HubSpot renamed its flagship conference from INBOUND to UNBOUND this year, after fifteen years under the old name. More than 13,000 people showed up in Boston to watch what the company called its biggest product release in years. Dozens of features shipped. Most of them will not affect your business.
We were there. As a HubSpot partner agency working across higher ed, agriculture, and professional services, we sat through the keynotes and the sessions with a specific question in mind: what actually changes for our clients. Three ideas came out of it, and they showed up again and again across marketing, sales, and product tracks alike.
Idea 1: AI is only as good as the data behind it
HubSpot's own research, presented by CEO Yamini Rangan in the opening keynote, is the most useful data point out of the entire conference: 90% of companies are already using AI in some form. Only 6% describe the results as transformative.
That gap is not about which tool you picked. HubSpot put a number on it. Comparing customers with strong data quality against customers with poor data quality, across the same AI features:
- Companies with clean, complete CRM data saw 264% more qualified leads.
- Companies with messy or incomplete data saw 28% fewer.
- The worst single number: connected customer calls dropped 86% for companies running AI on top of bad data.
Read that last one again. Bad data does not mean AI does nothing. It means AI actively makes things worse, faster, at scale. An agent trained on incomplete records does not know it is wrong. It just sounds confident while being wrong.
HubSpot's answer to this is a new feature called Context Home, which scores how complete your data actually is and flags the specific gaps limiting AI performance. It's a genuinely useful diagnostic. It is also, notably, not a fix. Someone still has to do the work of cleaning up years of inconsistent records, duplicate contacts, and abandoned properties.
What this means by industry:
- Higher education: recruitment and admissions data is often split across a CRM, a student information system, financial aid platforms, and whatever spreadsheet an individual counselor has been keeping since 2019. Before any AI feature (content generation, prospective student chat, enrollment prediction) will do anything useful, that data needs to live in one place and mean the same thing everywhere. An AI tool that drafts outreach based on incomplete applicant records will produce outreach that gets the details wrong, in a moment where getting it wrong costs you the applicant.
- Agriculture & Manufacturing: sales cycles in ag and manufacturing are long, relationship-driven, and often run through dealer or distributor networks rather than direct contact. That means the CRM record for a given account is frequently incomplete by design, most of the relationship happens outside the system. Before layering AI on top, the higher-value move is making sure what does get logged (past purchases, seasonal buying patterns, account history) is consistent enough to actually feed a model.
- Professional services: deal and quote data tends to be the cleanest of the three, because it's tied to billing. The gap is usually on the customer success and service side: support tickets, project notes, and account health information that lives in someone's inbox instead of the CRM. If you're evaluating any of HubSpot's new service or account-health features, that's the data to audit first.
Our take: We wouldn't recommend Context Home to a client yet as a standalone project. It's a good diagnostic, but running it before a real data audit just produces a longer list of problems without a plan to fix them. Do the audit first, in whatever tool you already trust, then use Context Home to confirm you got it right, not to discover it.
Idea 2: The future is one connected system, not separate marketing and sales tools
The clearest structural theme out of UNBOUND was HubSpot pulling functions that used to sit in separate tools into a single, shared system with a single source of data. This is the RevOps argument, made explicit at the platform level.
A few specific moves back this up:
- The Smart CRM now updates itself automatically from calls, emails, and meetings, so the record a salesperson sees is the same one marketing and service see, updated without manual entry.
- Marketing Studio 2.0 connects a company's AI-search visibility score directly to campaign execution: a marketer sees where they're losing visibility and can direct an agent to build a fix from the same screen, no handoff to a separate SEO tool or a separate content request.
- Agent Hub is positioned as a single view of every AI agent running across marketing, sales, and service, explicitly built to replace the patchwork of outside automation tools (Zapier, Make, N8N) that most growing companies accumulate over time.
- Revenue Hub generates quotes using full deal context pulled from the CRM. HubSpot reported one customer going from quote creation to signed agreement in under 15 minutes, because the quoting tool already had everything it needed instead of someone re-entering deal terms by hand.
The underlying argument: if your marketing team, sales team, and service team are working from different systems, or different views of the same system, AI cannot close the gap between them. It just runs faster in each silo separately.
What this means by industry:
- Higher education: the traditional split is admissions (marketing) versus enrollment counselors (sales) versus student services (customer success), often on entirely different systems from different vendors. A connected system means an inquiry captured on a landing page, a conversation with a counselor, and a question to the financial aid office are the same record, not three. That's the difference between a prospective student getting a generic follow-up and one that reflects what they actually asked about.
- Agriculture & Manufacturing: RevOps in ag and manufacturing usually means connecting field sales (often independent reps or dealers) with marketing content and with customer service after the sale. Revenue Hub specifically is built for high-volume, low-complexity quoting, and it is not the right fit for ag's variable, negotiated, relationship-based deal structure. The more relevant move here is making sure whatever CRM data exists is shared cleanly between whoever runs marketing and whoever runs the sales relationship, not adopting every new hub HubSpot ships.
- Professional services: this is where Revenue Hub is actually built to fit, specifically professional services firms with a limited, standardized set of SKUs or service packages. If your quoting process involves configuring dozens of variables per deal, this won't save you time. If your services or products are more standardized, connecting the quote directly to CRM deal data is a real efficiency gain, not just a feature.
Our take: Revenue Hub is the one announcement this year we'd actually push a Pro Services or Advanced Manufacturing client toward now. The 15-minute quote-to-signature example only works because the deal structure is simple enough for automation to help instead of get in the way. For ag clients specifically, we'd advise against Revenue Hub. It's the wrong tool for a relationship-driven, negotiated sales process, and forcing it in would slow deals down, not speed them up.
Idea 3: AI helps, but it still needs strategic direction and human judgment
The most repeated line out of the keynote was some version of "bad context is worse than no AI." HubSpot was explicit that none of what they announced replaces a strategist. Rangan's framing: companies getting real outcomes ask "what business result are we driving" before asking "where can we use AI," and they treat which tasks stay human as a deliberate decision, not an afterthought.
The clearest example: HubSpot's own data shows the tasks most teams automated first (drafting content, sending routine emails, prepping meeting notes) are the lowest-value use cases. The highest-value ones (prioritizing which deals to chase, flagging accounts at risk, deciding what a specific customer actually needs) all require judgment the model doesn't have on its own. Those are also the tasks HubSpot's own agents are built to assist, not replace, with a human review step built into most of the new features by design.
What this means by industry:
- Higher education: AI can draft admissions content and personalize outreach at a scale no team could match manually. It cannot decide what your institution's actual differentiators are, or which student populations you should be prioritizing given enrollment goals and Title III or HBCU-specific funding considerations. That strategic layer has to come from someone who knows the institution, not the model.
- Agriculture & Manufacturing: AI can help a smaller marketing team produce more content and follow up on more leads than they could otherwise handle. It cannot know which relationships matter most in a market where trust is built over years and a single bad interaction with a distributor can cost an account. Automating the wrong parts of that relationship is a real risk, not just an inefficiency.
- Professional services: AI-generated quotes and proposals move faster, but the underlying pricing strategy, service scoping, and account prioritization still require someone who understands the client relationship and the margin implications. Speed without judgment just means bad decisions happen faster.
Our take: The industries most tempted to automate the wrong thing first are the ones with the least room for a wrong guess: higher ed admissions and ag account relationships. Both run on trust built over a long cycle. We'd tell any client in those two verticals to keep a human reviewing every AI-drafted message before it sends, at least for the first two full cycles, not as a permanent rule, but long enough to know where the model actually gets it wrong for your specific audience.
The takeaway
Every tool HubSpot announced at UNBOUND 2026 is real. None of it is a shortcut. The gap between the 90% of companies using AI and the 6% getting real results isn't a tooling problem, it's a readiness problem: clean data, a genuinely connected system, and someone steering the strategy.
If you're evaluating any of what shipped this year, that's the order to work through it in: audit your data, connect your systems, then decide where AI actually earns its place in the workflow.