10 Prompt Qualifiers of ChatGPT for Monitoring and Evaluation
EVALCOMMUNITY ACADEMY TUTORIAL
10 Prompt Qualifiers That Make ChatGPT More Useful for Monitoring and Evaluation
Small instructions can make AI-generated M&E outputs clearer, more focused, more actionable, and easier to verify.
ChatGPT can support monitoring and evaluation professionals with indicator design, report summaries, qualitative analysis, evaluation questions, learning products, and stakeholder communication.
The quality of the response, however, depends heavily on how the request is framed. Longer prompts are not automatically better. In many situations, the fastest improvement is to add a short qualifier.
A prompt qualifier is a brief instruction that controls the length, audience, structure, evidence boundaries, degree of certainty, or expected next action.
Simple example
Summarise the main findings from this evaluation report.
Answer in no more than 150 words.
The first sentence defines the task. The second controls the output.
Learning objectives
By the end of this tutorial, you should be able to:
- control the length and structure of AI outputs;
- reduce generic or overly broad responses;
- prioritise decision-relevant information;
- generate more actionable recommendations;
- adapt content for different M&E audiences;
- surface uncertainty, assumptions, and evidence gaps.
Why qualifiers matter in M&E work
M&E work may involve evaluation reports, interview transcripts, survey results, theories of change, indicator tables, stakeholder feedback, programme documents, and workshop notes. When instructions are broad, ChatGPT may produce an answer that is relevant but not operationally useful.
Too broad
Review this monitoring report.
ChatGPT does not know whether “review” means correcting grammar, identifying missing data, testing conclusions, checking indicators, summarising findings, or developing recommendations.
More controlled
Review this monitoring report. Focus only on evidence gaps, unsupported conclusions, and inconsistent indicator data.
The 10 qualifiers
- Answer in 100 words or less
- Be clear and concise
- Tell me the three things I need to know
- Recommend the best next step
- Ask me three questions before answering
- Assume I already understand the basics
- Write for a smart but busy professional
- Avoid generic advice
- Give me one direct recommendation
- If you are uncertain, say so explicitly
QUALIFIER 1
“Answer in 100 words or less”
What it does
Places a clear limit on the response and forces ChatGPT to prioritise the most important information.
When to use it
Use it for executive summaries, dashboard narratives, management updates, donor briefings, email summaries, and presentation text.
M&E prompt example
Summarise the main implementation challenge described in these field-monitoring notes.
Answer in 100 words or less and include only information supported by the notes.
Use with care
Do not use an extremely short word limit when the task requires nuance, detailed evidence, or methodological explanation.
QUALIFIER 2
“Be clear and concise”
What it does
Asks ChatGPT to simplify the response without imposing a strict word limit.
When to use it
Use it for emails, guidance notes, indicator explanations, survey instructions, meeting summaries, and short methodological explanations.
M&E prompt example
Explain the difference between an output, an outcome, and an impact.
Be clear and concise. Use one development-sector example throughout.
Stronger version
Be clear and concise. Avoid long introductions, repeated points, and unnecessary definitions.
QUALIFIER 3
“Tell me the three things I need to know”
What it does
Forces prioritisation. Instead of a long list, ChatGPT selects the issues that appear most important.
When to use it
Use it when reviewing evaluation findings, performance updates, programme risks, data-quality problems, stakeholder concerns, or implementation delays.
M&E prompt example
Review this quarterly indicator table.
Tell me the three things the programme manager needs to know before the review meeting. Prioritise issues that could affect targets, reporting credibility, or implementation decisions.
Important
Define what “important” means. ChatGPT may not know which issue is most politically sensitive, operationally urgent, or relevant to a donor.
QUALIFIER 4
“Recommend the best next step”
What it does
Shifts the response from description and analysis to practical action.
When to use it
Use it when an indicator is off track, data are incomplete, partner figures conflict, response rates are low, findings are disputed, or implementation is delayed.
M&E prompt example
The current indicator value is 42%, compared with a target of 65%. Two partner reports are still missing, and the reporting period ends in five days.
Recommend the best next step the M&E officer should take. Explain the reason in two sentences.
Add a realistic constraint
Recommend the best next step that can be completed within five working days.
QUALIFIER 5
“Ask me three questions before answering”
What it does
Encourages ChatGPT to clarify missing information before producing a final response.
When to use it
Use it when the task depends on the intended audience, programme context, evaluation purpose, available data, donor requirements, or the decision to be supported.
M&E prompt example
Help me develop an evaluation matrix for a livelihoods programme.
Ask me three essential questions before creating the matrix.
Use with care
Do not use this qualifier for every prompt. When the task is simple and the information is already available, extra questions can slow the work down.
QUALIFIER 6
“Assume I already understand the basics”
What it does
Tells ChatGPT not to spend time explaining introductory concepts.
When to use it
Use it for methodology selection, indicator design, contribution analysis, qualitative coding, statistical interpretation, mixed-methods integration, or evidence synthesis.
M&E prompt example
Explain how to assess contribution in a complex policy programme.
Assume I already understand theories of change and basic evaluation terminology. Focus on practical design choices and evidence requirements.
For beginners, reverse it
Assume I have no previous experience. Explain each step in plain language and define technical terms when they first appear.
QUALIFIER 7
“Write for a smart but busy professional”
What it does
Adjusts tone, depth, and structure for readers who understand complex issues but have limited time.
When to use it
Use it for programme directors, donors, government officials, evaluation commissioners, board members, technical advisers, and partner organisations.
M&E prompt example
Write a briefing note explaining why the current outcome indicator cannot demonstrate programme impact.
Write for a smart but busy programme director. Use plain language, short paragraphs, and one practical recommendation.
Focus the answer
A busy decision-maker usually needs to understand the issue, why it matters, what decision is required, and what should happen next.
QUALIFIER 8
“Avoid generic advice”
What it does
Discourages vague recommendations such as “improve communication,” “strengthen capacity,” or “engage stakeholders.”
When to use it
Use it for evaluation recommendations, programme-improvement actions, risk responses, data-quality solutions, stakeholder-engagement plans, and learning actions.
M&E prompt example
Propose recommendations based on these evaluation findings.
Avoid generic advice. Each recommendation must identify the action, responsible actor, expected result, and suggested timeframe.
Make recommendations actionable
Instead of “strengthen partner capacity,” specify who should do what, by when, and how the next submission will be checked.
QUALIFIER 9
“Give me one direct recommendation”
What it does
Asks ChatGPT to take a position instead of presenting a long list of alternatives.
When to use it
Use it when choosing an indicator, chart, data-collection method, reporting approach, or whether more evidence is needed.
M&E prompt example
We have resources for either 20 key-informant interviews or four focus-group discussions. The evaluation question concerns how national and local institutions experienced implementation.
Give me one direct recommendation, not a list of options. State the main trade-off.
For higher-risk decisions
Add: “State your assumptions and identify one condition that would change your recommendation.”
QUALIFIER 10
“If you are uncertain, say so explicitly”
What it does
Asks ChatGPT to distinguish supported information from interpretation, assumptions, missing evidence, and uncertainty.
When to use it
Use it for evaluation findings, statistical interpretation, causal claims, performance conclusions, and analysis based on incomplete documents.
M&E prompt example
Review these interview summaries and identify the main factors affecting service uptake.
If the evidence is insufficient or contradictory, say so explicitly. Do not present assumptions as findings.
Stronger version
For each conclusion, label it as strongly supported, partially supported, or uncertain. Explain which evidence is missing.
Add sources and traceability
When using external information, ask for sources. When analysing documents you provide, ask ChatGPT to connect every conclusion to a page, section, paragraph, table, or quotation.
Verify every AI-generated reference. Confirm that the source exists, is current, and supports the claim.
How to combine qualifiers
Combine qualifiers when they control different parts of the output. Avoid many overlapping instructions.
Executive summary
Summarise the main findings from this evaluation report.
Write for a smart but busy programme director.
Answer in no more than 200 words.
Tell me the three things they need to know.
If the evidence is uncertain, say so explicitly.
Evaluation recommendation
Review the finding below and propose one recommendation.
Avoid generic advice.
Identify the responsible actor and timeframe.
Give one direct recommendation, not several options.
Indicator analysis
Analyse this indicator-performance table.
Assume I already understand basic M&E terminology.
Identify the three most important performance or data-quality issues.
Recommend the best next step for each issue.
Do not make assumptions about missing data.
Reusable qualifier menu
Control length
Answer in no more than 100 words. Use no more than five bullets. Give me a one-paragraph summary.
Control priority
Tell me the three most important findings. Rank issues from highest to lowest risk. Focus only on issues requiring management action.
Improve actionability
Recommend the best next step. Identify the responsible actor and timeframe. Convert each finding into a practical action.
Control the audience
Write for a programme manager, donor, experienced evaluator, or new M&E officer.
Improve transparency
State your assumptions. Identify missing information. Separate findings from interpretations. Do not invent data, quotations, sources, or stakeholder views.
Common mistakes to avoid
- Using a qualifier without enough context. Explain the situation, objective, constraints, and available evidence.
- Asking for concision when detail is required. Match the word limit to the purpose of the output.
- Requesting certainty from weak evidence. Ask for the most defensible conclusion, limitations, and confidence level.
- Requesting a recommendation without criteria. Define whether “best” means fastest, least costly, most rigorous, safest, or most useful.
- Treating a well-written answer as verified. Check important outputs against the original data and programme context.
Practice exercise
A programme has achieved 58% of its annual target. Two implementing partners have not submitted their latest data, and several beneficiary records contain duplicate identification numbers. The programme manager needs a short update before tomorrow’s meeting.
Weak prompt
Analyse this information.
Improved prompt
Analyse the programme-performance information below.
Write for a smart but busy programme manager.
Tell me the three things they need to know before tomorrow’s meeting.
Answer in no more than 150 words.
Recommend the best immediate next step.
Separate confirmed facts from possible explanations.
If the evidence is insufficient, say so explicitly.
A practical prompt formula
- Context: describe the programme, document, task, or situation.
- Task: state exactly what ChatGPT should do.
- Evidence: define the information the model may use.
- Qualifiers: control audience, length, priority, and uncertainty.
- Format: define how the final response should be presented.
You are supporting a midterm evaluation of a youth-employment programme. Review the interview summaries and identify factors affecting participation.
Use only the supplied evidence.
Tell me the three most important factors.
For each factor, provide supporting evidence, an alternative explanation, and your level of confidence.
Write for an experienced evaluator and avoid generic recommendations.
Present the answer in a structured list.
Final prompt checklist
- Is the task clear?
- Is enough context provided?
- Is the intended audience defined?
- Are length and priorities specified?
- Are the evidence boundaries clear?
- Should uncertainty be identified?
- Is an actionable next step required?
- Is the output format defined?
- Will a human review the result?
Conclusion
Better AI outputs do not always require longer prompts. A short, well-chosen qualifier can improve focus, clarity, usefulness, and transparency.
Qualifiers provide control, but they do not replace professional judgement. AI-generated findings, interpretations, recommendations, and citations should always be checked against the original evidence and programme context.
Continue learning with EvalCommunity Academy
Explore practical training on the responsible and effective use of AI in monitoring, evaluation, accountability, and learning.
