.
That creates an important distinction. AI isn’t only something marketers use to produce campaigns faster. It is becoming part of the environment in which customers discover, evaluate and potentially purchase from businesses.
If that continues, marketers may eventually manage AI surfaces in much the same way they currently think about search, social and paid media. The difference is that the interface isn’t necessarily a feed or a list of results. It’s a conversation.
That could make context and intent significantly richer than the keyword-based targeting digital marketing has spent decades optimising around.
AI has already made producing another email, landing page or piece of content dramatically easier. That means simply producing more isn’t much of an advantage.
HubSpot’s direction with Marketing Studio and its specialist AI capabilities suggests the marketer’s role is increasingly about orchestrating the system: understanding the audience, deciding what should happen, coordinating channels, setting the right context and measuring whether the work actually produces an outcome.
There is an irony here. The more content AI can produce, the less valuable content production itself becomes. The scarce skills move towards strategy, customer understanding, positioning, taste and judgement.
AI can create ten campaigns. Someone still needs to know which campaign shouldn’t exist.
One of the most practical applications of AI discussed at UNBOUND was helping salespeople decide where their attention should go.
Sales teams have enormous amounts of information available to them, but finding the relevant piece at the right moment is difficult. AI can increasingly perform that analysis in the background and surface the signal instead.
Rather than giving a salesperson more information, the system can help identify what changed, why it matters and which account deserves attention.
This could fundamentally change what a good CRM interface looks like. Historically, CRM software has largely asked the salesperson to go looking for information. An intelligent CRM should increasingly do the opposite: bring the salesperson the small number of things that deserve attention today.
The interface moves from database towards decision queue.
Another strong theme was the role of AI after a deal closes.
With enough connected customer context, HubSpot can increasingly identify patterns around engagement, service, churn and expansion. A business might spot that a previously active customer has gone quiet, recognise that another account is showing signs of needing an additional service, automate parts of onboarding or use AI to handle routine support.
That moves customer success closer to the centre of the commercial system.
It also exposes one of the strange boundaries businesses have historically created in their systems. Marketing owns someone until they become a lead. Sales owns them until they become a customer. Service owns them afterwards. The customer, of course, experiences none of those departmental boundaries.
AI works considerably better when the data doesn’t either. A connected customer platform creates the possibility of treating the relationship as one continuous lifecycle rather than three departmental databases stitched together.
When buyers can access enormous amounts of information before speaking to a salesperson, the salesperson’s role inevitably changes too.
One of the ideas discussed at UNBOUND was the shift from selling towards helping people buy.
AI can provide information, compare options and perform research extremely well. That makes the human contribution more valuable when a decision becomes complex: understanding circumstances, applying judgement, navigating trade-offs and helping someone determine what is actually right for them.
AI doesn’t necessarily eliminate the salesperson. It eliminates some of the reasons buyers previously needed one. The remaining reasons are arguably the more valuable ones.
There is an inconvenient reality underneath almost every exciting AI demonstration: the AI needs something reliable to work with.
A CRM full of duplicates, outdated records, inconsistent properties and disconnected customer information doesn’t suddenly become useful because an agent has access to it. In some cases, AI simply allows a business to make the wrong decision faster.
But there’s a bigger consequence. Historically, poor CRM data created reporting problems and annoyed employees. In an agentic business, poor CRM data can create actions.
An inaccurate field might previously have produced the wrong dashboard number. Tomorrow, it could cause an agent to contact the wrong person, offer the wrong thing or take the wrong action.
The cost of bad data therefore rises as AI becomes more autonomous. Data quality moves from housekeeping to risk management.
As agents start performing real work, businesses need to think differently about governance.
If an organisation eventually has dozens of agents operating across marketing, sales, service and operations, those agents need defined permissions and responsibilities. Businesses need to know what information they can access, what they are allowed to change, when a human needs to approve an action and how an agent should escalate something outside its scope.
That starts to look surprisingly similar to managing people. An agent needs an owner, a job, permissions, boundaries, escalation rules and a way of determining whether it is performing well.
We suspect one of the next operational disciplines businesses will need to develop is effectively agent management: not managing the underlying AI model, but managing a digital workforce of specialised agents operating throughout the organisation.
Another interesting development was HubSpot Work, an AI-first approach to bringing projects, documents, tables, forms, agents and processes together while keeping them connected to CRM context.
The direction is worth paying attention to. HubSpot has traditionally been the place where the customer relationship is recorded and managed. Increasingly, it is moving closer to the actual work that happens around that relationship.
That potentially changes the boundary of what a CRM is. If customer data, communication, projects, workflows, documents and agents increasingly operate from the same context, the CRM starts looking less like one category of business software and more like an operating layer around the customer.
That is a much bigger ambition.
Partner Day made another shift particularly clear.
If HubSpot continues making the core platform easier and more capable, the value of a partner can’t simply be knowing how to configure HubSpot. The opportunity moves towards understanding the customer’s business deeply enough to design the right system around it.
That includes industry knowledge, process design, data architecture, integrations, custom development and increasingly the ability to identify where AI and agents can create a meaningful outcome.
There is an interesting paradox here: the easier the technology becomes to use, the more valuable deep business understanding becomes.
Clients may need less help clicking the buttons. They may need considerably more help deciding what the system should actually do. That changes what good HubSpot consulting looks like.
For years, HubSpot services have largely been framed around implementation, onboarding, integration, optimisation and ongoing support.
AI adds another layer.
A partner can now look at how work moves through a business, identify bottlenecks or repetitive processes, determine what context exists inside HubSpot and then decide whether an agent could take responsibility for part of that work.
That isn’t really “AI consulting” in isolation. It is closer to operating-model design, with AI becoming another capability available when designing the system.
The best AI opportunities probably won’t begin with an AI workshop. They’ll begin with somebody complaining about a process: something that takes six hours every Friday, leads that nobody follows up properly, a customer problem nobody notices until it’s too late, salespeople spending half their week researching, or valuable data nobody has time to look at.
Those are much more useful starting points than “we need an AI strategy.”
Find the friction first. Then decide whether AI deserves a job there.
With 18 themes, hundreds of sessions and a lot of new technology announced, it would be easy to leave UNBOUND with a very long list of features. We think the more important story is what happens when you connect them.
AI models are becoming widely available, which means access to intelligence itself becomes less differentiating and context becomes more valuable. As context becomes more valuable, the CRM becomes more important. As agents become capable of acting on that context, the quality of CRM architecture and data starts affecting not only reporting but actual business decisions.
At the same time, building is becoming cheaper and HubSpot is becoming more extensible. Businesses have more scope to extend platforms around specific problems rather than automatically adding another piece of software. And as execution becomes easier, human judgement, industry knowledge and understanding which problems are worth solving become more valuable.
That leads us to a slightly different interpretation of where HubSpot is heading: the CRM is becoming less of a system of record and more of a system of context and action.
It doesn’t simply tell you what happened with a customer. Increasingly, it understands what is happening, identifies what matters and helps determine what should happen next.
For businesses, that means the AI conversation shouldn’t begin with which model, agent or tool to adopt. Start with the business. Where does work get stuck? Where does knowledge disappear? What information is fragmented across systems? What does your team repeatedly have to work out manually? What signals are sitting in your data that nobody has time to watch?
Then look at the architecture underneath it. Do you have the context? Is it reliable? Is it connected? Can HubSpot understand enough of the relationship to make an intelligent decision? Only then does the agent become interesting.
That’s probably our biggest takeaway from UNBOUND 2026.
The next stage of AI isn’t about adding AI to everything. It’s about giving it the right context, a clearly defined job and a business problem worth solving.