Mission goal: understand what the AI-enhanced version of your role is and why it is emerging.
Step 1 of 5Where are you starting from?
Choose your starting point+10 XP
Where are you today with AI in your M&E work?
Show answer
Every starting point is valid. Follow the mission suggested for your situation.
Step 2 of 5Meet your new role
Today's role
Monitoring and Evaluation Specialist
AI-enhanced role
AI-Enabled M&E Specialist
Level
Specialist · 3–6 years
How the role uses AI
Integrates AI into monitoring systems and is accountable for the quality of AI-supported results
"AI-Enabled M&E Specialist" is a proposed career title from the EvalCommunity framework, not a standardised occupational classification or a guaranteed vacancy.
Step 3 of 5What changes most in your job?
Your guess+10 XP
When routine data checking and first-draft reporting move to AI, where does your time go?
Show answer
In the profile, routine data checking, consolidation and first-draft reporting move to AI tools and automated workflows. More of your time goes to methodology, interpretation and helping managers act on findings. You stay accountable for indicator definitions, analytical choices and every figure that leaves the team.
Step 4 of 5AI-Enabled, not AI-Dependent
The goal is not to hand your work to AI. It is to become a stronger M&E specialist who uses AI and stays in charge of it. Here are the five questions, asked about your own work:
1QualityDo AI checks and drafts make your figures more accurate and traceable, or only faster?A traceability table, and an agreement score next to any AI coding.
2AccountabilityWhen AI miscodes answers or invents a cause, can you show what you checked?An exceptions log and a hand-checked sample for every round.
3IndependenceDo you still decide whether an anomaly is a data error or a real change?You query the field team before anything is corrected.
4EthicsDo respondents' names, phone numbers and locations stay out of AI tools?Anonymise first, and use approved tools only.
5Professional developmentCould you still do the analysis without AI, and explain every line of your script?Test every script and prompt on data where you already know the answer.
The five questions as the series states them
Quality: Does AI improve the quality and credibility of evaluation findings?
Accountability: Who is responsible when AI produces an incorrect interpretation?
Independence: Are evaluators maintaining their ability to think critically and make independent judgments?
Ethics: Are confidentiality, human rights and the interests of affected populations protected?
Professional development: Are we developing stronger evaluators who understand AI, or creating professionals who increasingly depend on it?
Step 5 of 5Your goal for this tutorial
By the end, you will know how to move your own daily tasks to AI step by step, which skills to build, and you will have a 12-week plan.
Next: see which of your tasks change, and how.
End of Mission 01
MISSION02
Traditional role vs AI-enhanced role
Before and after · Badge: Role Shifter
8 challenges
Mission goal: see which responsibilities become AI-assisted, which need new skills and which stay human.
Step 1 of 4Meet your case study
Case study · Rural Resilience Programme Illustrative
A fictional programme you will follow through missions 2, 3, 4, 8 and 9.
The programme
A three-year livelihoods and education programme in eight districts, with five implementing partners. Quarterly donor reports; mid-term evaluation in month 18.
The data
About 4,000 KoboToolbox household submissions per quarter, partners' monthly Excel returns, and a dropout survey with 1,800 open-ended answers.
The tools
KoboToolbox, Excel, Python (pandas), Power BI and NVivo, plus an approved AI assistant.
Your objective
Move routine checking, coding and drafting to AI, while every reported figure stays traceable to source.
Step 2 of 4Flip your tasks: today and with AI
Most of your work is not replaced. AI takes the routine part, and you check it. Tap each card to see how it changes.
Step 3 of 4Mini-game: sort the work
For each task, choose: AI-assisted, a new skill for you, or stays human.
Sort it+10 XP
Flagging duplicate IDs and impossible ages in 4,000 KoboToolbox submissions
Show answer
AI-assisted. Routine screening is exactly what the profile moves to a script with an AI step.
Sort it+10 XP
Writing the first draft of the narrative for an off-track indicator
Show answer
AI-assisted. Copilot can draft it; you correct it and add the field explanation.
Sort it+10 XP
Consolidating five partners' monthly Excel returns
Show answer
AI-assisted. In the profile, an automated workflow replaces two days of copy-pasting each month.
Sort it+10 XP
Writing instructions that make AI coding of open-ended answers reliable
Show answer
New skill. Prompt design: instructions, examples and output formats that give reliable results.
Sort it+10 XP
Hand-coding a random sample of AI-coded answers to check agreement
Show answer
New skill. Validating AI outputs is the profile's first learning priority.
Sort it+10 XP
Choosing indicators, definitions and targets with programme teams
Show answer
Human. Listed in the profile as a judgement that stays human.
Sort it+10 XP
Deciding whether an anomaly is a data error or a real change on the ground
Show answer
Human. AI can flag it. You, the data and the field team decide what it means.
Sort it+10 XP
Deciding which respondent data may be sent to an AI service
Show answer
Human. Protecting respondents' personal data stays with you.
Step 4 of 4Your turn: two minutes
Think about last week. Which of your own tasks would go in each pile?
Think about it for a moment, then reveal.
A quick way to do it: three columns on a sheet of paper, AI-assisted, new skill, stays human. Keep the list: you will use it in mission 4 (to pick a task to try) and mission 6 (to rate your gaps).
End of Mission 02
MISSION03
What AI actually changes in your daily work
Checks, risks and red flags · Badge: Red Flag Spotter
2 challenges
Mission goal: learn the checks that make AI-supported analysis usable, and practise spotting problems in AI drafts.
Step 1 of 4What can AI really do for you?
Before you look: which of your tasks could AI help with, and which must stay with you?
Think about it for a moment, then reveal.
AI can assist with
Screening submissions for missing and unusual values
First drafts of indicator narratives
First-pass coding of open-ended answers
Trend analysis by site and partner
Can be partly automated
Consolidating partners' monthly returns
Dashboards with plain-language KPI notes
Simple rules that flag sites at risk
New responsibilities
Writing prompts and codebooks AI can follow
Hand-checking samples of AI output
Keeping every figure traceable
Safe handling of respondent data in AI tools
Stays with you
Indicators, definitions and targets
Data error or real change?
Interpretation and recommendations
Every figure that leaves the team
Step 2 of 4The three checks you must always run
AI output is only usable when you can check it. Here are the checks, one at a time.
Traceability. Every AI-supported figure in a donor report can be traced to source data. This is the profile's first learning priority.
Hand-checked samples. You code or check a sample by hand before using AI results (150 of 1,800 answers in the profile's example).
Field evidence. AI findings are validated against the original datasets and what field teams report, before anything is reported.
Step 3 of 4Mini-game: spot the red flags
Case study · Rural Resilience Programme Illustrative
Copilot drafted the Q2 narrative for indicator 1.2. Here is the source data and the draft.
Click every sentence you would not send to the donor.
What the IPTT and field notes show for Q2
Farmers trained (indicator 1.2)
840 against a target of 1,200 (70%)
Districts
8 reported; 2 held no trainings (sessions postponed after flooding)
Women among trainees
45%
Explanation from farmers
None collected yet
Red flags found: 0 of 3+30 XP
Show answer
Three sentences should not be sent: the invented cause ("farmers are not interested"), "all districts on track except one" (two held no trainings) and the promise that the target will be met.
Step 4 of 4Mission check
Your call+20 XP
Your screening script reports zero problems in this week's 600 submissions, for the first time ever. What do you do first?
Show answer
Treat it as a signal to check: confirm the script ran on every submission and that the form or field names did not change.
End of Mission 03
MISSION04
Your daily tasks, step by step with AI
Six practical guides · Badge: Workflow Builder
6 challenges
Mission goal: learn how to move six of your real daily tasks to AI, step by step, with the checks that keep every figure traceable.
Case study · Rural Resilience Programme Illustrative
Every worked example below uses this programme's data, so you can follow one system from collection to the donor report.
Six real tasks from your working week. Pick one and go through it in six short steps: a task prompt, then a QA prompt that checks the AI's work.
YouAIYour teamAutomated shows who does each step. Examples are illustrative.
Choose a task
Guide 1 of 6 From the career profile
Screen KoboToolbox submissions before the report
From the profile: reviewing last week's KoboToolbox submissions for missing values and out-of-range entries; before each quarterly report, a Python script with an LLM step reviews 4,000 submissions, lists duplicate IDs, impossible ages and outlier distribution quantities, and drafts follow-up queries to field teams.
Step 1 of 6Today, and with AI
How it is usually done today Illustrative
You export the data and scroll through it in Excel, filter by filter.
Some problems are found, others reach the IPTT unnoticed.
Writing follow-up messages to each field team takes another afternoon.
How it works with AI
A script runs the same checks on every submission, every week.
It produces an exceptions list: duplicates, impossible values, outliers.
AI drafts the follow-up query for each field team.
You spend your time deciding what each exception means.
Step 2 of 6Step by step
YouWrite the rules firstExcel
List what counts as a problem: required fields, valid ranges (for example age 0–110), ID format, and how far from the median a quantity can be before you look at it.
AutomatedExport the submissionsKoboToolbox
Download the latest data from KoboToolbox as a CSV or Excel file, always in the same format.
AIGet the script writtenClaude · Python (pandas)
If you don't code yet, ask Claude to write a pandas script for your rules and to explain each line. Test it on a file where you already know the problems.
AutomatedRun the checksPython (pandas)
The script lists duplicate IDs, impossible ages and outlier quantities in an exceptions file, with the submission ID and the reason.
AIDraft the follow-up queriesClaude
Give the AI the exceptions list, without names or phone numbers, and ask for one short query per field team (prompt below).
YouDecide what each exception means
An outlier can be an entry error or a real change on the ground. Read each query before it is sent, and confirm with the field team before you correct anything.
Step 3 of 6Worked example
What you give the AI Illustrative
Submission ID
Team
Problem found
HH-2231
Team North
Household ID submitted twice (3 May and 10 May)
HH-2458
Team North
Respondent age 140
HH-3017
Team East
Received 60 kg of maize; median is 20 kg
What would you ask the AI to do with this?
Think about it for a moment, then reveal.
The task prompt (copy and adapt)
You are helping an M&E specialist follow up on data problems.
Below is a list of flagged KoboToolbox submissions with the team and the problem found.
For each team, write one short, polite message listing its submission IDs and asking the team to check each problem against the paper form or the respondent.
Rules: do not say the data is wrong, do not suggest a corrected value, and do not add problems that are not in the list.
[paste the exceptions list: IDs, team, problem; no names or phone numbers]
What comes back
To Team North: Please check two submissions: HH-2231 appears twice (3 May and 10 May), and HH-2458 records a respondent age of 140. Could you confirm both against the forms?
To Team East: Submission HH-3017 records 60 kg of maize, well above the usual 20 kg. Could you confirm the quantity with the distribution list?
Notice the queries ask, they don't correct. The 60 kg may be real (a larger household, a double ration). That call is yours, after the team replies.
Step 4 of 6Check the AI output
Independence first: read the draft and note the problems you see yourself, before you run the QA prompt. Then compare.
Another draft came back like this Illustrative
To Team North: HH-2458 has a wrong age of 140; please correct it to 40. HH-2231 is a duplicate and has been deleted. Please also check HH-2500, which looks suspicious.
To Team East: No problems this week.
The QA prompt (run it on every draft)
Act as an independent data quality reviewer.
Compare the draft follow-up queries below with the exceptions list.
List: every submission ID in the list that is missing from the queries, every ID in the queries that is not in the list, every corrected value or deletion the queries announce, and any name, phone number or location.
Do not rewrite the queries. List the problems only.
Exceptions list: [paste]
Draft queries: [paste]
What should the check catch?
Think about it for a moment, then reveal.
What the check finds
It suggests a corrected value (age 40) that nobody confirmed.
It says HH-2231 was deleted: deleting before the team replies breaks traceability.
HH-2500 is not in the exceptions list: an invented problem.
Team East's flag (HH-3017, 60 kg of maize) is missing.
Corrected output
To Team North: Please check two submissions: HH-2231 appears twice (3 May and 10 May), and HH-2458 records a respondent age of 140. Could you confirm both against the forms?
To Team East: Submission HH-3017 records 60 kg of maize, well above the usual 20 kg. Could you confirm the quantity with the distribution list?
Step 5 of 6Your checks
Tick each check as you do it on your own work. Each one earns +5 XP.
Independence: the script finds the exceptions. You decide what each one means, with the field team.
Step 6 of 6Quick check
Quick check+25 XP
The script flags a distribution quantity three times the median. What do you do?
Show answer
Query the field team first. Deciding whether an anomaly is an error or a real change stays with you.
Try it this week
Take last week's KoboToolbox export and write down your rules (step 1). Then ask an AI assistant to turn them into a pandas script, and test it on that export.
From the profile: an AI assistant applies a 12-code framework to 1,800 open-ended answers on reasons for dropout; the specialist hand-codes 150 of them to check agreement before using the results.
Step 1 of 6Today, and with AI
How it is usually done today Illustrative
Open-ended questions are dropped or only skimmed, because coding takes weeks.
Or AI is used to summarise answers, and nobody checks what it decided.
How it works with AI
You write the codebook, and AI applies it to every answer.
You hand-code a random sample without looking at the AI codes.
You compare the two and use the results only if agreement is good enough.
Step 2 of 6Step by step
YouWrite the codebookExcel · NVivo
Define each code with a one-line definition and one example answer (the profile's example uses 12 codes). Add UNSURE and OTHER.
YouRemove personal details
Strip names, places small enough to identify people, and phone numbers before any answer goes into an AI tool.
AICode the answers in batchesChatGPT
Give the AI the codebook and a batch of answers with IDs. Ask for one code and a short reason per answer (prompt below).
YouHand-code a random sample blindExcel
Pick a random sample (150 of 1,800 in the profile) and code it yourself without looking at the AI codes.
AutomatedCalculate agreementExcel
Agreement = answers where your code matches the AI code ÷ answers checked. Look at which codes disagree most.
YouFix, then decideNVivo
If agreement is below the level you set in advance, sharpen the codebook and run it again. Use the results only when it passes, and say how they were checked.
Step 3 of 6Worked example
What you give the AI Illustrative
ID
Answer (anonymised)
R-0112
"My daughter had to help at home after her father got sick."
R-0347
"The school is too far and the road floods."
R-0981
"We could not pay for the uniform this year."
What would you ask the AI to do with this?
Think about it for a moment, then reveal.
The task prompt (copy and adapt)
Apply this codebook to each answer about why a child stopped attending school.
Codebook: [paste your codes, each with a one-line definition and one example].
Rules: give exactly ONE code per answer. If no code clearly fits, write UNSURE. Do not create new codes.
Return a table: ID, code, a reason of 10 words or fewer.
[paste anonymised answers with IDs]
What comes back
ID
AI code
Reason
Your code
R-0112
Household labour
Child needed to help at home
Household labour
R-0347
Distance and access
School far, road floods
Distance and access
R-0981
Cost of schooling
Could not pay for uniform
Cost of schooling
In a real round you compare all 150 sampled answers. Say 129 of 150 match: agreement = 129 ÷ 150 = 86%. Whether that is good enough is a level you set before you start, not after.
Step 4 of 6Check the AI output
Independence first: read the draft and note the problems you see yourself, before you run the QA prompt. Then compare.
Another draft came back like this Illustrative
ID
AI code
Reason
R-0112
Household labour; Health
Father sick, child helps at home
R-0347
Distance and access
School far, road floods
R-0520
Early marriage
"She left to live with her husband's family"
R-0981
Cost of schooling
Could not pay for uniform
The QA prompt (run it on every draft)
Act as an independent reviewer of qualitative coding.
Check the AI coding below against the codebook.
List: every answer with more than one code, every code that is not in the codebook, every answer that should be UNSURE because no code clearly fits, and every answer ID missing from the output.
Do not recode the answers. List the problems only.
Codebook: [paste]
AI coding: [paste]
What should the check catch?
Think about it for a moment, then reveal.
What the check finds
R-0112 has two codes; the codebook allows exactly one.
"Early marriage" is not in the codebook: the AI created a new code.
R-0520 should be OTHER or UNSURE, for you to decide, and possibly a new code agreed before the next round.
Corrected output
ID
AI code
Reason
R-0112
Household labour
Child needed to help at home
R-0347
Distance and access
School far, road floods
R-0520
UNSURE
Possible new code: to review
R-0981
Cost of schooling
Could not pay for uniform
Answers marked UNSURE go to you, not back to the AI.
Step 5 of 6Your checks
Tick each check as you do it on your own work. Each one earns +5 XP.
Quality: report the agreement score next to the results, so readers can judge how far to trust them.
Step 6 of 6Quick check
Quick check+25 XP
AI coded all 1,800 answers in ten minutes. When can the theme percentages go in the report?
Show answer
Use the results only after a random hand-coded sample (150 of 1,800 in the profile) shows good enough agreement.
Try it this week
Take 50 answers from a past survey you already coded by hand. Run them through the prompt above and calculate agreement. It tells you how good your codebook really is.
From the profile: updating the IPTT ahead of the quarterly donor report and querying partners on late data; Power BI pulls IPTT figures and Copilot drafts a paragraph per off-track indicator, which the specialist corrects and adds field explanations to.
Step 1 of 6Today, and with AI
How it is usually done today Illustrative
You update the IPTT, then write a paragraph for every indicator from scratch.
Off-track indicators get the same attention as on-track ones.
Field explanations arrive late, by email, and get lost.
How it works with AI
Power BI shows which indicators are off track.
Copilot drafts one factual paragraph per off-track indicator.
You check every figure and add what only the field team knows.
Step 2 of 6Step by step
Your teamClose the IPTTExcel
Update the IPTT and query partners on late data. Mark any figure that is still missing instead of estimating it.
AutomatedFlag the off-track indicatorsPower BI
Power BI reads the IPTT and lists indicators below the threshold you set against target.
AIDraft one paragraph per indicatorMicrosoft 365 Copilot
Copilot drafts a factual paragraph per off-track indicator, with a placeholder for the explanation (prompt below).
YouCheck every figureExcel
Compare each number in the draft with the IPTT cell it came from.
Your teamGet the field explanation
Ask the programme team why each indicator is off track, and what they will do about it.
YouWrite the explanation and the next step
Replace each placeholder with the field explanation. Remove any cause or promise the evidence does not support.
Step 3 of 6Worked example
What you give the AI Illustrative
Indicator
Target
Actual
% of target
1.2 Farmers trained
1,200
840
70%
2.1 Households with a kitchen garden
900
520
58%
3.1 Savings groups formed
40
39
98%
What would you ask the AI to do with this?
Think about it for a moment, then reveal.
The task prompt (copy and adapt)
Draft one short paragraph for each indicator below that is under 80% of target.
Use ONLY the figures in the table. State the target, the actual and the percentage.
Do not explain why it is off track and do not predict the year-end result: write [FIELD EXPLANATION NEEDED] where the explanation belongs.
Table: [paste the IPTT extract]
What comes back
1.2 Farmers trained. 840 farmers were trained against a target of 1,200 (70%). [FIELD EXPLANATION NEEDED]
2.1 Households with a kitchen garden. 520 households have a kitchen garden against a target of 900 (58%). [FIELD EXPLANATION NEEDED]
Indicator 3.1 (98%) is left out, as asked. The placeholders are your to-do list for the conversation with the programme team.
Step 4 of 6Check the AI output
Independence first: read the draft and note the problems you see yourself, before you run the QA prompt. Then compare.
Another draft came back like this Illustrative
1.2 Farmers trained. 840 farmers were trained against a target of 1,200 (75%), due to low interest among farmers. The target will be met by December.
3.1 Savings groups formed. 39 of 40 groups were formed (98%).
The QA prompt (run it on every draft)
Act as an independent M&E quality reviewer.
Compare the draft narrative below with the IPTT extract.
List: every figure or percentage that differs from the IPTT, every cause or prediction that is not in the source, every indicator included that should not be, and every off-track indicator that is missing.
Do not rewrite the narrative. List the corrections only.
IPTT extract: [paste]
Draft: [paste]
What should the check catch?
Think about it for a moment, then reveal.
What the check finds
840 of 1,200 is 70%, not 75%.
"Low interest among farmers" is an invented cause: the placeholder was filled in.
"The target will be met by December" is a prediction the data cannot support.
Indicator 3.1 is on track (98%) and should not be there; indicator 2.1 (58%) is missing.
Corrected output
1.2 Farmers trained. 840 farmers were trained against a target of 1,200 (70%). [FIELD EXPLANATION NEEDED]
2.1 Households with a kitchen garden. 520 households have a kitchen garden against a target of 900 (58%). [FIELD EXPLANATION NEEDED]
Step 5 of 6Your checks
Tick each check as you do it on your own work. Each one earns +5 XP.
Accountability: every figure in your narrative traces to an IPTT cell you checked, whoever drafted the text.
Step 6 of 6Quick check
Quick check+25 XP
The Copilot draft says an indicator is behind "due to low community interest". What do you do?
Show answer
Remove the invented cause and get the explanation from the programme team.
Try it this week
Take last quarter's IPTT and run the prompt above on the off-track indicators. Compare the drafts with what you actually wrote: what would you have had to fix?
Consolidate five partners' monthly returns automatically
From the profile: an n8n workflow collects five partners' monthly Excel returns, standardises indicator codes and loads them into one database, replacing two days of copy-pasting each month.
Step 1 of 6Today, and with AI
How it is usually done today Illustrative
Each partner emails an Excel file with its own indicator names.
You spend two days a month copy-pasting into the IPTT.
Late files and changed columns are found only when totals look wrong.
How it works with AI
Partners save their returns in one agreed place.
A workflow standardises indicator codes and loads the data into one database.
Unknown codes, missing partners and odd totals go to an exceptions list for you.
Step 2 of 6Step by step
YouAgree the indicator codesExcel
Make one list of standard indicator codes with their definitions from your reference sheets, and share one return template with partners.
AIMap each partner's labels onceClaude · ChatGPT
Ask an AI assistant to match each partner's indicator labels to your codes and mark anything uncertain (prompt below).
Your teamBuild the workflown8n · PostgreSQL
With IT or a data colleague, set up an n8n workflow: pick up each new file, apply the mapping, and load the rows into one database.
AutomatedValidate every loadn8n
The workflow sends unknown codes, missing partners and totals that change sharply to an exceptions list, instead of loading them.
YouWork the exceptions
Review the exceptions list each month and query partners on late or unclear data.
YouKeep the mapping up to date
When a partner changes a label, update the mapping and note the change, so every figure stays traceable.
Step 3 of 6Worked example
What you give the AI Illustrative
Partner
Their indicator label
Partner 1
No. of farmers trained (M+F)
Partner 2
Training participants - agriculture
Partner 3
Farmers reached
What would you ask the AI to do with this?
Think about it for a moment, then reveal.
The task prompt (copy and adapt)
Our standard indicator codes are: [paste codes with their definitions, e.g. IND-1.2 Farmers trained: number of farmers who completed at least one training session].
Below are the indicator labels used by three partners.
Match each label to one of our codes. If a label could mean something different from our definition, write UNSURE and explain why in one line.
Do not guess.
[paste partner labels; no personal data]
What comes back
Partner
Their label
Our code
Note
Partner 1
No. of farmers trained (M+F)
IND-1.2
Matches
Partner 2
Training participants - agriculture
IND-1.2
Check: may count one farmer twice
Partner 3
Farmers reached
UNSURE
Reached is not the same as trained
Step 4 of 6Check the AI output
Independence first: read the draft and note the problems you see yourself, before you run the QA prompt. Then compare.
Another draft came back like this Illustrative
Partner
Their label
Our code
Note
Partner 1
No. of farmers trained (M+F)
IND-1.2
Matches
Partner 2
Training participants - agriculture
IND-1.2
Matches
Partner 3
Farmers reached
IND-1.2
Matches
The QA prompt (run it on every draft)
Act as an independent data reviewer.
Check this label mapping against our indicator definitions and the partners' original labels.
List: every mapping marked as a match where the label could count something different from our definition, every partner label left out, and every partner missing from the mapping.
Do not fix the mapping. List the problems only.
Definitions: [paste]
Partner labels: [paste]
Mapping: [paste]
What should the check catch?
Think about it for a moment, then reveal.
What the check finds
Partner 2 counts participants, so one farmer at two sessions may be counted twice: not a sure match.
Partner 3's "reached" is not the same as "trained": it should be UNSURE.
Partners 4 and 5 are missing from the mapping.
Corrected output
Partner
Their label
Our code
Note
Partner 1
No. of farmers trained (M+F)
IND-1.2
Matches
Partner 2
Training participants - agriculture
IND-1.2
Check: may count one farmer twice
Partner 3
Farmers reached
UNSURE
Reached is not the same as trained
Partners 4 and 5
[to add]
Labels not yet received
Step 5 of 6Your checks
Tick each check as you do it on your own work. Each one earns +5 XP.
Quality: a workflow that guesses a mapping silently changes your totals. It must stop and ask.
Step 6 of 6Quick check
Quick check+25 XP
A partner sends a new indicator label the workflow doesn't recognise. What should happen?
Show answer
Unknown labels go to the exceptions list, and you confirm the meaning with the partner before loading.
Try it this week
Write down the indicator labels each partner used last month, next to your codes. That table is the brief for whoever builds the workflow.
Prepare for a meeting on an indicator 30% behind target
From the profile: meeting a programme manager to discuss why an output indicator is 30% behind target; AI-assisted trend analysis to spot delays and diverging sites earlier in the quarter; simple predictive indicators to flag sites at risk of missing targets.
Step 1 of 6Today, and with AI
How it is usually done today Illustrative
The gap is noticed at the end of the quarter, in the report.
The meeting starts with the total and ends with guesses.
Recommendations are general: "accelerate implementation".
How it works with AI
Trends by site show the gap weeks earlier, and which sites drive it.
A simple rule flags sites at risk of missing their target.
You arrive with checked facts, questions and options the manager can act on.
Step 2 of 6Step by step
AutomatedLook at the trend by sitePower BI
Break the indicator down by site and month in your dashboard, not just the total.
AIAsk for the patternClaude · Copilot
Give the AI the anonymised table by site and ask which sites diverge from the rest and since when (prompt below).
YouRule out a data problemExcel
For the diverging sites, check reporting completeness and recent DQA findings. Is the gap real, or are reports missing?
AutomatedFlag sites at riskExcel · Power BI
Use a simple rule you can explain, for example: a site below half of its quarterly target at mid-quarter is flagged at risk.
YouPrepare questions and options
Write two or three questions for the manager and two options with what each would take. AI can suggest options; you judge which fit the context.
YouAgree actions and follow up
In the meeting, agree who does what by when, and track it in the next update.
Step 3 of 6Worked example
What you give the AI Illustrative
Site
Target to date
Actual
% of target
Sites 1–5 (combined)
500
470
94%
Site 6
100
18
18%
Site 7
100
12
12%
Total
700
500
71%
What would you ask the AI to do with this?
Think about it for a moment, then reveal.
The task prompt (copy and adapt)
Below is an output indicator by site, against target to date.
Describe which sites diverge from the others and by how much.
Then list what I should check in the data before concluding the gap is real, and three questions to ask the programme manager.
Do not explain the cause and do not recommend actions yet.
[paste anonymised table by site]
What comes back
Pattern: sites 1–5 are at 94% of target. Sites 6 and 7 are at 18% and 12% and account for almost all of the gap.
Check first: did sites 6 and 7 submit all their reports? Did their reporting forms or indicator definition change?
Ask the manager: Did activities start on time at sites 6 and 7? Have staff or partners changed there? What support do these sites need?
The overall "30% behind" turns into a precise question about two sites. That makes the meeting short and the actions specific.
Step 4 of 6Check the AI output
Independence first: read the draft and note the problems you see yourself, before you run the QA prompt. Then compare.
Another draft came back like this Illustrative
Overall the indicator is at 75% of target. Sites 6 and 7 are behind because of staff turnover, and they should be closed so resources can go to sites 1 to 5.
The QA prompt (run it on every draft)
Act as an independent M&E reviewer.
Compare the trend summary below with the table by site.
List: every figure that differs from the table, every cause that is not in the data, and every recommendation the summary makes although it was asked not to.
Do not rewrite the summary. List the problems only.
Table: [paste]
Summary: [paste]
What should the check catch?
Think about it for a moment, then reveal.
What the check finds
The total is 500 of 700, which is 71%, not 75%.
"Staff turnover" is a cause the data does not show: it is a question for the manager.
Closing sites is a recommendation the prompt ruled out, and a decision for managers, not the AI.
Corrected output
The indicator is at 71% of target (500 of 700). Sites 1 to 5 are at 94%. Sites 6 and 7 are at 18% and 12% and account for almost all of the gap. Before the meeting: check reporting completeness at sites 6 and 7, and ask the programme manager what has changed there.
Step 5 of 6Your checks
Tick each check as you do it on your own work. Each one earns +5 XP.
Independence: AI shows the pattern. The explanation comes from your data checks and the field.
Step 6 of 6Quick check
Quick check+25 XP
The trend analysis shows two sites far behind. What do you take to the programme manager?
Show answer
Bring checked facts by site, your questions and actionable options. The interpretation is yours.
Try it this week
Pick the indicator furthest behind target in your IPTT. Break it down by site and write the three questions you would ask the programme manager.
From the profile: drafting the terms of reference for a mid-term evaluation; commissioning or co-leading baseline, midline and endline studies and managing evaluation consultants. The OECD-DAC evaluation criteria are among the role's tools.
Step 1 of 6Today, and with AI
How it is usually done today Illustrative
You start from an old ToR and edit it for days.
Evaluation questions pile up until no budget could answer them all.
Data protection rules for consultants are an afterthought.
How it works with AI
AI drafts the structure and candidate questions from your programme documents.
You cut the questions down to the few the evaluation can really answer.
You set the methods, budget and data rules yourself.
Step 2 of 6Step by step
YouGather the inputs
Theory of change, logframe, latest IPTT, the donor's evaluation requirements, budget and timeline.
AIDraft the structure and questionsClaude · ChatGPT
Ask the AI for a ToR outline and candidate questions grouped by the OECD-DAC criteria (prompt below).
YouPrioritise the questionsOECD-DAC evaluation criteria
Keep the questions decision-makers need answered at mid-term and that the budget and time allow. Cut the rest.
YouSet methods, budget and timeline
Write these yourself, from your own budget and context. Never accept days or costs the AI made up.
YouAdd data rules for the consultants
State how respondent data must be protected, and whether and how the consultants may use AI tools on it.
Your teamReview with the programme manager
Check that the questions match what managers need to decide, then finalise.
Step 3 of 6Worked example
What you give the AI Illustrative
A three-year livelihoods programme in eight districts. Main outcomes: farmers adopt improved practices; households increase income. Mid-term in month 18. The donor asks for an independent mid-term evaluation. [Theory of change and logframe attached, no personal data.]
What would you ask the AI to do with this?
Think about it for a moment, then reveal.
The task prompt (copy and adapt)
Using the programme summary, theory of change and logframe below, draft an outline for the terms of reference of a mid-term evaluation.
Include: background, purpose and use, scope, evaluation questions grouped by the OECD-DAC criteria (relevance, coherence, effectiveness, efficiency, impact, sustainability), methods, deliverables, and a section on data protection.
Propose no more than two questions per criterion. Leave budget, number of days and dates as [TO CONFIRM].
[paste programme summary, theory of change and logframe]
What comes back
Section
Draft content (excerpt)
Purpose and use
Inform mid-course decisions by programme management and the donor
Effectiveness
To what extent are farmers adopting the improved practices, and what helps or hinders adoption?
Efficiency
Are activities delivered on time and within budget across the eight districts?
Budget and days
[TO CONFIRM]
A good draft leaves gaps for you to fill. If it invents a budget or a number of days, delete them.
Step 4 of 6Check the AI output
Independence first: read the draft and note the problems you see yourself, before you run the QA prompt. Then compare.
Another draft came back like this Illustrative
Section
Draft content (excerpt)
Effectiveness
Five questions on adoption, income, gender, markets and nutrition
Methods
Randomised controlled trial in all eight districts
Budget and days
USD 45,000; 30 consultant days
Data protection
(section missing)
The QA prompt (run it on every draft)
Act as an independent evaluation reviewer.
Check the draft ToR outline below against the instructions it was given.
List: every criterion with more than two questions, every budget, number of days or date that was not provided, every method that does not fit a mid-term evaluation of this scope, and every required section that is missing.
Do not rewrite the ToR. List the problems only.
Instructions: [paste]
Draft ToR: [paste]
What should the check catch?
Think about it for a moment, then reveal.
What the check finds
Effectiveness has five questions; the limit was two.
The budget and days were invented: they should read [TO CONFIRM].
A randomised trial does not fit a mid-term evaluation of a running programme: methods are yours to set.
The data protection section is missing.
Corrected output
Section
Corrected content (excerpt)
Effectiveness
To what extent are farmers adopting the improved practices, and what helps or hinders adoption?
Methods
[SET BY THE M&E SPECIALIST]
Budget and days
[TO CONFIRM]
Data protection
How respondent data is protected, and whether consultants may use AI tools on it
Step 5 of 6Your checks
Tick each check as you do it on your own work. Each one earns +5 XP.
Ethics: the ToR states how respondent data is protected, and whether consultants may use AI on it.
Step 6 of 6Quick check
Quick check+25 XP
The AI proposes 24 evaluation questions for your mid-term ToR. What next?
Show answer
Prioritise the questions that matter for mid-term decisions and fit the budget and timeline.
Try it this week
Run the prompt above on a programme you know. Count how many of the draft questions you would keep, and write one line on why for each.
Pick a recurring task that is not covered above, such as the enumerator refresher before a post-distribution monitoring round, and use this prompt to draft your own step-by-step guide.
I am an M&E specialist. Here is a recurring task I do: [describe the task, how often, who is involved, which tools and data].
Break it into numbered steps. For each step, say who should do it (me, a colleague, an automated workflow or an AI assistant), which tool, and what could go wrong.
Mark every step where a person must check the result before it is used, and say what they should check.
Do not suggest putting respondents' personal or sensitive data into an AI tool.
End of Mission 04
MISSION05
The skills you need to develop
Skill inventory · Badge: Skill Builder
1 challenge
Mission goal: find out which of the 14 profile skills you already have and which need new learning.
Step 1 of 5Guess first
Your guess+20 XP
Which skill is the first learning priority for this role in the career profile?
Show answer
Validating AI outputs, followed by Python for data, AI-assisted qualitative analysis, Power BI and data protection.
Step 2 of 5Skills you bring with you
Tick the ones you already have. Open a skill to see how to build it.
Your skills match
0 of 14
A. Existing M&E and professional competencies
Transfers from your role
Logframes, results chains and their assumptions: the test for any causal claim an AI draft makes.
Build it: Check one AI-drafted narrative against your results chain.
Transfers from your role
SMART indicators and reference sheets. Clear definitions are what AI checks and code mappings depend on.
Build it: Turn one reference sheet into validation rules for your screening script.
Transfers from your role
M&E plans, IPTTs, data flow maps and reporting calendars: the base for automation.
Build it: Redraw your data flow map with automated steps and check points.
Transfers from your role
DQAs against validity, reliability, timeliness, precision and integrity. The same thinking applies to AI outputs.
Build it: Include one AI-assisted step in your next DQA.
Transfers from your role
Descriptive and inferential statistics. You need them to judge outliers, trends and AI-suggested patterns.
Build it: Check one AI-suggested trend with a simple calculation of your own.
Transfers from your role
KIIs, FGDs, codebooks and thematic analysis. A good codebook is what makes AI coding reliable.
Build it: Write a codebook with definitions and examples for one open question.
Transfers from your role
Question wording, skip logic and piloting. Good forms prevent the errors your scripts would otherwise catch.
Build it: Add constraints (ranges, required fields) to one KoboToolbox form.
Transfers from your role
Clear findings and actionable recommendations, now including editing AI drafts.
Build it: Edit one AI draft and list every change you made, and why.
Step 3 of 5New skills for the AI era
These are the ones most specialists still need to learn.
Your skills match
0 of 14
B. New AI competencies
Needs new learning
Checking AI results against source data, sampling for errors, keeping outputs traceable. The profile's first learning priority.
Build it: Build a traceability table for one past report (guide 3).
Needs new learning
Writing instructions, examples and output formats that give reliable results from AI assistants.
Build it: Build and test a library of five prompts for your recurring tasks.
Needs new learning
First-pass coding and theme extraction with human review. Where AI saves the most time, and where errors are easiest to miss.
Build it: Run the agreement check from guide 2 on 50 past answers.
Needs new learning
Triggers, scheduled flows and integrations that remove repetitive manual steps.
Build it: Map your partner consolidation as a workflow (guide 4).
C. Data and technical skills
Needs new learning
Data models, measures and interactive dashboards. Most organisations now expect live indicator dashboards (Power BI is a learning priority).
Build it: Build one Power BI page from your IPTT, with a plain-language note per KPI.
D. Ethics, governance and quality assurance
Needs new learning
Consent, anonymisation, DPIAs and safe handling of personal data. Sending beneficiary data to AI services raises questions you must be able to answer.
Build it: Write a one-page rule: which data may go into which AI tool, after which anonymisation.
Step 4 of 5Your five learning priorities
In this order, from the career profile:
Validating AI outputs. Every AI-supported figure in a donor report has to be traceable to source data.
Python for data. Basic scripting lets you automate data checks instead of doing them by hand.
AI-assisted qualitative analysis. Where AI saves the most time, and where coding errors are easiest to miss.
Power BI. Most organisations now expect live indicator dashboards instead of static tables.
Data protection and privacy. Sending beneficiary data to AI services raises consent and privacy questions you must be able to answer.
You don't need to become a developer. Basic pandas scripts for data checks are enough to start, and an AI assistant can write and explain them. You still test every script on data where you know the answer.
End of Mission 05
MISSION06
Interactive skills gap assessment
Know your gaps · Badge: Self-Aware Analyst
1 challenge
Mission goal: measure your readiness in four areas and find your top three gaps.
Step 1 of 3How it works
10 quick statements, about 3 minutes. Rate yourself honestly: this is a learning tool, not a test, and your answers stay in your browser.
Beginner: little or no experience
Developing: I understand it but need practice
Competent: I can apply it independently
Step 2 of 3Questions 1 to 5
Step 3 of 3Questions 6 to 10
Prefer to score it yourself?
Beginner = 0, Developing = 1, Competent = 2. Add them up (maximum 20): 0–7 Getting started · 8–14 Developing · 15–20 Ready to integrate AI. For every Beginner answer, start with the matching action:
Use an AI assistant on low-risk tasks for two weeks and keep an error log (roadmap stage 1).
Build and test a library of five prompts (roadmap stage 2).
Run the agreement check on 50 past answers (mission 4, guide 2).
Build a traceability table for one past report (mission 4, guide 3).
Ask an AI assistant to write your screening script and explain each line (mission 4, guide 1).
Build one dashboard page with a plain-language note per KPI (roadmap stage 3).
Map your partner consolidation and its exceptions (mission 4, guide 4).
Use the screening routine and query the field before correcting (mission 4, guide 1).
Agree AI data rules with your data protection focal point (roadmap stage 3).
Prepare your next indicator meeting with the routine in mission 4, guide 5.
Mission goal: commit to a realistic 12-week plan and track it week by week.
Step 1 of 5Weeks 1–2: Understand AI
About 3–4 hours a week (a planning suggestion). Tick steps as you do them: +5 XP each, saved in this browser. The mission completes when every step in all four stages is ticked.
Stage 1 · Weeks 1–20 / 5 done
Understand AI
Build AI literacy and see where AI touches your monitoring data.
Loot: an error log, a data flow map and a list of manual checks.
Step 2 of 5Weeks 3–4: Apply AI
Stage 2 · Weeks 3–40 / 4 done
Apply AI
Practise on past or anonymised data, where you already know the answers.
Loot: a tested script, an agreement score and a prompt library.
Step 3 of 5Weeks 5–8: Integrate AI
Stage 3 · Weeks 5–80 / 5 done
Integrate AI
Put one routine into regular use, with its checks.
Loot: a working routine, a dashboard, a workflow design and data rules.
Step 4 of 5Weeks 9–12: Demonstrate competence
Stage 4 · Weeks 9–120 / 5 done
Demonstrate competence
Show you can run AI-supported monitoring responsibly.
Loot: a portfolio project, one documented reporting cycle and a development plan.
Step 5 of 5Check
Mission check+20 XP
In which stage do you first test AI coding and your screening script, on data where you already know the answers?
Show answer
Stage 2 (weeks 3–4).
End of Mission 07
MISSION08
Your practical learning project
Boss level · Badge: Evidence Architect
1 challenge + 8 deliverables
Mission goal: produce a portfolio piece that proves you can run AI-supported reporting responsibly.
Step 1 of 6Your boss-level project
Boss level project
Build an AI-assisted quarterly reporting routine for one programme
Objective: show you can use AI to check, analyse and draft monitoring results, while keeping every figure traceable and every judgement your own.
Case study · Rural Resilience Programme Illustrative
No programme of your own to use? Build the routine for this one, using the examples from mission 4.
Context: use your own programme and one reporting cycle, or an anonymised copy of last quarter's data.
Resources: a KoboToolbox export, your IPTT and indicator reference sheets, one open-ended question, last quarter's report; Excel; an approved AI assistant; Power BI and Python if available; your data protection policy.
Step 2 of 6Deliverables, one by one
Deliverables checklist0 / 8 done
8 steps to a portfolio piece
The mission and the certificate unlock only when all 8 deliverables are ticked.
Step 3 of 6Templates to start from
Four ready-made files for your deliverables. Download, adapt and keep them.
Before you share it: what should a person check in your project?
Think about it for a moment, then reveal.
Every figure, against the IPTT and source data.
The agreement score, recalculated from your sample.
The data protection note, reviewed by your data protection focal point.
Any AI-generated text, reviewed and marked as AI-assisted.
No personal or confidential data left in the portfolio sample.
Step 5 of 6How it's judged
Criterion
What good looks like
Traceability
Every reported figure can be traced to source in one step.
Validation
AI outputs were checked with a sample and a level set in advance.
Judgement
Each anomaly was decided with the field team, not by the script.
Data protection
Rules name the data type, the tool and the anonymisation step.
Usefulness
Narratives give managers explanations and next steps, not just numbers.
Honesty
The memo reports problems and limits, not just time saved.
Step 6 of 6Check
Mission check+20 XP
Which deliverable proves that every figure in your narrative can be checked?
Show answer
The traceability table.
End of Mission 08
MISSION09
Responsible AI: what you must never delegate blindly
Guardrails · Badge: Responsible AI Guardian
10 challenges
Mission goal: know exactly which judgements stay with you, and the control for each AI risk.
Step 1 of 6Would you approve this report?
Click every sentence you would not confirm. There are five problems: an invented figure, an unsupported causal claim, an incorrect percentage, a missing indicator and a misleading interpretation.
Case study · Rural Resilience Programme Illustrative
Copilot drafted the Q2 donor report. Your manager asks you to confirm it today.
Source (Q2 IPTT and coding check)
Result
1.2 Farmers trained
840 against a target of 1,200 (70%)
2.1 Households with a kitchen garden
520 against a target of 900 (58%)
3.1 Savings groups formed
39 against a target of 40 (98%)
Dropout survey
1,800 answers AI-coded; 150 hand-checked, 86% agreement; cost of schooling cited in 31% of answers
Household income
No indicator
Problems found: 0 of 5+40 XP
Show answer
Five problems: the 85% (it is 58%), the invented income figure, the unsupported claim about training, the misleading reading of the 31%, and the claim that every indicator is reported (3.1 is missing).
Step 2 of 6Your decision
Your decision+20 XP
You found the problems. Can you confirm this report to your manager?
Show answer
Return it for correction. Fix each problem, recheck every figure against the IPTT in the traceability table, then confirm.
Use the traceability table on every AI-supported report:
The report is corrected. Before you confirm it, answer the five questions. Tick each one you can honestly answer yes to.
Step 4 of 6Mini-game: delegate or never?
AI can check, sort and draft. It cannot be accountable. For each task: can AI help (with checks), or must you never delegate it?
Delegate?+10 XP
Flagging duplicate IDs and impossible ages in 4,000 submissions
Show answer
AI can help. A script with an AI step does this well; you review every flag.
Delegate?+10 XP
Deciding whether an unusual quantity is an error or a real change
Show answer
Never delegate. Listed in the profile as human judgement, made with the field team.
Delegate?+10 XP
Coding 1,800 open-ended answers, with a hand-checked sample
Show answer
AI can help. With a codebook, a blind sample and an agreement check, this is a good use of AI.
Delegate?+10 XP
Choosing indicators, definitions and targets
Show answer
Never delegate. Agreed with programme teams and stakeholders, never left to a tool.
Delegate?+10 XP
Drafting the narrative for an off-track indicator
Show answer
AI can help. Fine as a first draft that you check figure by figure and complete with field explanations.
Delegate?+10 XP
Writing the recommendations a manager will act on
Show answer
Never delegate. AI can suggest options; interpreting results in context stays with you.
Delegate?+10 XP
Sending respondents' names and phone numbers to a public AI tool
Show answer
Never delegate. Protecting respondents' personal data stays with you. Anonymise first, use approved tools only.
Delegate?+10 XP
Drafting candidate evaluation questions for a ToR
Show answer
AI can help. A useful start, as long as you prioritise the questions and set the methods and budget.
Step 5 of 6Six risks, six controls
What can go wrong when AI helps with your monitoring data? Name three risks before you look.
Think about it for a moment, then reveal.
Risk
What it looks like
Your control
Hallucination
A figure or claim with no source
Traceability table; check every figure
Invented causes
A reason for under-performance nobody reported
Get explanations from the field team
Miscoding
Answers put under the wrong code, at scale
Blind hand-coded sample; agreement level
Bias
AI misreads answers from some groups or languages
Check samples across groups and languages
Privacy
Respondent data pasted into a public tool
Anonymise first; approved tools only
Silent failure
A script or workflow skips data after a form change
Count records in and out; test after every change
Step 6 of 6Tell readers how AI was used
Transparency is a control too. Add a short statement to reports that used AI, and adapt it to what you actually did.
AI tools were used to screen monitoring data for errors, to code open-ended responses and to draft parts of the indicator narrative. All figures were checked against the indicator tracking table, a random sample of [X] coded responses was checked by hand ([Y]% agreement), and the final text was reviewed by the M&E Specialist.
Agree which data may go into which tool before you start:
The AI-Enabled M&E Manager is covered in Tutorial 01, the AI-Enabled Senior M&E Specialist in Tutorial 03, the AI-Assisted M&E Officer in Tutorial 04, the AI-Powered Monitoring Specialist in Tutorial 07 and the AI-Enabled MEAL Specialist in Tutorial 13.
Step 2 of 5What you need to show
Strengths you build on
Indicator design and reference sheets
DQA experience, the basis of AI validation
Survey and qualitative methods
Donor reporting and knowing what credible results look like
What you need to show
A screening routine you built and run
AI coding checked with an agreement score
A report where every figure is traceable
A dashboard managers actually use
Data rules you follow for AI tools
Step 3 of 5Tool: build your LinkedIn headline
Pick the pieces that are true for you. Only claim what you have actually done.
Your headline
Step 4 of 5CV bullets and qualifications
CV bullet templates
Built an automated screening routine for [4,000] KoboToolbox submissions per quarter, flagging duplicates, impossible values and outliers before data reached the IPTT.
Coded [1,800] open-ended survey answers with AI against a [12]-code framework, validated with a hand-coded sample of [150] ([X]% agreement).
Introduced AI-drafted indicator narratives with a traceability table linking every reported figure to source data.
Designed an automated consolidation of [five] partners' monthly returns, saving [two days] of manual work per month.
Replace the brackets with what you have actually done.
Qualifications in the profile
Master's degree in development studies, public policy, statistics, economics or a related field.
3–6 years in programme M&E, including indicator design, data collection and donor reporting.
Useful certifications: MEAL DPro (PM4NGOs); IPDET evaluation training programme.
Step 5 of 5Check
Mission check+20 XP
What's the best way to use "AI-Enabled M&E Specialist" on your CV?
Show answer
Use it to describe your capabilities, backed by evidence, unless your employer actually uses the title.
End of Mission 10
MISSION11
Your next steps
Commit and celebrate · Badge: Transition Ready
1 challenge
Mission goal: turn what you learned into a two-week plan and unlock your certificate.
Step 1 of 2Commit to your next two weeks
Pick the actions you commit to for the next two weeks. Your plan builds itself below.
Start learning
Do this
Show this
My plan
Recommended first project: the AI-assisted quarterly reporting routine from mission 8, starting with screening your next KoboToolbox export.
Step 2 of 2Your completion certificate
Complete all 11 missions to unlock it. Add your name, take a screenshot and share your progress.
AI & M&E Career Transition Series · Tutorial 02
Your name
has completed the learning path
From Monitoring and Evaluation Specialist to AI-Enabled M&E Specialist
This is a learning-progress certificate. It records what you completed in this free tool, in your own browser; it is not independently verified and not a formal qualification. Your mission 8 project is the evidence to show an employer.
End of Mission 11
Questions and answers
What does an AI-Enabled M&E Specialist do?
Designs and runs the monitoring and evaluation system for one or more programmes, so that managers, donors and partners know whether activities are delivered and whether intended results are happening. The role exists because funding, adaptation and accountability decisions depend on credible, timely performance evidence.
How is AI changing the Monitoring and Evaluation Specialist role?
Routine data checking, consolidation and first-draft reporting move to AI tools and automated workflows, so more of the specialist's time goes to methodology, interpretation and helping managers act on findings. The specialist remains accountable for indicator definitions, analytical choices and every figure that leaves the team.
Which judgements should an AI-Enabled M&E Specialist never hand over to AI?
The career profile lists these as human judgement:
Choosing indicators, definitions and targets with programme teams and stakeholders.
Deciding whether an anomaly is a data error or a real change on the ground.
Interpreting results in context and writing recommendations managers can act on.
Protecting respondents' personal data when using AI tools.
Which skills should I learn first?
The learning priorities in the career profile, in order:
Validating AI outputs. Every AI-supported figure in a donor report has to be traceable to source data.
Python for data. Basic scripting lets the specialist automate data checks instead of doing them by hand.
AI-assisted qualitative analysis. Qualitative data is where AI saves the most time, and where coding errors are easiest to miss.
Power BI. Most organizations now expect live indicator dashboards instead of static tables.
Data protection and privacy. Sending beneficiary data to AI services raises consent and privacy questions the specialist must be able to answer.
The profile lists 14 skills in all: Results frameworks and theory of change, Indicator design, M&E planning, Data quality assessment, Statistical analysis, Qualitative methods, Survey and questionnaire design, Prompt design, Validating AI outputs, AI-assisted qualitative analysis, Dashboard development, Workflow automation, Data protection and privacy and Analytical report writing.
Which tools does this role use?
The career profile names: KoboToolbox, Excel, Python (pandas), Power BI, NVivo, ChatGPT, Claude, Microsoft 365 Copilot, n8n and OECD-DAC evaluation criteria. The tutorial shows how to check every AI output, whichever tool you use.
What qualifications and experience does the profile list?
Master's degree in development studies, public policy, statistics, economics or a related field.
3–6 years in programme M&E, including indicator design, data collection and donor reporting.
Useful certification: MEAL DPro (PM4NGOs) and IPDET evaluation training programme.
How long does the transition take?
This tutorial suggests a 12-week plan of about 3 hours a week, in four stages: understand AI (weeks 1–2), apply AI (weeks 3–4), integrate AI (weeks 5–8) and demonstrate competence (weeks 9–12). It is a planning suggestion, not a requirement.
Is "AI-Enabled M&E Specialist" an official job title?
No. It is a proposed career title from the EvalCommunity framework, not a standardised occupational classification or a guaranteed vacancy. Use it to describe your capabilities, backed by evidence, unless your employer actually uses the title.
