Catalogue › AI Strategy › ai-vision

AI vision

Articulates the organisation's AI ambition - a deliberately chosen level of ambition, a short vision statement staff can repeat, a narrative of how work will change and a handful of guiding principles - agreed by the leadership team and written for the whole workforce. Use when leaders cannot say in a sentence why the organisation is adopting AI, when functions hold competing pictures of it, or before outcomes and a roadmap are set.

Department

AI Strategy

Lifecycle

Discover, Envision

Maturity

explorer, adopter, builder, scaler, ai-enterprise

Risk tiers

0, 1, 2, 3, 4

Platforms

platform-neutral

Editions

starter, scale, enterprise

Reviewers

none

Status

draft

Copilot Cowork

AI Strategy plugin for Copilot Cowork, or this skill on its own: Download SKILL.md

Licence

CC BY 4.0, by Simon Foote

Purpose

This skill gives the organisation one shared answer to "why are we doing AI, how far do we intend to go, and what kind of organisation do we want to be while we do it". It produces an explicit level of ambition, a short vision statement, a narrative of how work will change for people, and guiding principles that set the tone for everything after. Vision work usually goes wrong in one of three ways: it is a product slogan ("Copilot for everyone"), it is so grand that nobody can connect it to their own job, or it is written by a small group and announced, so the workforce reads it as a headcount plan in disguise.

When to use

• The leadership team has decided to take AI seriously but cannot say in one sentence why, or how far.
• Different functions describe AI in contradictory ways (cost cutting, innovation, risk) and staff are hearing all of them.
• A strategy, CoE or adoption programme is about to start and needs a shared starting point.
• Staff anxiety about AI is rising and leaders need to state their intent plainly.
• A merger, new chief executive or major change of direction makes the existing vision obsolete.

When not to use

• Turning the vision into three to five measurable outcomes, decision-rule principles and the "where AI is not appropriate" list - use ai-strategy:strategic-outcomes.
• Establishing the honest current baseline the ambition must start from - use ai-strategy:ai-maturity-assessment.
• Securing and equipping the executive who will champion the vision - use ai-strategy:executive-sponsorship.
• Planning how the vision is communicated across the organisation - use change-and-communications:ai-communication-strategy.
• Revisiting an existing vision against results and Microsoft change - use ai-strategy:strategy-refresh.

Questions to ask first

1. What is the organisation for, and which of its existing strategic priorities could AI genuinely help with?
2. How ambitious do leaders intend to be over the next two to three years, and what are they prepared to invest and change to get there?
3. What do leaders want staff to believe about AI and their jobs, and is that belief true?
4. Which commitments (to customers, regulators, staff representatives, the public) must the vision respect?
5. Who must agree the vision for it to hold, and who will be expected to repeat it?

Required inputs

• The organisation's mission, current business strategy and its existing values or behaviours framework.
• The maturity assessment from ai-strategy:ai-maturity-assessment, or a note that none exists yet.
• Time with the leadership team, ideally a facilitated working session, plus short conversations with staff representatives where they exist.
• Any existing AI statements, policies, public commitments or regulatory obligations.

Where the maturity baseline is missing, mark the ambition level as provisional and record the gap. Never invent competitor positions, market statistics or productivity figures to make the case for a bolder ambition.

Decision framework

Choose the ambition level deliberately, using the maturity baseline and the investment leaders will actually commit. The level shapes what the vision may promise.

A vision statement passes when it fits in two sentences, names a change in how work or service happens rather than a product, can be repeated by a manager without notes, and is consistent with the chosen ambition level. Guiding principles pass when there are no more than six, each states a stance the organisation will hold under pressure, and each could be explained to a sceptical member of staff.

Practitioner approach

Start from the organisation's purpose and current strategy, not from AI. Ask leaders what the organisation exists to do and where it is under strain, then ask where AI could plausibly help. A vision that cannot be traced to the existing strategy will be treated as a technology side project and abandoned at the first budget round.

Make the ambition choice explicit and early. Present the three levels, the maturity evidence and what each level would demand in investment, ownership and process change. Leaders often say "transformed" and fund "foundational"; naming the gap in the room is the most useful thing this skill does. Record the chosen level and the reasoning, and say what would have to be true to move up a level later.

Draft the vision statement with the leadership team, not for them. Offer two or three contrasting drafts and let them argue. Good statements describe people and service: "Our people spend their time on judgement and care, with AI handling the searching, drafting and summarising, so customers get faster, better-informed answers." Weak ones describe tools or adjectives. Test each draft by asking a manager to explain it to their team in their own words; if they cannot, it is too abstract.

Write the narrative next: half a page on how a typical working day, a customer interaction and a manager's role will feel different in two to three years. This is where most staff will look for themselves, so be honest about roles that will change. If leaders intend AI to reduce roles, the vision must not imply otherwise; employment questions go to HR and, where relevant, staff representatives before anything is published.

Then agree the guiding principles. These are stances rather than rules: how the organisation treats its people, how openly it uses AI with customers, who stays accountable, how it treats data, and its attitude to experimentation and risk. Keep them few and plain. ai-strategy:strategic-outcomes will later turn them into decision rules and add the explicit exclusions, so do not try to write policy here. Check coverage against the responsible AI and risk management source records, summarised in your own words.

Play the result back on one page using the ai-vision-and-principles template and test it with a small, mixed group of staff before publication. Watch for words that alarm people ("efficiency", "headcount", "replace") and for promises the organisation cannot keep. Record what changed after testing. Finally, agree who owns the vision (normally the executive sponsor), where it will be published and when it will be revisited.

Outputs

• An AI vision on the ai-vision-and-principles template: purpose link, chosen ambition level with reasoning, vision statement, narrative of change, guiding principles, commitments to staff, owner and review date.
• A record of the staff test: who was consulted (by role or group), what they said and what changed.
• A short list of open questions for ai-strategy:strategic-outcomes and ai-strategy:ai-roadmap to resolve.

Microsoft platform considerations

Not applicable to the vision itself, because the ambition must survive changes in Microsoft's products and names. Keep product names out of the statement and principles. Where leaders want to illustrate the narrative with today's tools, check current capability and naming against the source records in references/sources.md and label the examples as illustrative.

Adoption considerations

The vision is the first thing most staff will hear about AI, and it sets their expectations of everything that follows. Hand it to change-and-communications:ai-communication-strategy for a deliberate rollout through managers, and make sure managers hear it, with the reasoning, before their teams do.

Knowledge considerations

Not applicable in detail at vision stage. Where the ambition depends on AI grounded in organisational knowledge, note that dependency so the outcomes and roadmap address knowledge ownership early.

Security and privacy considerations

A guiding principle on data should state that AI use respects existing confidentiality and data protection obligations. Check the wording with the data protection officer and the security lead; legal interpretation belongs with them, not with this skill.

Responsible AI considerations

The guiding principles are the organisation's first public statement of how it intends to use AI responsibly. Check that they address accountability, transparency, fairness and human oversight, and that nothing in the vision contradicts them. Where the narrative implies AI informing decisions about people, flag it for responsible-ai:ai-impact-assessment before publication.

Evaluation requirements

Not applicable to solution testing. The vision is evaluated by whether a mixed group of staff can restate it accurately after reading it once, and whether leaders can use it to settle a real disagreement about direction.

Value measures

The vision does not carry benefit figures. Its effect is measured through staff understanding and sentiment about AI (through existing surveys, reported at group level above an agreed minimum size), with the executive sponsor as owner, and through the strategic outcomes that follow from it.

Scale notes

Starter

One leadership session and a one-page vision with three or four principles. The chief executive or a director owns it, and it is tested informally with a handful of staff.

Scale

Run the leadership session, then test drafts with staff from each major function and with staff representatives. Publish with a short narrative per function showing what it means locally.

Enterprise

Set the enterprise vision and principles centrally; business units write their own narrative beneath it without changing the statement or principles. The board approves the ambition level, and the vision is reviewed with the annual strategy cycle.

Hand-offs

• To ai-strategy:strategic-outcomes to turn the vision into measurable outcomes, decision-rule principles and exclusions.
• To ai-strategy:executive-sponsorship to confirm the sponsor who will own and champion the vision.
• To change-and-communications:ai-communication-strategy to plan how it reaches the workforce.
• To ai-strategy:investment-case so the funding request matches the chosen ambition level.

Reviewer gates

Not applicable. The vision is an executive decision, not a release. The responsible-ai department may be asked for advice on the principles; the data protection officer and HR review anything touching personal data or roles.

Failure modes

• The vision names a product ("Copilot everywhere"), so it dates within a year and says nothing about outcomes.
• Leaders choose a transformed ambition but fund a foundational one, and the gap surfaces as missed expectations.
• The narrative avoids the question of jobs, staff fill the silence with the worst interpretation, and adoption stalls.
• Principles are so generic ("we use AI responsibly") that they never shape a decision or reassure anyone.
• The vision is announced without manager briefing, and managers cannot answer their teams' first questions.

Sources

references/sources.md in this skill lists the source records behind this skill, what each is authoritative for, how volatile it is and what you may do with it. Product availability, licensing, limits and preview status change often, so check the current source before stating them, and cite what you used.

Never

• Write a vision statement that names a product or vendor feature instead of a change in work or service.
• Invent market, competitor or productivity statistics to justify an ambition level.
• Imply job security, or job reductions, that leaders have not decided and HR has not reviewed.
• Publish the vision before managers have been briefed and a group of staff has tested it.