EvalCommunity Tutorial

How to Use AI to Write a MEL Plan

A practical step-by-step tutorial for MEL professionals who want to use AI to draft, structure, review, and improve a Monitoring, Evaluation and Learning plan.

Part of the Planning Guides

Back to AI in Monitoring & Evaluation
|
Browse AI Prompt Library

On this page

• Takeaway
• Why use AI for a MEL plan?
• Before you start
• The 5-section MEL plan workflow
• Weak vs. strong AI-assisted drafting
• 5 rules for AI-assisted MEL plans
• MEL plan starter prompt
• Follow-up prompts
• Quality review checklist
• FAQ
• Related resources

Never ask AI to “write a MEL plan” in one prompt. Break the MEL plan into sections, feed each section the context it needs, and treat every AI output as a first draft that requires professional review.

AI can help you:

• Create a MEL plan outline.
• Translate a project description into a results framework.
• Suggest indicators, baselines, targets, and data sources.
• Draft an Indicator Performance Tracking Table.
• Design a data collection plan.
• Write data quality and data management procedures.
• Develop learning questions and adaptive management routines.
• Review the draft for gaps, inconsistencies, and weak assumptions.

AI cannot replace:

• Stakeholder consultation.
• Donor compliance review.
• Ethical judgement.
• Contextual knowledge of the programme.
• Validation of indicators, baselines, targets, and assumptions.
• Final approval by a qualified MEL professional.

Recommended inputs

• Project proposal or concept note.
• Theory of Change or logical framework.
• Donor guidance or reporting requirements.
• Target population and geographic scope.
• Key activities and implementation timeline.
• Existing indicators, if available.
• Staffing, tools, and budget constraints.
• Known risks, assumptions, and ethical considerations.

Data protection reminder: Do not upload confidential, sensitive, or personally identifiable data into public AI tools unless your organization, donor rules, consent procedures, and data protection policies clearly allow it.

Step 1

Create the results framework

Start by asking AI to map the Theory of Change into a results hierarchy. This usually includes the goal, outcomes, outputs, and sometimes activities.

Output to request: a table with result level, result statement, suggested indicators, data source, frequency, assumptions, and gaps.

Step 2

Develop indicators and targets

Feed the results framework back to AI and ask it to propose 2–3 SMART indicators per result level. Include disaggregation requirements such as sex, age, disability status, location, or population group.

Output to request: an Indicator Performance Tracking Table with definitions, baselines, targets, data sources, frequency, responsibility, and validation notes.

Step 3

Build the data collection plan

Ask AI to recommend a data collection method, data source, frequency, tool, responsible person, and limitation for each indicator. Provide realistic constraints so the plan is implementable.

Output to request: a data collection matrix linked directly to the indicator table.

Step 4

Draft data quality and data management procedures

Ask AI to draft procedures for validation, cleaning, storage, access control, spot checks, data quality assessments, and correction of errors. Mention the tools your team uses, such as KoboToolbox, Excel, DHIS2, CommCare, Power BI, or Google Sheets.

Output to request: a practical data quality assurance section with clear responsibilities and review routines.

Step 5

Design learning and adaptive management routines

Ask AI to design a learning cycle that explains what evidence will be reviewed, who participates, what decisions can be made, and how adaptations will be documented.

Output to request: learning questions, review meetings, pause-and-reflect moments, decision points, and documentation methods.

Example 1: Results framework

Weak prompt: “Write a results framework for a nutrition project.”

This usually produces a generic framework with vague outcomes and little connection to actual activities, target groups, donor requirements, or implementation context.

Strong prompt:

Example 2: Indicator selection

Weak prompt: “Give me indicators for a health project.”

This may produce generic indicators that are not linked to decisions, data systems, disaggregation needs, or team capacity.

Strong prompt:

Example 3: Learning section

Weak prompt: “Write the learning section of my MEL plan.”

This often generates boilerplate language about “promoting a culture of learning,” but without practical mechanisms or decision points.

Strong prompt:

1. Work section by section, not all at once

Draft the results framework first. Then use that output to develop indicators. Then use the indicators to develop the data collection plan. This creates a logical chain across the MEL plan.

2. Specify the donor upfront

USAID, FCDO, EU, World Bank, UN agencies, foundations, and government donors may use different terminology, reporting formats, and indicator expectations. Tell AI the donor from the beginning so the structure and language fit the expected format.

3. Include team capacity and constraints

AI may recommend unrealistic data collection plans if it does not know your staffing, tools, geographic coverage, budget limits, and partner roles. Give these constraints before asking for methods or schedules.

4. Validate indicators with the Decision Test

For every indicator AI proposes, ask: “What decision will this indicator inform, and who will make that decision?” If the indicator does not support a real management, learning, accountability, or reporting decision, it may not be worth collecting.

5. Treat the AI draft as scaffolding

AI can produce a useful first draft, but it cannot fully understand your implementation context, stakeholder dynamics, donor relationship, political environment, or field realities. Use the output as structure, not as the final document.

AI-Assisted MEL Plan: Results Framework Prompt

Copy and adapt the prompt below.

I need to draft a Monitoring, Evaluation and Learning plan for a project. Start with the results framework.

Project details:

• Project title: [PROJECT NAME]
• Duration: [e.g. 3 years, October 2026 to September 2029]
• Donor: [USAID / FCDO / EU / World Bank / UN agency / foundation / other]
• Sector: [e.g. nutrition, WASH, education, livelihoods, governance, health]
• Target population: [e.g. 15,000 children under 5 in Northern Kenya]
• Geographic scope: [e.g. 3 districts in Turkana County]
• Key activities: [list 4–6 main activities]
• Implementing partners: [if relevant]
• Available MEL capacity: [e.g. 1 MEL manager, 3 field monitors, KoboToolbox, Excel]

Please produce:

1. A results framework table with these columns: Result Level, Result Statement, Suggested Indicators, Data Source, Frequency, Assumptions, and Gaps.
2. Use result levels appropriate for the donor.
3. Ensure every output logically contributes to an outcome, and every outcome contributes to the overall goal.
4. Flag any gaps where activities do not clearly map to outputs or where the logic is weak.
5. Keep the language clear, professional, and suitable for a MEL plan.
6. Present the output as a clean table.

Indicator table prompt

Using the results framework below, create an Indicator Performance Tracking Table. For each result statement, propose 2–3 SMART indicators. Include: Indicator, Definition, Baseline, Year 1 Target, Year 2 Target, Year 3 Target, Data Source, Collection Method, Frequency, Disaggregation, Responsible Person, and Notes. Flag any indicator that requires further validation.

Data collection plan prompt

Using the indicator table below, create a data collection plan. For each indicator, recommend the data source, collection method, frequency, responsible person, tools required, and potential limitations. Make the plan realistic for this team capacity: [describe team, tools, budget, and geographic coverage].

Data quality prompt

Using the indicator table and data collection plan below, draft a Data Quality Assurance section for the MEL plan. Include procedures for validation, spot checks, data cleaning, data storage, access control, correction of errors, review meetings, and documentation.

Learning and adaptation prompt

Using the project description, results framework, and indicator table below, draft a Learning and Adaptive Management section. Include learning questions, review meetings, pause-and-reflect sessions, participants, decision points, documentation methods, and how findings will inform programme adaptation.

Full plan review prompt

Review the MEL plan below as a senior MEL advisor. Identify gaps, weak indicators, unrealistic data collection methods, missing responsibilities, unclear learning processes, data quality risks, ethical concerns, and inconsistencies between sections. Provide practical recommendations for improvement.

• Does the plan align with the project objectives?
• Are the result levels logical and consistent?
• Are indicators specific, measurable, achievable, relevant, and time-bound?
• Are indicators realistic for the team’s capacity?
• Are baselines and targets included or clearly marked as pending?
• Are data sources available and reliable?
• Are collection methods feasible?
• Are responsibilities clearly assigned?
• Is the reporting schedule included?
• Are data quality procedures practical?
• Are ethical and data protection risks addressed?
• Does the learning section include real decision points?
• Has the AI-generated content been reviewed by a MEL professional?
• Has the draft been validated with relevant stakeholders?

Can AI write a complete MEL plan?

AI can draft parts of a MEL plan, but it should not be used to produce a final plan without review. The strongest approach is to draft section by section, then validate the full document with a MEL professional and relevant stakeholders.

What is the best AI prompt for a MEL plan?

The best prompt is specific. It should include the donor, sector, project duration, target population, key activities, geographic scope, team capacity, available tools, and the exact output format you want.

Should I ask AI to create indicators?

Yes, AI can suggest indicators, but every indicator must be reviewed. Check whether it is SMART, feasible, aligned with the result statement, connected to a decision, and realistic for your data system.

Can I upload donor documents to AI?

Only if your organization’s data protection rules, donor agreements, and confidentiality requirements allow it. For sensitive or restricted documents, use approved tools and remove confidential information where needed.

How do I avoid a generic AI-generated MEL plan?

Provide detailed context, specify the donor, include constraints, request tables, ask AI to flag gaps, and review every section. Generic prompts produce generic plans.

Continue learning with EvalCommunity Academy

Explore practical AI tutorials, prompt tools, and courses designed for evaluators, MEL professionals, researchers, and development practitioners.

Explore the AI in M&E Course