That is half three of a three-part collection on how HubSpot remodeled with AI. Half one covers how we construct with AI. Half two covers how we develop with Agent-first GTM.
Constructing the proper engineering platform and rebuilding your go-to-market movement are meaningless if the group working them isn’t prepared. That’s the half most transformation playbooks skip. It’s additionally the half that determines whether or not any of it sticks.
We didn’t skip it; we doubled down. Consequently, 94% of HubSpotters use AI weekly, workers have constructed over 3,900 AI brokers, and our expertise profile seems to be essentially completely different than it did three years in the past.
That is our playbook for HubSpot’s organizational transformation that made all the things else potential.
Stage 1: Constructing AI Fluency (2023–2025)
The primary stage is about fluency throughout the complete group, and it has to begin with dedication from the highest. Leaders need to mannequin the conduct, share their very own experiments, and create the situations for everybody else to observe, not mandates.
We ran three performs to get there, and every is repeatable for any group:
Present the toolset. Each HubSpotter acquired enterprise licenses for a core set of AI instruments. A central AI technique staff manages vendor relationships, units safety requirements, and streamlines adoption of latest instruments, which eliminates procurement and safety bottlenecks that gradual transformation at most corporations. AI fluency can’t be a aggressive benefit you reserve for sure groups. It must be a baseline expectation for all groups.
Shift the mindset. This included fostering a tradition of experimentation, through which workers really feel empowered to try to to embrace new methods of working. We up to date our firm values to encourage this angle, including ‘be daring, be taught quick’ as a core worth. Staff share use circumstances and experiments in devoted chat channels. Leaders take part alongside their groups, typically getting reverse-mentored by folks additional alongside of their experimentation, and executives share their very own learnings in weekly updates. We additionally modified our organizational clock pace, shifting from annual planning cycles to six-week sprints to maintain tempo with the know-how.
To trace our progress, we additionally set a transparent, company-wide utilization objective: 80% weekly energetic AI utilization by the top of 2025. Then we tracked it overtly — by staff, by device, by use case — and made the information seen to everybody. Transparency drove accountability in each instructions: groups that had been behind had a transparent sign, and groups that had been forward grew to become fashions for others.
We need to be clear about why we tracked utilization reasonably than outcomes at this stage. Stage 1 was about constructing AI fluency. You may’t measure final result enchancment from instruments folks aren’t utilizing but. Utilization was a number one indicator, not the vacation spot. This wasn’t tokenmaxxing; it was a crucial step on the best way to outcome-maxxing in Stage 2.
Construct the skillset. We carved out protected time for studying. This included hackathons and 20 company-wide AI studying days in 2025. AI was woven into onboarding from day one and into ongoing supervisor improvement. The objective was easy: shift the query from “ought to I exploit AI for this?” to “how do I exploit AI higher?”
The end result of Stage 1 was a brand new expertise profile. By the top of this stage, we had a corporation that was changing into AI-fluent, with 94% of HubSpotters utilizing AI weekly, with over 3,900 AI brokers created by workers to enhance their very own work.
Stage 2: Workforce-Degree Transformation (2025–Current)
When workers every use AI in several methods for various use circumstances, you get particular person productiveness however not enterprise outcomes. To attain team-level transformation, you want clear priorities with actual accountability behind them.
To start out, we plotted groups towards two dimensions:
- AI maturity: How have they adopted instruments? Are they seeing measurable outcomes?
- AI readiness: What’s the potential of the staff’s work for automation? Is there enterprise threat? Are the information infrastructure and AI capabilities there to help?
That evaluation produced three classes for us: Tempo setters, or groups that had been already shifting quick. We don’t need to gradual these groups down; we need to help them. Close to-in wins, or groups which have apparent automation alternatives however haven’t acted. The bottleneck for these is nearly all the time management consideration, not tooling. And lastly, Massive bets. These are the groups with highest potential however probably the most dependencies. They want devoted funding in information, techniques, and alter administration.
Right here’s the place our groups fell, every requiring a special playbook:
Tempo setters: Engineering, Help, and Advertising had already seen main productiveness and effectivity positive factors by way of confirmed AI use circumstances, management sponsorship, and measurement. They wanted minimal help and continued their momentum by way of AI fluency investments.
Advertising is the clearest instance. The staff reimagined workflows throughout the board: AI-powered e-mail personalization drove an 82% enchancment in e-mail conversions, an AI chatbot now handles over 82% of web site inquiries and generated 10,000+ gross sales conferences per quarter by This fall 2025. A video advert manufacturing check delivered AI-generated spots at $300–$3,000 versus $300K–$500K with conventional manufacturing, and AI-assisted weblog manufacturing reduce author hours per article by 60%.
Close to-in wins: Recruiting and Operations had clear automation alternatives that could possibly be unlocked with the proper instruments. The important thing lever was management consideration: “gemba walks,” stepping into the work alongside groups to establish precisely the place AI may substitute or increase particular duties, and driving adoption hands-on reasonably than from a distance.
An instance of that is Expertise Acquisition. By embedding AI immediately into the hiring funnel, we noticed a 10-day discount in time to rent and a 30% discount in software evaluate time. We absolutely automated 80% of interview scheduling duties, leading to a 90% improve in scheduling quantity with no extra headcount. The share of sourced hires from previous candidates grew from 8% to 18% within the first 90 days, a direct results of AI resurfacing expertise that will have in any other case been invisible.
Massive bets: Gross sales, Customer Success, and Product has the best potential however wanted important funding in information, techniques, and alter administration. These groups acquired devoted AI pods, cross-functional groups of purposeful specialists, information scientists, and ops engineers targeted on reimagining particular workflows by way of speedy experimentation and iteration.
The deeper lesson of Stage 2 is that not each staff wants the identical help. The maturity and readiness evaluation is what tells you the place to push, the place to help, and the place to take a position. With out it, you find yourself making use of the identical strategy in every single place and questioning why solely a few of it really works.
Stage 3: Institutional Transformation (2026 and Past)
We’re early in Stage 3. However the path is obvious, and it is going to be an important stage of all.
Phases 1 and a pair of solved for particular person and staff productiveness. Stage 3 is about constructing institutional AI. The excellence issues. Making each worker 10x extra environment friendly doesn’t make an organization 10x extra productive, until the establishment itself is redesigned round new AI capabilities.
The muse of Stage 3 is institutional context. It means giving everybody entry to the proper instruments, information, and knowledge, and encoding firm processes into brokers that may act on them at scale.
The distinction turns into seen in how work will get achieved everyday. When an engineer wants context on a codebase, they don’t ask a colleague; they ask HubSpot’s inside coding agent. When a gross sales supervisor needs to know why a deal stalled, they don’t pull a report; they ask our native Guided Promoting Assistant. When a brand new rent wants to know how HubSpot makes choices, they don’t look ahead to onboarding; they ask our inside AI device. That’s what institutional AI seems to be like in apply: the collective context of the group, obtainable to everybody, for the time being they want it.
Transferring to this stage additionally requires confronting questions that earlier levels don’t. When brokers personal steps in a workflow end-to-end, governance issues extra. Who can see what? What choices require human sign-off? How do you catch dangerous outputs earlier than they compound? We’ve needed to construct for these questions intentionally, establishing clear permissions, audit trails, and escalation paths in order that the pace of brokers doesn’t outpace our skill to supervise them.
We’re nonetheless on this journey. However we perceive what’s at stake. The businesses that construct institutional AI are those that can have a bonus. However to do it, don’t begin with AI. Begin with the work. Discover the workflow that’s gradual, costly, or brittle. Discover the staff that’s most prepared. Run the experiment, measure it truthfully, then decide to what the information exhibits.
AI transformation begins with a powerful basis
The identical precept runs by way of all the things on this collection: the instruments are simply the start line. Constructing the inspiration – technically, structurally, and culturally – is what means that you can scale.
In engineering, that basis is a platform. In go-to-market, it’s a flywheel. In how you use, it’s the group itself. The businesses that determine this out gained’t simply use AI higher, they’ll develop higher.
