STRATEGY OPERATIONS · OCTOBER 2026
The strategy memo should make the uncertainty easier to inspect.
Use AI to reduce information assembly and challenge assumptions, while leaders own the decision. Build a workflow around source quality, scenarios, approval and execution instead of treating a particular model or tool as a substitute for strategic judgment.
Start with a decision and an evidence boundary
Write the decision in one sentence: enter a market, change a price, acquire a business or reallocate capital. Identify who can approve it, when the decision is needed and which outcomes matter. A request to summarize everything about an industry is an information task, not a strategy process.
Separate known facts, third-party claims, internal assumptions and open questions. For each material fact, retain the source, date, scope and owner. A competitor’s announcement may prove that it announced a product; it does not prove product traction. An analyst forecast is a scenario, not observed demand.
Give the assistant an explicit evidence boundary. It may summarize authorized documents and prepare questions. It must flag missing evidence rather than fill it with plausible detail. For a deal or legal interpretation, involve the qualified reviewer who owns that judgment.
Organize the stack into four useful layers
The intelligence layer gathers current primary sources and authorized internal records. The compression layer extracts findings, disagreements and implications. The scenario layer checks assumptions and consequences. The execution layer turns an approved decision into accountable milestones and review triggers.
Current model catalogs help select candidates for document analysis, tool use and multimodal inputs. OpenAI, Anthropic and Google document different capabilities and availability status. Use those catalogs as procurement references, then test the actual document pack; a longer advertised context window does not prove complete or accurate analysis. OpenAI catalog; Claude catalog; Gemini catalog
McKinsey’s analysis of agentic foundations emphasizes governed data, and its technology agenda discusses operating-model change. These sources support considering data and execution together. They do not establish a universal return on a particular strategy assistant or justify replacing a leadership team with a model. Governed data foundations; Technology agenda
Use a memo template that exposes assumptions
Ask for a decision statement, options, evidence ledger, financial implications, operational requirements, risks, unresolved questions and recommendation conditions. Require source references for factual claims and labels for assumptions. The recommendation should identify what would change the decision rather than merely argue for the preferred option.
For a long document pack, create a coverage register. List each source, which sections were considered and the claims extracted. Verify important provisions directly in the original documents. Do not assume that uploading every file means the model found every material clause or reconciled contradictions.
Have the assistant prepare a challenge memo from a different viewpoint: customer demand, operational feasibility or financing risk. Keep that analysis grounded in the same evidence. A theatrical debate between agents is less useful than a concrete list of unsupported assumptions, missing permissions and resources required to execute.
Keep calculations outside persuasive prose
Use controlled spreadsheets or planning software for scenario calculations. Define the units, period, currencies and formulas. The model can propose variables and explain sensitivity, but verify the calculations before using them in a board paper.
Build a base case, a downside case and a condition that would justify expansion. Separate probability estimates from facts. If the team cannot support a probability, use a break-even threshold or range instead. Ask which operational constraint dominates each scenario and which evidence would reduce the uncertainty.
Make the implementation requirements visible: accountable owners, staffing, budget, data access, decision authority and dependencies. A strategy can be attractive and still infeasible with available resources. The memo should show that gap explicitly so leadership can fund it, change the sequence or reject the option.
Connect approval to an operating rhythm
Once approved, record the decision, assumptions and review date in the system of record. Translate commitments into owned milestones and acceptance conditions. Keep a change log so new evidence can update the plan without rewriting the history of what leadership originally approved.
Review leading signals that are specific to the decision. A new market entry may require signed distribution capacity, qualified demand and service capability; generated content volume is not a substitute. Define stop, continue and expand conditions before sunk-cost pressure appears.
Restrict access to deal, personnel and competitive information. Confirm retention and vendor terms for the workspace you use. AI assistance should make the evidence more inspectable while preserving confidentiality and authority. It should never quietly broaden who can see a sensitive strategy pack.
Compare the choices
| Work | Tool capability to test | Failure risk |
|---|---|---|
| Internal evidence synthesis | Permission-aware retrieval and source references | Incomplete coverage or inaccessible source claims. |
| Current external research | Primary-source discovery and dated evidence | Stale pages and unsupported secondary summaries. |
| Financial scenarios | Controlled formulas and versioned assumptions | Confident prose around incorrect arithmetic. |
| Execution follow-through | Owned milestones and decision history | An approved memo without resources or accountable delivery. |
Worked example: make the decision concrete
Illustrative market-entry decision: a leadership team assumes 200 customers, $1,000 annual contribution per customer and $150,000 annual fixed cost. The base-case contribution after those fixed costs is $50,000. The break-even customer count is 150, ignoring setup cost, financing and tax. These are illustrative inputs, not market evidence.
The assistant should identify the assumptions that need validation: customer willingness to buy, acquisition cost, support capacity and collection timing. The team can approve a limited demand test with a capped budget instead of committing to the full expansion. A useful memo states the evidence needed to move beyond that test and the owner responsible for obtaining it.
Failure modes to plan for
The most common failure is compressing away uncertainty. A sourced caveat becomes a confident headline, or a projected number becomes an observed result. Compare the draft’s certainty with its evidence ledger.
Another failure is letting a polished recommendation outrun delivery capacity. Require staffing, authority and dependencies alongside the financial case. Keep historical incident stories and survey claims out of the memo unless the actual records support them; this guide does not rely on an invented acquisition failure or an undisclosed executive survey.
Your useful next step
Take one pending leadership decision and create an evidence ledger, option table and break-even analysis. Ask AI to identify missing assumptions, then have the responsible owners validate the financial and operational requirements before approval.
Questions and answers
Which model is best for strategy?
Evaluate current candidates on your real document pack and decision workflow. Accurate source coverage, confidentiality and useful uncertainty handling matter more than a general model ranking.
Can AI perform legal or acquisition diligence?
It can organize documents and identify questions. Qualified legal, financial and operational reviewers must validate material conclusions and own the diligence decision.
Should we assign probabilities to every scenario?
Only when there is a defensible basis. Otherwise use ranges, break-even thresholds and explicit conditions rather than invented precision.
How do we know the assistant helped?
Measure evidence preparation and correction work, decision traceability and execution clarity. Do not attribute business results to AI without a credible baseline and causal comparison.
Sources and evidence scope
Primary sources checked on 11 October 2026. Product pages establish vendor-described features; they do not establish independent comparative performance. Worked examples are explicitly illustrative. Implementation recommendations are editorial guidance.
- Openai Models — Official catalog documents current model identifiers, capabilities and supported interfaces.
- Claude Models — Official catalog documents current Claude models, identifiers and capabilities.
- Gemini Models — Official catalog documents Gemini models, availability status and capabilities.
- Mckinsey Data — Primary consultancy analysis supports governed data transformations and agent scaling, not guaranteed ROI.
- Mckinsey Agenda — Primary 2026 strategy analysis supports operating-model transformation scope.
Research Path