Copilot for Strategic Decisions: 5 Workflows to Stress-Test Your Next Big Call
The short answer: yes, if you use it to attack your decision instead of approving it. Run a pre-mortem, a devil's advocate, scenarios, option scoring and a board-question rehearsal, and keep the final call. Below: 5 workflows, each with a full prompt and what to check.
This is the written version of the strategic-adviser part of my executive session. I teach executives to use AI as a strategic and managerial adviser, like a McKinsey consultant at your side. It is not an IT project. Whoever owns it decides.
Copilot for strategic decisions: 3 setup moves
1Write the decision in 1 paragraph
What you're deciding, by when, the options on the table, and what you already believe. You'll paste it into all 5 workflows. If you can't write it, that's your first finding.
2Build every prompt in 4 parts
Microsoft's framework is goal, context, expectations and source. Only the goal is required. Put sources last, because later parts of a prompt get more emphasis. And run the important ones twice, since the same prompt can give different answers.
3Pick the tool
Free Copilot Chat works on the web and on what you paste or upload. Researcher (generally available since 2 June 2025) builds a cited report from your files, emails, meetings and the web. It needs a paid Copilot licence (Microsoft 365 Copilot, now called Microsoft Copilot), and a query typically takes under 5 minutes if it's simple and 10 to 45 if it's complex. Microsoft's published limit is 25 queries per user per month (an October 2025 figure, and I couldn't confirm a newer one), so spend them on decisions that matter. Analyst works on files you attach, and Microsoft's 2025 launch post says you can view the Python it runs. More in Researcher vs Analyst.
1. Pre-mortem: tell Copilot the decision already failed
Use it before you sign. Imagine the decision failed 12 months from now, and ask why. (Yes, it feels morbid. That's the point.)
Run a pre-mortem on this decision. Decision: [paste your one-paragraph decision]. Imagine it is 12 months from now and the decision has clearly failed. Write the 7 most likely reasons, ranked by likelihood. For each one, give the first warning sign I would see in the first 90 days and 1 action I can take now to reduce the risk. Mark each reason as based on my facts or as your own assumption. Use the attached memo and the facts above.
2. Devil's advocate: make it argue your board's side
Your board will do this anyway, in public. Better to hear it first, in private. Give Copilot a role: the most sceptical director in the room.
Act as the most sceptical member of our board. Proposal: [paste your one-paragraph decision]. Give me the 5 strongest objections, each in one sentence, the way a director would say it out loud. For each one, tell me what evidence would prove it right and what would prove it wrong. Keep the tone blunt and hold back alternatives until I ask. Base it on the attached proposal and last quarter's results.
3. Scenarios: 3 futures, each with a trigger
A scenario is only useful if it has a trigger, the moment you change course. So ask for one in each.
Build 3 scenarios for [decision] over the next 18 months: base case, downside and upside. For each, give the 3 assumptions that define it, what I would see in the first 2 quarters if it is playing out, the decision I would take at that point, and what it would cost to reverse course. Put the assumptions in a table. Use the attached forecast and market notes, and label anything not from them as an assumption.
4. Option scoring: set the weights before you see the answer
Weights are where politics hides. Write them down first, then let Copilot build the table. The weights in this prompt are an example, so use yours.
Score these options for [decision]. Option A: [name and one line]. Option B: [name and one line]. Option C: [name and one line]. Criteria and weights, adding up to 100: revenue impact 30, risk 25, time to value 20, fit with strategy 15, effort 10. Give each option a score from 1 to 5 per criterion in a table with a one-line reason, then the weighted total. Then tell me the single weight change that would flip the ranking. Use the attached options paper.
5. Board-question rehearsal: the 10 questions you'd rather hear today
Run it the day before the meeting, then again after you've fixed the weak spots.
Act as a director who has read the attached board paper on [decision] and is not convinced. Ask me the 10 hardest questions, in the order I am most likely to get exposed: numbers first, then risk, then people, then timing. After the list, show how a weak answer and a strong answer would sound for the top 3 questions. Use the attached board paper and the latest results.
Here is my answer to question 3: [type or paste it]. Score it from 1 to 5 for clarity and evidence, and tell me what a sceptical director would still ask.
What Copilot should not decide
1The verdict
Copilot drafts, you decide. Microsoft tells you to review and verify its output, because mistakes are hard to spot.
2Decisions about people
If the decision names people (a restructuring or a promotion), keep Copilot out of the verdict. Microsoft's application card says to avoid uses with a consequential impact on employment. In the EU, AI systems used to decide on promotion or termination are on the AI Act's high-risk list (Annex III, point 4), with obligations from 2 December 2027. Ask a lawyer where your case falls.
For the weekly routine, see my Copilot prompts for executives. I have no commercial interest in Microsoft, for the record.
Questions about using Copilot for strategic decisions
Can Copilot help with strategic decisions?
Yes, as an adviser that challenges your thinking. The decision stays with you, because Microsoft tells you to verify its output.
What is a pre-mortem, and how do you run one with Copilot?
A pre-mortem assumes the decision has already failed and asks why. Give Copilot the decision, the context and 1 source. Ask for the 7 likeliest reasons, a first warning sign for each, and a mark on which are assumptions.
Should executives use Researcher or regular Copilot Chat for decisions?
Use regular Copilot Chat when you can paste the facts. Use Researcher when the evidence is spread across your files, emails, meetings and the web. It needs a paid Copilot licence, and Microsoft's published limit is 25 queries per user per month (an October 2025 figure).
Does Microsoft have Copilot scenarios for executives?
Yes. Microsoft's Scenario Library has an Executive Offices section with 6 scenarios, last modified in June 2025: preparing an all-hands meeting, and 5 days in the life of an executive. Microsoft marks its examples as for demonstration purposes only, so use them for ideas.
Is it safe to put board papers into Copilot?
Check your company's rules and your sharing settings first. Microsoft says Copilot only surfaces organisational data a person already has at least view permission for, and that prompts, responses and Microsoft Graph data aren't used to train its foundation models (Microsoft Learn, July 2026). More in my EU data residency guide.
Want your leadership team to practise this live?
This is part of my executive session. A big room gets the keynote. The leadership team alone gets the 90-minute executive briefing.