You don’t need to have AI figured out to run an offsite that moves AI adoption forward. That said, you’re likely being asked to build a one-day agenda to discuss a technology that changes monthly, for a room of enthusiastic users and people hoping this AI transformation wave passes soon.
It’s not an easy task. Getting the agenda right comes from knowing which conversation your particular group can finish, and building the day around that one thing.
At Clarinet, we ask every client two questions before designing an offsite:
- Who is in the room?
- What is the hardest thing about AI at your company right now?
"I often find that a lot of offsites are really focused on either the first or the second, but they don't think about the two in combination," says Diane Sadowski-Joseph, Clarinet's Co-founder and Head of Product.
The first question matters because it tells you what the room has the authority to decide. The second tells you which problem is worth the day. These are the foundation of any offsite agenda focused on AI adoption. How you refine from there depends on what type of offsites you’re running, typical challenges unique to that format, and your goals.
Four types of offsites
Based on the type of offsite you’re running, build the agenda around what is broken for that specific group rather than around activities. Then decide what happens afterward before you decide what happens in the room.
Below, for each of the four types (executive, all-leadership, team, and all-company) we share who is in the room in the context of AI adoption, typical goal, where it can go wrong, and some ideas for where to focus.
The executive offsite
Who's in the room: The small group that can commit resources.
Typical goal: Alignment on AI direction. What AI-native means for your company, by when, and what you will spend.
Where it goes wrong:
- Debates on tools instead of discussing the destination.
- Alignment is assumed rather than measured, so everyone nods and nobody agrees.
- Leaving with a handful of tactics but no shared vision and no owner.
Success story: Vouch, a 250-person insurance company, came in with under 40% of people using AI weekly and a stalled internal argument. Their tech-forward contingent thought the company was moving too slowly. Their traditional insurance side saw no reason to rock a boat that had floated for a century. Diane calls this pattern the swirl: "How do we stop having the same conversation over and over again? Is it safe? Should we do it? I don't know. Maybe we should…"
The fix was making it concrete. Every leader placed the organization on the AI Adoption Spectrum where it sat that day and where they wanted it in six months, then the group worked through the disagreements that surfaced. Three months later, 89% of employees reported productivity gains from AI.
Where to spend the time:
- Have each leader place the company on the spectrum privately, then show the spread.
- Agree on the vision and the message the whole organization will hear.
- Name an owner for each key decision.
The all-leadership offsite
Who's in the room: The managers who translate direction into behavior.
Typical goal: Cascade the direction, get managers to own it, and build fluency at the manager level.
Where it goes wrong:
- Executive direction gets diluted in translation, so managers relay a vibe instead of an expectation.
- The skill spread in the room is enormous, and content pitched to the middle bores the top and loses the bottom.
- Managers are asked to champion something they cannot personally do yet.
"If you tell your managers, now go teach your employees how to use AI well, and they're thinking, but I don't know how to use AI well…That is a very challenging scenario to be in," says Diane.
Success story: Juniper Square built their leadership offsite around two things. First, clarity on what the company expected of leaders in the transition, what it did not expect, and what support existed. Second, fluency training for the leaders themselves rather than talking points about fluency.
"The leader does not have to be the strongest AI user, but they need to know enough to have good AI instincts," Diane says. AI initiatives launched across every team with internal owners, and the full AI fluency training is part of every new hire's onboarding.
Where to spend the time:
- State what you expect of leaders, what you do not, and what resources they have.
- Give leaders their own training, not a script.
- Send every manager out with one dated commitment.
The team offsite
Who's in the room: The people who share a workflow.
Typical goal: Redesign how this specific team works, and ship one or two things.
Where it goes wrong:
- Hackathon energy with no owner. Lots of ideas, but nothing actually happens.
- Building something that falls apart in practice, because building with AI is its own skill.
- Spending the day debating what to do instead of doing it. And nobody has room for homework afterward.
Success story: Start from where the team feels friction, then run what we call “workflow forensics.” Break the workflow into three parts: the inputs, the judgment in the middle, and the outputs. That structure forces the group to surface the context that lives in people's heads and never got written down, which is the context AI needs and does not have.
The side effect is the real payoff. "Every single time we do it, someone ends up cutting or eliminating a step before AI even gets involved," Diane says.
Then spend most of the day building, with expert help beside the team rather than doing the work for them. That can be an outside partner or a strong internal builder from another team. Either way the team owns what gets made, because, "There is no such thing as a plug and play AI workflow," Diane says. "So it's better if you know how it was made."
Where to spend the time:
- Map current frictions before proposing solutions.
- Reinvent the workflow before automating it.
- Ship one or two V1s with named owners.
The all-company offsite
Who's in the room: Everyone, at wildly different starting points.
Typical goal: A shared baseline and language, permission, and momentum at scale.
Where it goes wrong:
- Inspiration with nothing concrete behind it.
- Trying to build before the organization has agreed on what people are allowed to do.
- All talk and no hands on keys, which signals that adoption is somebody else's homework.
Success story: Retention.com ran a company-wide build-a-thon, but the move that made it work happened beforehand. They assessed AI fluency across the company, sorted people into bands, and gave the lower bands training in advance. Everyone walked in with a baseline instead of a deficit.
"If there's one momentum killer, it is people feeling scared and behind," Diane says. "And you don't want everyone all in the same place feeling scared and behind at the same time." The session ended with fluency up across the whole company and one new workflow shipped per department.
Where to spend the time:
- Assess fluency levels before designing the day.
- Pair the day with clarity on what is policy and what is guideline.
- Offer shared training plus tailored support by level.
Before your next offsite
Come back to the two questions we recommend as your starting point: who is in the room, and what is the hardest thing about AI right now. Three decisions follow from those answers:
- Decide the next steps before you pick the activities. Working backward from what you want to be true a month later rules out most of the agenda ideas on its own.
- Build the agenda around what this group is uniquely positioned to offer. An executive team cannot redesign a workflow and a team cannot approve a budget. Asking a room for something it has no authority to give is how offsites turn into venting.
- Favor immediate progress over eventual progress. One shipped workflow with an owner beats ten ideas in a document.
And if you are having the same conversation for the third time, the format is the problem. Interrupt the pattern, and consider bringing in someone from outside to stop the swirl.
None of this requires you to be the most AI fluent person in the room. It requires you to design the right conversation for the people in it. "Part of the value of the exercise is not the answers," Diane says. "The value is the discussion."
