The CEO wants a growth strategy. Finance wants an investment decision. HR wants clarity about jobs and learning. Technology leaders want a manageable set of priorities. Put all four expectations under the heading “AI summit” and an impressive programme can still leave everyone waiting for a different answer.
An AI strategy summit for executives needs a shared decision brief before it needs a speaker lineup. Its purpose is to help the leadership team make a limited set of strategic choices, expose the assumptions behind them and assign the work needed to resolve uncertainty.
That purpose differs from a corporate AI learning summit, where employees need opportunities to practise and develop capabilities. Executives may need introductory learning too. But the distinctive contribution of their offsite is authority: deciding what the organisation will pursue, support, defer or stop.
What the evidence can contribute to the discussion
The October 2026 Wharton–GBK AI adoption report compares responses from managers and more junior employees, finding different perceptions of organisational alignment and training adequacy. Its online survey covered employees in large US private-sector organisations. These are separate groups' self-reports, not matched accounts of the same teams or a causal test of leadership practices.
That distinction matters when preparing a summit: a senior team's confidence should be checked against what people doing the work experience. The study does not show that holding an executive event improves alignment or returns.
In August 2026, MIT Sloan reported George Westerman's advice on AI transformation, drawing on executive conversations and case studies. It connects strategy with organisational change, technology foundations and workforce capabilities. This is useful expert context, rather than evidence for a particular conference design.
The seven questions below are our practical planning recommendations. They are not a validated assessment or a promise that a half-day meeting produces a successful AI strategy.
Prepare the decisions before preparing the slides
Ask the executive sponsor to distinguish decisions the summit can make from matters requiring further approval. An executive team may authorise a bounded exploration while a larger investment needs a board decision. Calling both “approval” invites confusion.
Prepare a concise evidence pack with three ingredients: current business priorities, what is actually happening with AI, and unresolved choices. Include operational friction, employee perspectives and disappointing results alongside successful demonstrations. Label estimates and assumptions clearly.
For each proposed decision, request a short note explaining the options, the consequence of delay, the evidence available and whose authority is required. A decision with no credible options needs more preparation. A question with no authorised decision-maker needs a different invitation list.
Include the leaders responsible for strategy, operations, finance, technology and people, plus relevant risk expertise. Bring in employee or customer evidence even when the decision session itself is small. Where candid discussion is difficult, gather concerns beforehand through an appropriate confidential channel.
Seven questions to put on the executive summit agenda
1. Which business ambition deserves priority—and what will receive less attention?
“Become AI-first” does not tell a team how to allocate scarce management time. Ask what the organisation wants to improve for customers, employees or its wider mission, and why AI is a plausible means.
Consider a fictional industrial distributor. Sales wants faster quotations to pursue more opportunities. Operations wants fewer order errors. Finance wants lower processing costs. Each may be legitimate, but they require different evidence and can place different demands on the same people.
The summit should make those competing aims visible. If quotation speed wins initial priority, what happens to the other proposals? Can they proceed independently, or must they wait for the same data and operational staff?
Leave with: a specific priority, the business rationale and an explicit account of the work being deferred. Agreement is more meaningful when it includes an opportunity cost.
2. Which assumptions could make that ambition a poor investment?
Give the room something more demanding than a polished success story. Ask each proposal's sponsor to identify the assumption most likely to undermine it.
For the distributor, faster quotation drafting may be irrelevant if customers wait mainly for stock confirmation. An AI demonstration could work perfectly while targeting the wrong constraint. Another assumption might be that staff can review additional quotations without delaying existing orders.
Invite operational colleagues to challenge the account. What do they know about exceptions, missing information or informal workarounds? What evidence would change the sponsor's mind?
Leave with: a short list of consequential uncertainties and a way to investigate them. This is not a request to eliminate uncertainty before acting. It is a choice about which uncertainty must be reduced before committing more resources.
3. What kind of commitment are we ready to make?
A promising idea can justify investigation without justifying a full deployment. Separate permission to learn from permission to expand.
Ask what the next commitment actually includes: funding, staff time, data preparation, procurement work, operational access or a change to customer service. Specify which commitments are reversible and which would be difficult to unwind.
Return to the distributor. Leaders might support testing draft quotations on approved historical cases while withholding permission to send generated quotations directly to customers. That is a meaningful decision if it answers a genuine uncertainty. It is less useful if the exercise simply repeats an already convincing demonstration.
Leave with: an appropriately bounded commitment, its resource implications and the evidence required before the next decision. Avoid describing an exploratory activity as an approved company-wide programme.
4. Who can resolve conflicts between functions?
Naming a project owner does not automatically settle competing incentives. Sales may want speed; operations may carry the cost of correction; technology may control access to essential systems.
Ask who can resolve those conflicts and which issues require executive intervention. Define the relationship between the business owner, technical contributors and people responsible for assurance. The answer should reflect existing organisational authority rather than add an impressive title with no practical mandate.
Also decide how concerns reach that person. A frontline employee who notices repeated errors needs a usable route, not a distant committee name.
Leave with: clear authority for the next phase, an escalation route and an account of unresolved cross-functional dependencies. This discussion concerns how decisions get made; detailed technical design belongs with the people responsible for delivering it.
5. What must people become better able to do?
An AI strategy is incomplete if its workforce assumption is “people will adapt.” Ask which capabilities the proposed change depends on, who currently has them and what support would be required.
For the distributor, reviewing quotations may demand product knowledge, judgment about unusual orders and the confidence to challenge a plausible recommendation. Tool familiarity alone would not settle those questions.
The human learning advantage offers a useful lens: consider what people need to understand and learn as the work changes, alongside what the system can produce.
Discuss time and workload honestly. If managers are expected to coach unfamiliar practices, what can they deprioritise? If roles may change, which decisions are settled and which remain open?
Leave with: the capability requirements and the support leaders are prepared to resource. Avoid promising job security or role changes that the organisation has not actually decided.
6. Which boundaries must hold while we learn?
Some constraints should shape the experiment from the outset. Ask which data, actions and customer interactions are appropriate for the proposed phase, what requires review and when activity should pause.
The NIST AI Risk Management Framework provides voluntary guidance for incorporating trustworthiness into AI design, use and evaluation. Its 2023 framework is under revision, according to NIST's current overview. It is a reference for discussion, not a legal compliance certificate.
Use relevant specialists to translate organisational requirements into practical boundaries. In the distributor example, permission to create an internal draft need not include permission to alter prices or make customer commitments.
Leave with: an understandable scope of permitted activity, the approval still needed and responsibility for checking that the boundaries remain appropriate. This guide cannot determine the legal or sector-specific requirements for your organisation.
7. What would make us continue, change direction or stop?
Before enthusiasm attaches itself to a project, ask what an acceptable result would look like and how it would be assessed.
For quotation work, faster drafting is one observation. Customer response, total turnaround, correction work and the burden on reviewers may change the overall interpretation. Agree which outcomes matter to the chosen ambition, what comparison is feasible and which costs must be included.
Do not force false precision where a baseline is missing. Assign someone to establish it. Equally, avoid moving the goalposts whenever results disappoint.
Leave with: the next review date, the evidence to be brought and the decisions that review can trigger. Include stopping or changing direction as credible outcomes. A review that can only endorse expansion is not a serious learning opportunity.
A focused half-day agenda
The following sequence is illustrative. It assumes that participants have read the evidence pack and can make the proposed decisions; it is not an empirically established optimal duration.
09:00–09:20 — Sponsor's brief: confirm the decisions in scope, authority and competing business priorities.
09:20–09:50 — Shared perspective: a keynote or expert briefing challenges a central assumption and clarifies the discussion.
09:50–10:20 — Reality check: operational evidence, employee perspectives and questions about what remains unknown.
10:35–11:35 — Choices and commitments: work through priorities, assumptions, resources and cross-functional authority.
11:35–12:15 — Conditions for proceeding: address capability needs, boundaries and evidence for the next review.
12:15–12:45 — Decision readback: confirm agreements, record disagreement and assign missing work.
Seven questions need not become seven equal sessions. Spend time where the choices are contested. If the evidence pack reveals substantial gaps in basic understanding, add a separate preparation session rather than compress learning and investment decisions into the same conversation.
Brief the external speaker for a precise contribution
A keynote can introduce a challenging perspective and a common vocabulary. Internal leaders remain responsible for the company's decisions.
Ask the speaker what assumption their contribution will help the room examine, what evidence supports the argument and which company context they need beforehand. Share the decision brief without disclosing material that should remain confidential. Agree how the moderator will connect the talk to the subsequent discussion.
For a detailed procurement checklist, see what HR and L&D buyers should ask a corporate learning keynote speaker.
Choose a keynote when the main need is shared perspective. Choose a facilitated working session when leaders need to examine competing options. A technical workshop serves another purpose: developing the detail needed to deliver a chosen direction. One programme can combine them, provided each contribution has a clear job.
Make the follow-through visible
Circulate a decision record promptly enough for participants to correct misunderstandings. Record what was agreed, who has authority to act, resources committed, unresolved questions and the next review. Give a deferred decision an owner and a condition for returning to it.
Communicate the relevant implications to affected teams. They need to know what has changed, what remains uncertain and how they can contribute evidence or raise concerns.
Evaluate the summit first against its actual purpose: were important choices clarified, was disagreement recorded honestly, and did authorised follow-through begin? Later business outcomes require their own evaluation. Attendance and executive satisfaction cannot establish that the summit caused a financial return.
The strongest closing question is practical: Which decision will be different because we spent this time together—and who now needs to hear about it?
Planning an executive AI event? Explore Learning Coach's international keynotes and discuss your executive event brief.
References
Wharton Human-AI Research and GBK Collective. The Adaptive Advantage: Who's Winning With AI and How, October 2026. See methodology and organisational alignment sections. US enterprise survey; observational self-reports, not an event evaluation.
Vereckey, Betsy. 6 questions to guide your AI strategy. MIT Sloan, 3 August 2026. Institutional reporting of George Westerman's expert perspective.
National Institute of Standards and Technology. AI Risk Management Framework. Framework released January 2023; current overview checked 28 October 2026. Voluntary guidance; revision underway.
