Prompt engineering is more than asking an AI system a good question. It translates business context, role, task, constraints, data boundaries and output standards into a clear instruction. In enterprise use, a strong prompt must design for accuracy, security, traceability and human oversight—not speed alone.
1. Six components of an effective prompt
Defines the professional perspective the model should adopt.
Explains the organisation, sector, audience and problem.
States one clear and measurable output.
Sets the data, text and assumptions to be analysed.
Defines the required format such as a table, memo or deck.
Adds evidence, uncertainty and human-approval rules.
2. prompt template
Act as [ROLE]. Context: [ORGANISATION, SECTOR, AUDIENCE AND PROBLEM] Task: [ONE CLEAR OUTPUT] Input: [DATA/TEXT] Constraints: [REGULATION, PRIVACY, TIME, SCOPE] Output format: [TABLE/MEMO/PRESENTATION] Quality criteria: [EVIDENCE, CONSISTENCY, RISK, UNCERTAINTY] Do not fabricate missing data. State assumptions and flag decisions that require human approval.
3. Practices that improve prompt quality
- Break complex work into stages instead of combining many ambiguous tasks.
- Do not upload confidential, personal or commercially sensitive data to unapproved systems.
- Use measurable quality criteria rather than abstract words such as “perfect”.
- Ask for evidence and uncertainty, and independently verify generated sources.
- Require human review for high-impact HR, legal, financial, HSE and investment decisions.
- Manage proven prompts with owner, version, purpose and usage notes.
4. 40 ready-to-use enterprise prompts
Replace the bracketed fields with your own context. Validate every output against policies, real data and authorised expert judgement.
Board Brief
Act as a senior management adviser. Review the report below and produce a one-page board decision brief with: critical developments, target deviations, root causes, top five risks, decisions required and 30-day actions. Link each finding to evidence. Do not invent missing data. Report: [PASTE TEXT]
Strategy Map
Act as a strategy adviser. Build a three-year strategy map for [ORGANISATION/SECTOR] using financial, stakeholder, internal process, and learning and growth perspectives. For each objective, show cause-and-effect links, KPI, target, owner and first 90-day initiative.
KPI Design
Act as a performance management specialist. Propose no more than 12 critical KPIs for [PROCESS/UNIT]. Include purpose, definition, formula, source, frequency, target, warning threshold, owner and potential gaming risk. Separate leading and lagging indicators.
OKR Quality Review
Review the OKRs below. Score strategic relevance, measurability, ownership and impact from 1 to 5. Identify key results written as activities and rewrite them as measurable outcomes. OKRs: [PASTE]
Risk Register
Act as an enterprise risk manager. Identify strategic, operational, financial, regulatory, HSE, environmental, cyber and reputational risks for [SECTOR/PROCESS]. For each risk include cause-event-impact statement, controls, likelihood, impact, control effectiveness, residual risk, KRI, action and owner.
Scenario Planning
Create base, upside and stress scenarios for [TOPIC]. For each, state assumptions, trigger indicators, financial and operational effects, decision points and pre-agreed actions. Present uncertainty explicitly and avoid false precision.
Audit Programme
Act as a risk-based internal auditor. Prepare an audit programme for [PROCESS], including objective, scope, criteria, key risks, control objectives, interview questions, sampling, tests, evidence and likely finding classes.
Nonconformity Analysis
Act as an ISO lead auditor. Rewrite the finding below as requirement, objective evidence, nonconformity statement and risk effect. Do not assume the root cause. Then create investigation questions for Five Whys and fishbone analysis. Finding: [PASTE]
Management Review Agenda
Design an integrated management review agenda covering ISO 9001, 14001, 45001, 50001 and 27001. For each item include input, decision question, expected output, owner and follow-up KPI.
SIPOC
Create a SIPOC for [PROCESS]. Show suppliers, inputs, high-level steps, outputs and customers. Then define boundaries, critical-to-quality characteristics, top three risks and five metrics.
RACI Matrix
Create a RACI matrix for [PROJECT/PROCESS]. Break work into decision-relevant activities and ensure one accountable role per row. Highlight missing or conflicting R and A assignments.
Delegation of Authority
Design a delegation of authority matrix for [ORGANISATION/UNIT] covering transaction type, threshold, preparer, reviewer, approver, informed parties, deputy rule, exceptions and records. Flag segregation-of-duties risks.
Organisation Design
Act as an organisation design adviser. Analyse role overlap, layers, spans, decision delay and centre-site tension in the structure below. Propose three models and compare benefits, risks and transition conditions. Structure: [PASTE]
Role Profile
Create an outcome-focused role profile for [POSITION], including purpose, accountabilities, decision rights, key interfaces, KPIs, technical and behavioural capabilities, and first 100-day expectations.
Capability Matrix
Build a capability matrix for [UNIT/ROLE GROUP]. Classify technical, managerial, digital and behavioural capabilities. Define observable behaviour at four levels and recommend target levels by role.
Learning Needs Analysis
Use the performance data and role requirements below to conduct a learning needs analysis. Separate training-solvable gaps from process, resource or management issues. For training needs, define target behaviour, audience, method, duration, assessment and business KPI. Data: [PASTE]
Leadership Feedback
Act as an executive coach. Thematically analyse the 360 feedback below; identify strengths, development areas and possible blind spots without diagnosing personality. Build a 90-day plan and weekly practice routine. Feedback: [PASTE]
Change Plan
Build a change management plan for [TRANSFORMATION]. Segment stakeholders by impact and resistance. For each group define required behaviour, likely resistance, message, channel, sponsor, feedback mechanism and adoption metric.
Meeting Design
Design a decision-focused [MEETING NAME] with purpose, participant roles, pre-read, agenda, decision question per item, timing, decision log, action follow-up and cancellation criteria.
Decision Memo
Prepare a decision memo of no more than two pages for [TOPIC] using problem, context, options, criteria, cost-benefit, risks, reversibility, recommendation and decision required.
Investment Case
Build an investment case for [INVESTMENT] covering strategic fit, problem, scope, alternatives, CAPEX/OPEX, benefits, data required for NPV/payback, risks, permits, resources, milestones and success measures. Do not fabricate financial data.
Specification Review
Act as a procurement and engineering specialist. Review the specification below for scope clarity, performance criteria, acceptance tests, standards, deliverables, warranty, maintenance, HSE, environment, cybersecurity and vendor neutrality. Flag ambiguity and brand bias. Text: [PASTE]
Bid Comparison
Compare the bids below using total cost of ownership, technical compliance, lead time, warranty, service, references, contract risk and financial resilience. Build a weighted matrix and leave missing data blank. Bids: [PASTE]
Supplier Scorecard
Create a balanced supplier scorecard for [SUPPLIER/CATEGORY] using quality, delivery, cost, HSE, environment, ethics, innovation, continuity and collaboration. Define measure, weight, source and improvement threshold.
Root Cause Analysis
Act as a problem-solving facilitator. Rewrite the event below as an observable, measurable problem statement. Create a timeline, change analysis, Five Whys, evidence questions across people-equipment-method-material-environment-management, and a verification plan. Event: [PASTE]
Maintenance Strategy
Recommend a maintenance strategy for [ASSET/SYSTEM]. Compare corrective, preventive, predictive and design-change options using failure mode, criticality, safety, production effect, detectability and cost. Justify with RCM logic.
Downtime Analysis
Classify the downtime records below into failure, planned maintenance, process, fuel/material, grid, environment, HSE, external and other. Produce a two-word cause, duration, lost output and recurrence risk for each; then create a Pareto summary and top five actions. Records: [PASTE]
Energy Performance
Act as an ISO 50001 energy specialist. For [FACILITY/PROCESS], identify significant energy uses, relevant variables, baseline, EnPI formulas, normalisation and saving opportunities, with a measurement and verification plan.
HSE Risk Assessment
Act as an HSE risk specialist. Identify hazards for [ACTIVITY] across routine, non-routine, maintenance, contractor, human factors and emergencies. Evaluate controls and recommend additional controls in the hierarchy: eliminate, substitute, engineer, administer, PPE.
Incident Learning
Analyse the incident or near miss below without blame. Identify immediate causes, contributing conditions, systemic causes, failed and successful barriers, high-potential indicators and organisation-wide learning actions. Event: [PASTE]
Environmental Aspects
Create an environmental aspect-impact register for [ACTIVITY/SITE] covering normal, abnormal and emergency conditions; air, water, soil, waste, resources, noise and biodiversity. Propose significance criteria and operational controls.
ESG Metrics
Design a board-level ESG metric set for [SECTOR]. For each metric define rationale, calculation, boundary, owner, frequency, target, assurance need and greenwashing risk.
Cyber Incident Playbook
Act as an information security manager. Build an incident response playbook for [INCIDENT TYPE], covering detection, classification, containment, evidence preservation, recovery, communications, regulatory notification, escalation and lessons learned with roles and time targets.
Data Governance
Design a governance model for [DATA DOMAIN], including data owner, steward, system owner, quality rules, master definitions, access, retention, classification, change control and quality KPIs.
AI Use Cases
Generate AI use cases for [FUNCTION]. Score each by business problem, user, data, model task, human control, benefit, error impact, privacy risk, feasibility and pilot duration. State controls before recommending automation of high-impact decisions.
AI Risk Assessment
Assess the AI application below for privacy, security, bias, accuracy, explainability, IP, regulation, human oversight, vendor dependency and continuity. Build a risk-control-evidence-residual risk-approval table. Application: [PASTE]
Corporate Communication
Act as a corporate communications specialist. Rewrite the technical text below for the board, employees and the public. Preserve the core message but adapt tone and detail. Avoid hype, unsupported claims and jargon. Text: [PASTE]
LinkedIn Post
Turn the article below into a professional LinkedIn post. Use a strong but non-sensational opening, one management problem, three practical insights and a reflective closing question. Write 900-1,300 characters and use no more than five relevant hashtags. Article: [PASTE]
Presentation Storyline
Act as a management consultant. Create a 10-slide executive storyline for [TOPIC]. For each slide provide one outcome-led headline, evidence required, recommended visual and speaker note.
Report Quality Review
Review the report below from an executive perspective. Identify unnecessary detail, repetition, unsupported conclusions, unclear ownership, unmeasurable actions and missing decision requests. Restructure it to be concise, evidence-led and decision-focused. Report: [PASTE]
5. Prompt review checklist
- Is the business problem and target user clear?
- Are the model role and boundaries explicit?
- Is the source and currency of input data known?
- Are the format and quality standard measurable?
- Are missing data, assumptions and uncertainty visible?
- Have privacy, IP and regulatory risks been assessed?
- Is human approval required for high-impact decisions?
- Will the result be verified against real evidence?
6. Managing an enterprise prompt library
Prompts should be governed as controlled digital assets rather than personal notes. Record purpose, process, owner, approval level, permitted data classification, target model, version, last test date and sample output. This improves reuse while preventing outdated or risky prompts from spreading.
Does AI use in your organisation depend on individual trial and error, or is there a standard prompt approach that jointly governs accuracy, data security and human oversight?
