What the session should produce, how Micron and a global data center operator designed theirs, and how to fit a working session into a leadership summit that is already on the calendar.
The short answer
An executive AI session for a leadership team should leave the room with three things: leaders who have applied AI to their own week, a short list of workflows the company will redesign at the organization level (each with an owner, a 30-day step, and a value metric), and a first draft of the operating rules for AI use. That takes a half day, roughly 70% of it hands-on, designed from interviews with the leaders who will be in the room. Fluency alone is the output of a lecture; decisions are the output of a working session.
What's inside
- What should an executive AI offsite actually produce?
- How does a leadership-team session differ from an open-enrollment executive program?
- What do leadership teams say in discovery?
- How did a global data center operator design its executive AI session?
- How did Micron sequence leadership enablement into an enterprise system?
- How do you run the session inside an existing leadership offsite?
- What does the pre-work look like?
- How do you measure whether an executive session worked?
- Who should be in the room?
- Frequently asked questions
Correlation One has run executive AI sessions for leadership teams at Micron, Colgate-Palmolive, Pacific Life, MGX, and others, usually as the first installment of a longer enterprise program. This post sets out what those sessions produce, how two of those companies designed theirs, and the practical constraints of running one inside an existing leadership offsite. It also addresses the question we are asked most often: how this differs from sending executives to an open-enrollment program.
What should an executive AI offsite actually produce?
The design question is not "what should leaders learn" but "what should exist at the end of the day that did not exist in the morning." We hold sessions to three outputs.
- Confirmed personal intuition. Each leader has taken a currently manual task from their own week and worked it with AI in the room, on the tools the company has deployed. Leaders who have not felt the leverage firsthand cannot drive adoption down through their organizations; this is the foundation, not the goal.
- A named set of organization-level workflow redesigns. A shortlist, typically three to six, of workflows worth solving once for the whole company rather than function by function. Each leaves the room with an owner, a draft 30-day next step, and a value metric, usually hours returned or cycle time.
- A first-draft AI operating charter. The priorities, guardrails, and operating principles the room generated while doing the work: what counts as a source of truth, who is accountable for what AI produces, and where a team can run on its own versus where it escalates. This is a working draft that the leadership team carries out and refines, not a policy document written separately.
The third output is the one that distinguishes a session from a briefing. Governance that is written in a conference room by a separate committee tends to lag the use cases it is meant to govern. Governance that falls out of leaders scoring their own workflows for data sensitivity and accuracy risk is grounded in the work from the start.
How does a leadership-team session differ from an open-enrollment executive program?
They answer different questions. An open-enrollment executive program at a business school serves an individual executive who wants a credential, a peer network across companies, and a conceptual grounding in where AI is going. A leadership-team session serves a company that needs its executives to make decisions about its own workflows, its own data, and its own operating model, together, with follow-through that is measured.
| Dimension | Open-enrollment executive program | Leadership-team working session |
|---|---|---|
| Unit of enrollment | One executive | The leadership team as a group |
| Material | Cases and frameworks drawn from many companies | The participants' own workflows, captured in pre-work |
| Tools | Vendor-neutral or demonstration environments | The company's deployed stack and licenses |
| Primary output | Credential, network, conceptual fluency | Named redesigns with owners and metrics; draft operating charter |
| What follows | The executive returns to the company | Function-level labs build what the room prioritized; results are presented back to the same leaders |
| Measurement | Completion | Pre/post confidence, 30-day follow-up on commitments, hours returned per redesign |
Many executives do both, and there is no conflict. The failure mode is treating the first as a substitute for the second: a leadership team in which several members hold certificates but the company has no shortlist, no owners, and no charter.
What do leadership teams say in discovery?
Every session we run is preceded by structured interviews with the leaders who will be in the room, typically six to twelve conversations over four to six weeks. Across dozens of these engagements, the same themes recur, and they shape the design more than any curriculum does.
- Leaders want value, not usage. Finance and operations leaders in particular have already funded licenses. They want hours saved, errors reduced, and workflows redesigned, and they are explicit that without near-term evidence, organizational momentum on AI is at risk.
- Employees are tired of being encouraged. Several uncoordinated adoption pushes, each from a different group, blur into one another. Leaders ask for a single coherent design rather than another campaign.
- Leadership needs its own programming. Leaders need to use AI in their own work, identify which workflows in their function can be AI-augmented, understand how roles change when teams include agents, and exercise judgment about approving an agent for a workflow with the right human review. None of that is served by the same training the broader workforce receives.
- The fluency spread in one room is wide. In the same leadership team, one executive runs most of their job on an AI assistant and another describes looking at the technology through a keyhole. Teaching to the floor wastes the former; teaching to the ceiling loses the latter.
- The real friction is in handoffs. The highest-value redesigns are rarely inside one function. They sit in the request-and-wait cycles between functions and, for global companies, between regions.
- Training without measurement is not training. CEOs ask for expected outcomes stated up front and measured afterward. That request sets the format: commitments stated aloud, anchored to a baseline, followed up at 30 days.
Two engagements illustrate how these themes become a design. Both companies approached executive enablement earlier and more deliberately than most of their peers.
How did a global data center operator design its executive AI session?
Global data center operator
Senior leadership team · Executive session fall 2026 · Impact Lab followed
9
leaders interviewed, including the CEO, before the session was designed
4 hrs
in-person working session for a senior leadership team of ~14
3 → 5
people per team building a working agent in the Impact Lab that followed
Discovery first, then design
Over a six-week discovery period, Correlation One interviewed nine of the company's leaders, including the CEO, ahead of a half-day in-person session for a senior leadership team of roughly 14. The interviews produced eight design requirements, among them: start with the art of the possible on real executive tasks; spend most of the session on individual use and workflow redesign rather than enterprise strategy; teach a repeatable method for finding and analyzing workflows; anchor everything in an actual day, week, and month; design for cross-functional and cross-regional handoffs across the Americas, APAC, and EMEA; weave guardrails through the work rather than lecturing on governance; give every initiative an owner and a metric; and treat the leaders who were already fluent as leverage rather than as an audience.
Resolving top-down and bottom-up
The CEO's mandate was top-down: do more with less and reimagine how whole parts of the business work. The leaders closest to the work made the opposite case, that the best use cases come from the people doing the work. The session was built to hold both. Individual productivity wins and team-level upgrades were harvested bottom-up in peer groups, then converged top-down into a small set of cross-functional, cross-regional redesigns the whole room named together. Because the room named them, the bets were chosen by the leaders who own the work rather than handed down. That shortlist, with the guardrails surfaced along the way, became the first-draft AI enablement charter.
A parallel track for power users
Rather than run the room at one level, the company gave its most fluent leaders a distinct role at each phase: presenting their own AI-ified use case as proof from inside the room, coaching peers during the hands-on blocks, and breaking off with an AI coach for a stretch challenge, such as an agent prototype or a chained multi-tool task from their own week, during foundational blocks they had already outgrown. They rejoined for every synthesis and commitment moment so the room stayed whole.
What the half day looked like
Pre-work: a baseline AI-confidence survey, and a function-level AI Opportunity Map in which each leader captured the core workflows their function owns and flagged the manual, handoff-heavy ones.
| Time | Block | Type |
|---|---|---|
| 10 min | CEO opening. AI as core to the strategy, in the CEO's own words. | Strategy |
| 45 min | The art of the possible. Facilitators work one or two currently manual executive tasks from the pre-work live; leaders try one on their own task. | Hands-on |
| 15 min | Power-user spotlight. One leader walks through a use case they AI-ified themselves. | Proof |
| 45 min | Decompose your own workflows. Steps, handoffs, stall points; scored for AI readiness using an input-process-output lens and four criteria (repeatable, rule-based, resource-heavy, risk-light), with data sensitivity and accuracy risk identified in the same pass. | Hands-on |
| 40 min | Peer review. Cross-functional groups hunt for three patterns: work common to nearly every function, manual custom work ripe for automation, and cross-regional handoffs where ownership is unclear. | Hands-on |
| 30 min | From workflows to org-level redesigns. The full room decides which nominated workflows are worth solving once for everyone. | Decision |
| 25 min | Draft the charter. Consolidate the guardrails the day generated into priorities, guardrails, and operating principles. | Decision |
| 10 min | Commit. Each redesign gets an owner, a 30-day step, and a value metric, stated aloud. | Decision |
What came next
The redesigns named in the room seeded the company's AI Impact Lab, which opened two weeks later with teams of three to five building and testing a working AI agent for their own team's workflow across four 90-minute sessions and an innovation showcase presented back to the leadership team. A second cohort followed. The sequence matters: the executive session set the frame, and the labs delivered the build against it.
How did Micron sequence leadership enablement into an enterprise system?
Micron
Semiconductor manufacturer · Discovery fall 2025 · Applied pilot Jan 2026 · Function-level labs 2026
CFO · CPO · CIO
plus manufacturing, supply chain, corporate and procurement leaders in structured discovery
50
non-technical employees built measured agents in the January pilot
5% → 0%
error rate on one pilot agent, with a weekly task cut from ~1 hour to ~5 minutes per user
From AI usage to AI impact
In fall 2025, Correlation One ran structured discovery with Micron's CFO, CPO, CIO, and leaders across manufacturing, supply chain, corporate functions, and procurement. One message was consistent: usage of AI tools had risen across the company, and the next step was to convert usage into impact. Nine themes came out of those interviews, and two of them concerned leadership directly. Leaders wanted tangible value from the AI investment already made: hours saved, errors reduced, workflows redesigned. And leaders needed dedicated, white-glove programming of their own, covering personal fluency, workflow identification, leading people-plus-agent teams, and the judgment to approve an agent for a workflow with the right human review.
Leaders as the activation spine
Micron's approach is notable for where it placed leadership enablement. Rather than treating the executive session as a standalone event, the enablement blueprint positioned leaders as the primary accelerant of adoption, with a specific mechanism attached: role-based AI playbooks codifying a small set of high-frequency, high-impact use cases per role family, deployed through the management hierarchy in a cascade (leaders model, directors reinforce, managers drive) and reinforced in short, live functional huddles. Executive enablement in this model is not a day of learning; it is the point at which the cascade starts.
Coordinating IT, enablement, and governance
Discovery also surfaced a pattern we see in most large enterprises: as business users try to build agents, they hit data-connectivity limits, and the enablement effort and the IT roadmap are not sequenced together. Micron's blueprint addressed this with a central, prioritized backlog of AI use cases assessed for impact, feasibility, and risk, and a steering body approving additional data access against that list, so that each wave of upskilling focuses only on what is currently available.
Evidence from the applied pilot
The leadership work ran alongside an applied pilot in January 2026 in which 50 non-technical employees built working agents for their own workflows, with baseline and post-build measurement on each. One team's agent, which fills equipment vendor core-return forms from existing records, reduced a weekly task from about one hour to about five minutes per user and took the measured error rate from 5% to zero. By spring 2026, the model had been extended to a front-end manufacturing Impact Lab with ten leadership-approved opportunity zones, a three-gate vetting process for ideas (leadership impact, IT feasibility, coach buildability), and a target of at least 20% improvement per workflow.
How do you run the session inside an existing leadership offsite?
Most leadership teams do not want a separate event. They want the session inside an offsite or summit that is already on the calendar, and that is usually the better setting, because the CEO's framing and the team's priorities are already in the room. The constraints are practical, and we have learned them the hard way.
- Time. Two and a half hours is the floor for a format that includes workflow identification, a prototype, and a pitch round. Three and a half to four hours is the working target. When a wealth-management firm asked us to design a session for about 120 leaders at its leadership growth summit, the first version was two and a half hours; the executive sponsor's own reaction on seeing the agenda was that it was too short, and the request went back to the summit organizer to find more time. A morning block, a working lunch, and a post-lunch pitch round is a format that fits a two-day agenda and recovers the low-energy slot after lunch.
- Room size and the pitch math. At 120 leaders in tables of eight, that is 15 teams. Fifteen pitches do not fit in 30 minutes. Either the pitch round is sized to the number of tables, or tables are paired, or the showcase is moved to a working lunch. Decide this before the agenda is printed.
- Licensing. Each table needs at least one person with a full license to the deployed tool, including agent-building capability, so that every team can identify a workflow, choose one, and begin building it. The larger risk is after the session: leaders who leave without the licenses they used in the room lose momentum within weeks. Resolve licensing before the session.
- Priorities, not a list. The sponsor of that summit put it plainly: the organization could not leave with ten priorities; two or three was the number. That discipline applies to the AI session as much as to the rest of the offsite.
- Customization is not optional. A life insurer came to us after a vendor-led session on agentic AI, sponsored by a technology provider, landed poorly with its business leaders. The content was not wrong; it was not theirs. The replacement brief asked for a three-hour session built on the company's own workflows and its own definition of return. Generic executive AI content, however well produced, is the single most common reason these sessions fail.
- Logistics. Laptops, Wi-Fi, and licenses are confirmed in writing a week out. Sessions have been lost to each of them.
What does the pre-work look like?
Pre-work is what makes a half day sufficient. Without it, the first ninety minutes are spent discovering what could have been known in advance.
- Leader interviews. Six to twelve structured conversations, including the CEO, over four to six weeks. The output is a synthesis of themes and the points on which leaders diverge, and every design choice in the agenda traces back to one of them.
- Baseline confidence survey. A short pre-session survey that sets each leader's starting point and anchors the pre/post comparison.
- Function-level AI Opportunity Map. Each leader captures the core workflows their function owns, flags the manual and handoff-heavy ones, and notes a rough business case and the regions each touches. These are the raw material the session decomposes; the map is refined in the room, not created there.
- Power-user identification. The interviews surface which leaders already run much of their work on AI. They are given a role before the day, not discovered on it.
- CEO framing. The opening is the CEO's, in the CEO's words. The room needs to hear the conviction directly, and a facilitator cannot substitute for it.
How do you measure whether an executive session worked?
Three measures, in sequence.
- In the room: pre/post confidence against the baseline survey, and the count and quality of named redesigns with an owner and a metric attached.
- At 30 days: whether each owner has taken the stated next step, and whether the deployed tool's usage data shows the leaders themselves using it on the tasks they worked in the session.
- At the end of the following lab cohort: hours returned or cycle-time reduction per redesign, measured against the baseline captured before the build, and presented back to the same leadership team that prioritized it.
The last measure is the one that justifies the session. A leadership team that prioritized a redesign in September and sees a working agent with a measured baseline in October has evidence; a team that attended a briefing has notes. Our ROI methodology sets out how we capture baselines and attribute returns across a program.
Who should be in the room?
The senior leadership team, with the CEO present for the opening and the commitments. In our experience the session loses value when the CEO delegates attendance, because the top-down frame, which is what makes the bottom-up work converge, has to come from the person who owns the strategy. Below the leadership team, the right next audience is the function leaders who will own the labs, and they are often invited to the synthesis and commitment blocks rather than the whole day.
Frequently asked questions
A half day, roughly four hours with two short breaks, is the working minimum for a session that produces decisions rather than awareness. A session can be compressed to two and a half hours inside a larger summit, but the pitch-and-commit block is the first thing to suffer, and leaders consistently ask for more time once they see the agenda.
A senior leadership team of 10 to 20 produces the tightest decisions because the whole room can converge on a shortlist. Sessions for 100 or more leaders at a summit work as a table-based format with eight per table, one AI coach circulating per three to four tables, and a pitch round that is sized to the number of tables.
Each table or working group needs at least one person with a full license to the tool the company has chosen, including any agent-building capability. The larger issue is after the session: if leaders leave without the licenses they used in the room, momentum stalls within weeks. Resolve licensing before the session, not after.
Whatever the company has already deployed. Sessions at Correlation One run on Claude, ChatGPT, Gemini, and Microsoft Copilot depending on the client's stack; the method of decomposing a workflow and scoring it for AI readiness is the same across tools, and the tool-specific layer is kept thin.
Design for it rather than teaching to the floor. The most fluent leaders get a distinct role: they present their own AI-ified workflow as proof in the room, coach peers during hands-on blocks, and break off with an AI coach for a stretch challenge during foundational blocks they have already outgrown.
Each named redesign leaves the room with an owner, a 30-day next step, and a value metric, usually hours returned or cycle time. Those redesigns seed function-level AI Impact Labs, where teams of three to five build and validate working agents over four to five weeks and present them back to the same leaders.
Yes, and it often works better there because the CEO's framing and the team's priorities are already in the room. The constraints are time, Wi-Fi, laptops, and licensing. A morning block followed by a working lunch and a post-lunch pitch round is a format that fits a two-day summit agenda.
Open-enrollment programs serve an individual executive who wants a credential, a network, and conceptual grounding. A leadership-team session serves a company that needs its executives to make decisions about its own workflows, data, and operating model, together, with measured follow-through. Many leaders do both; they answer different questions.
Planning an AI session for your leadership team?
We design every executive session from interviews with the leaders who will be in the room, on the tools your company has already deployed, with measurement built in from the baseline.
- Executive AI Enablement: nine capabilities and the Agentic AI for Business Leaders track
- AI Impact Lab: function-level workflow redesign with working agents
- Measuring AI training ROI: how we capture baselines and attribute returns
- AI governance as a workforce capability
- Best enterprise AI training companies in 2026