Prompt engineering for leadership
The phrase "prompt engineering" sounds wrong to the executive team. It sounds like a technical discipline, and the executive job is not to learn technique. But there is a real communication discipline buried inside the phrase, and once it is learned properly, AI moves from being a quick writing tool to being a thinking partner at the strategy level. This pillar walks through how leaders use AI for strategy deliberation, board preparation, internal communication and business analysis.
Written by Jesper Sachmann, founder of EnterpriseIQ. 27 years of IT leadership across Oracle, Logica and Capgemini, combined with daily use of AI as a thinking partner in strategic work since 2023.
- →Prompt engineering for leadership is not technique. It is a communication discipline similar to briefing a skilled analyst.
- →Four categories: strategy, board preparation, internal communication, business analysis.
- →A good executive prompt has four elements: role, context, task structure, output requirements.
- →Productivity gain typically 5 to 10 hours per week once stable workflows are established.
- →AI is a drafting tool and a thinking partner, not a decision-maker. Three areas should be avoided entirely: individual employee decisions, financial reporting, legal advice.
Why "prompt engineering" sounds wrong
When executives hear "prompt engineering", most think technical specialist. That is understandable, because the phrase comes out of the developer community and is associated with coding and system prompts. To leaders it sounds like something the IT director should familiarise themselves with, not something that belongs at the executive level.
That is the wrong frame. The technique that actually lives inside the phrase is not coding. It is the ability to articulate a complex task clearly enough that a skilled analyst can solve it. When you brief a consultant on a strategy project, you do exactly the same thing: you set the context, define the task, specify the output you want, and bound what falls outside scope.
The difference is that AI is faster and cheaper than the consultant, and it does not forget what you said in the previous session. On top of that, there is not a single executive task without one or more AI-suited moments. Strategy deliberation has them. Board preparation has them. Internal communication has them. Business analysis has them. That is what this pillar gets into.
Four categories of executive prompts
For executive teams in knowledge-intensive organisations, value-creating AI use falls into four main categories. Each has its own pattern and its own discipline.
Category 1: Strategy deliberation
AI as a thinking partner on decisions with uncertainty. Typical use cases: scenario analysis on a strategic bet, pros and cons of an acquisition opportunity, evaluation of a competitor's move, judgement on whether to close a business unit.
The discipline here is to give AI enough context about the company, the market and the decision frame, and then ask for scenarios rather than direct recommendations. The best-in-class models (Claude Opus 4.7, GPT-5 o1-pro) are strong on this task type when they receive proper context to work with. The output is not the decision. It is a better-structured discussion the executive team can have.
Category 2: Board preparation
AI as a preparation assistant ahead of board meetings. Typical use cases: drafting the quarterly report, Q&A prep on hard questions, condensing an extensive report into a 2-page briefing, talking points for a strategic presentation.
The discipline is to give AI the historical material (previous board minutes, previous reports, the company's strategic context) and ask it to produce something that matches the existing tone and structure. That cuts preparation time from typically 4 to 8 hours down to 1 to 2 hours per meeting, while the quality of the preparation usually improves.
Category 3: Internal communication
AI as a writing assistant on announcements, emails and talking points for employees. Typical use cases: announcing strategic changes, talking points for the quarterly all-hands, email responses to employee questions, communication around hard decisions.
The discipline is to give AI the context of the situation (what has been decided, why, what the consequences are) plus the tone-of-voice requirements (business-like, empathetic, decisive). AI produces a draft that typically needs 30 to 50 percent editing, but it qualifies the starting point significantly compared with starting from a blank page. This is the category where leaders most quickly see time savings.
Category 4: Business analysis
AI as an analyst on financial data, competitor research and market insight. Typical use cases: walking through quarterly results and identifying notable movements, comparing competitors' public statements, summarising industry reports, identifying patterns across customer data.
The discipline is to give AI structured data (statements, reports) and ask for both factual summary and interpretation. It is important to distinguish between AI "interpreting data you give it" (good) and "finding data itself" (quality risk, requires verification). Perplexity Pro is best for the research phase with source citations, Claude or GPT-5 for the deeper analysis on data you supply yourself.
Anatomy of a good executive prompt
A good executive prompt has four elements. It is not a strict template that must always be followed. It is a checklist that helps ensure the prompt has what it needs.
1. Role
Tell AI what kind of expert it should be. "You are a senior strategy consultant" or "You are an experienced CFO" or "You are a communications adviser with 20 years of experience from listed companies".
Why: AI adapts tone, vocabulary and analytical framework to the role you give it. A "senior strategy consultant" produces something more abstract than a "junior analyst", even on the same question.
2. Context
Tell AI about the company, the industry and the specific situation. Include relevant numbers (revenue, growth, market position), recent events, and the type of decision that needs to be made.
Why: Generic prompts produce generic answers. Concrete prompts produce concrete answers. 80 percent of the quality difference lives here, not in polishing the phrasing.
3. Task structure
Specify what you want AI to do. "Identify three scenarios" or "Summarise the five most important points" or "Write a 200-word draft". Clear structure requirements give clear structured answers.
Why: AI wants to be helpful and will produce something on its own if you do not specify structure. That "something" is rarely what you actually needed.
4. Output requirements
Specify tone, format, length and what should be omitted. "Business-like tone, no bullet points, max 400 words, avoid filler words like 'exciting' and 'fantastic'."
Why: AI's default style is often too flowing, too packed with signalling words and too bullet-heavy. Explicit format instructions remove 80 percent of the friction moments.
The four elements are not necessarily separate sections in your prompt. They can flow together. What matters is that all four are represented. Missing context, you get generic answers. Missing structure, you get unstructured answers. Missing output requirements, you get AI's default style. Missing role, you get something at a middle-of-the-road level.
Five example prompts that work
The examples below are deliberately short. They should be adapted to your situation, but the structure is consistent across them.
1. Scenario analysis on a strategic decision
You are a senior strategy consultant with experience from mid-market knowledge-intensive firms in Scandinavia.
We are a law firm in Aarhus with 45 staff, revenue DKK 65 million, growth 8 percent. We are considering opening an office in Copenhagen next year. The argument in favour is access to larger clients. The argument against is that local competitors are strongly entrenched and standing it up will cost DKK 4 to 6 million.
Identify three realistic scenarios (positive, neutral, negative) over a 3-year horizon. For each scenario: which 3 to 4 assumptions underpin it, what are the signal indicators that it is unfolding, what is the economic consequence.
Format: business-like tone, no filler, each scenario at most 200 words. Closes with one question I should ask before deciding.
2. Board Q&A prep
You are an experienced board chair who has sat on 15+ boards in the SME segment.
I am presenting Q1 results tomorrow. Key points: revenue 12 percent below budget, costs on plan, therefore lower earnings. We have lost two larger clients to competitors on price, but signed three new smaller clients.
Predict the seven to eight most likely questions the board will ask. For each question, articulate the sharp question behind the question (what they are really worried about), and suggest how I respond honestly without sounding defensive.
Format: no small talk, no "That's a great question" types. Direct and useful.
3. Announcing a hard decision
You are a communications adviser with experience from organisations facing hard employee announcements.
We have decided to move our customer service team from Aarhus to Aalborg. That means 12 employees need to relocate or change jobs. The decision is made, it is not up for debate, but we want to handle it respectfully and with proper support.
Draft an email to the affected employees (about 350 words). Tone: respectful, honest that this is hard, clear about what the next steps are, not self-pitying.
Requirements: no filler, no "We understand this may be difficult" formulations that ring hollow. Concrete next steps with timing.
4. Competitor analysis from public sources
You are a researcher who specialises in competitor analysis for mid-market firms.
Our main competitor is Firm X (a law firm in Aarhus, about 35 staff). Over the past 6 months they have hired two tech-law specialists and launched a new IP portfolio service.
Summarise, based on publicly available information: what their moves signal about strategy, which market segment they are pursuing, what the implications are for our positioning. Source citations for every claim.
Tooling: use Perplexity for the research phase, cross-check against LinkedIn and the firm's website.
5. Summary of a long report
You are a strategic analyst who produces executive briefings for busy directors.
I am attaching an 80-page industry report from McKinsey on AI adoption in Nordic knowledge-intensive firms.
Produce a 2-page briefing that: 1) summarises the 5 most important findings in priority order, 2) identifies the 3 most relevant implications for a mid-market law firm, 3) notes 2 things in the report that look overrated or too generic.
Format: no marketing tone, evaluative and concrete. Cite with page numbers when you refer to the report.
The five prompts show the pattern. All of them have role, context, task structure and output requirements. None of them take more than half a minute to write. The output is typically at a level where you as a leader spend 10 to 20 minutes polishing rather than writing from scratch.
Where executive prompts typically fail
Five pitfalls recur in the leaders we train. They are not hard to avoid once you know them.
Pitfall 1: Too little context
"What should we do about competitor X?" produces a generic answer. "We are a 45-person law firm with DKK 65 million revenue, competitor X has just expanded into an IP portfolio service, our existing clients are beginning to ask whether we can offer the equivalent. What are our three options?" produces a useful answer. The difference is not in the AI. It is in the context.
Pitfall 2: Treating AI as an oracle
Leaders who ask AI "what should we do" expect a decision. AI gives a plausible direction that sounds sensible, the leadership team follows it without further deliberation, and when it does not work out, they lose trust in AI. Better approach: use AI to structure the alternatives and their assumptions. The decision is yours.
Pitfall 3: No iteration
Many leaders ask one prompt and accept the first answer. The real value emerges in the second, third and fourth iteration, where you add "expand scenario 2", "challenge assumption X", "what would the counter-argument be". It is like a conversation with an analyst, not a query against a database.
Pitfall 4: Consumer tier for confidential data
Consumer-tier ChatGPT, Gemini and Perplexity typically use inputs for training. Client-confidential strategy deliberations, M&A considerations or executive-team discussions must not enter consumer tier. Use Claude Enterprise, Microsoft Copilot for M365 Enterprise, or self-hosted Llama for that kind of work. Details in our governance pillar.
Pitfall 5: Expecting instant magic
Leaders who try AI once, get a mediocre result and conclude "AI does not work at my level" often miss the real learning curve. It takes 4 to 8 weeks of serious use before prompts begin to flow naturally and the optimal split of work between AI and your own time becomes clear. It is a competence on par with Excel or PowerPoint, not a quick fix.
Three areas where AI should be avoided
This pillar would be incomplete without explicitly naming three areas where leaders should not use AI as a thinking partner, however good the prompt.
1. Decisions about individual employees
Hiring, dismissal, performance reviews, pay adjustments, promotions. This is Article 22 GDPR territory ("automated processing producing legal effect"), and it is high-risk under the EU AI Act. AI can be used to structure the process or prepare conversations, but the decision about individual people must be human and documented. Use AI to prepare 1-on-1 conversations, not to make HR decisions.
2. Financial reporting to auditors or stock exchanges
Quarterly results, annual reports, stock-exchange announcements, regulatory reporting. The quality standard is higher than AI can stand behind without thorough human review, and errors here have significant consequences. AI can be used to draft from structured data you provide yourself, but every single number must be traceable to its source, and review must be documented.
3. Client-specific professional advice with legal effect
Contracts, audit findings, audit opinions, financial advice, legal counsel. AI is a drafting tool here, not a decision-maker. The professional standardisation embedded in the authorisations (auditor, attorney, financial adviser) is part of the quality the client is paying for. AI accelerates the drafting work, but the final delivery is the licensed person's responsibility.
Three steps you can take this week
Step 1: Establish your AI workflow as a leadership team
- Each executive sets aside 30 minutes to test one of the five example prompts above against a real task
- Use Claude Pro or ChatGPT Plus (not consumer tier if the data is confidential)
- Note what worked, where friction appeared, and where you spent the most time
Step 2: Hold a 60-minute reflection together
- The executive team meets to discuss the results from Step 1
- Identify 2 to 3 workflows where AI augmentation looks value-creating
- Agree who owns the further development of each workflow
- Set a follow-up meeting 4 weeks out where you review progress
Step 3: Define AI boundaries explicitly
- Write a 1-page document that specifies which data may be sent to which tools
- Explicitly confirm that the three forbidden areas (individual employee decisions, financial reporting, legal advice) are not AI-processed
- Share the document with the rest of the leadership team and the IT director
- Place it inside your AI policy (if you do not have one yet, see our governance pillar)
The three steps are deliberately light. The goal is not perfect strategy on day one. It is getting the leadership team going so the foundation can grow on the basis of real experience rather than hypotheses.
FAQ
Should the executive team learn prompt engineering?
Not as a technical discipline. But the executive team should learn to use AI as a thinking partner, and that requires a basic grasp of context-building. 2 to 4 hours of focused training is typically enough.
How much time does a typical executive spend on prompts?
30 to 60 minutes per session in serious strategy work with AI as a thinking partner. Quick prompts 2 to 5 minutes. Combined productivity gain typically 5 to 10 hours per week once stable workflows are established.
May the executive team send client data to AI?
It depends on the tier and provider. Claude Enterprise, Microsoft Copilot for M365 Enterprise and Gemini Workspace Business have DPAs and can receive confidential data. Consumer tier must not. For highly sensitive data: self-hosted Llama or Mistral.
How do you avoid AI hallucination?
Three techniques: keep the context verifiable, explicitly ask AI to flag uncertainty, use Perplexity for research with source citations. Hallucination most often occurs when AI is treated as an oracle rather than a thinking partner.
What should the executive team NOT use AI for?
Decisions about individual employees, financial reporting to auditors or stock exchanges, and client-specific professional advice with legal effect. AI can support preparation, but it cannot make decisions or carry the ultimate responsibility.
Does Jesper use AI as a thinking partner himself?
Yes, daily. Claude Opus 4.7 for strategy, ChatGPT Plus for iteration, Perplexity for research. Every deliverable (including this pillar) carries an audit trail. See our ai-stack page for full transparency.
Next steps
Three paths depending on where you stand:
Take the EnterpriseIQ Score
12 questions, 5 minutes. Baseline on AI maturity for the executive team plus five concrete quick wins.
AI Tools Bootcamp
One-day hands-on for the leadership group. Executive-focused version: 90-minute briefing plus 30 prompts.
30-minute conversation
No obligation. We work out which executive AI workflows make sense for your leadership team.
About the author
Jesper Sachmann is the founder of EnterpriseIQ. 27 years of IT leadership across Oracle, Logica and Capgemini plus 11 years of Archer experience as Alliance Director Europe and Integrated Risk Management Lead Nordics, combined with daily use of AI as a thinking partner since 2023.
AI attribution: This article is AI-assisted produced with Claude Opus 4.7, human review by Jesper Sachmann. See our AI transparency policy for how we use AI in every deliverable.
Citing this article? "EnterpriseIQ: Prompt engineering for leadership (2026-05-26)" or link to enterpriseiq.dk/en/insights/prompt-engineering-for-leadership.