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 performance 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
Programme Performance Specialist
AI-enhanced role
AI-Driven Programme Performance Specialist
Level
Specialist · 3–6 years
How the role uses AI
Integrates AI: builds forecasts and automated data flows around AI, and is accountable for the quality of the results
"AI-Driven Programme Performance 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
Forecasts that took days of spreadsheet work can now refresh automatically, with AI suggesting drivers of change. What becomes your job?
Show answer
In the profile, the specialist's job shifts to building sound forecasting logic, testing it against past performance, communicating uncertainty and making sure forecasts lead to realistic corrective actions.
Step 4 of 5AI-Enabled, not AI-Dependent
The goal is not to hand your work to AI. It is to become a stronger performance specialist who uses AI and stays in charge of it. Here are the five questions, asked about your own work:
1QualityDo AI-refreshed forecasts make your performance analysis more reliable, or only faster?Every forecast is back-tested against past years and shown as a range, not a single figure.
2AccountabilityWhen a forecast is wrong, can you explain why, and who acted on it?Methods, assumptions and past accuracy are written down, and you sign off every risk brief.
3IndependenceDo you still find the real reasons behind a trend?AI suggests drivers; you check them against procurement records and with programme teams.
4EthicsDo your risk briefs treat partners fairly and keep the people the programme serves in view?No speculative cause blames a team or partner, partner data stay in approved tools, and actions are checked for their effect on participants.
5Professional developmentAre you getting better at forecasting, or only at running a model?You read the code you run, understand overfitting and validation, and learn from every back-test.
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 forecast performance risks with AI, test the forecasts and turn them into realistic corrective actions, 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 · Water and Livelihoods Portfolio Illustrative
A fictional portfolio you will follow through missions 2, 3, 4, 8 and 9.
The portfolio
Six grants for water, farming and training in three districts, with nine outcome targets and 24 output indicators.
The data
Three years of monthly indicator data, the indicator tracking table, monthly finance data, the procurement log and partner reports.
The tools
Python (pandas), Jupyter, Excel, Power BI, Power Automate, Microsoft 365 Copilot, Claude and ActivityInfo.
Your objective
Show managers, in time to act, which targets are at risk, why, and what can realistically be done, with the uncertainty stated.
Step 2 of 4Flip your tasks: today and with AI
Less spreadsheet work, more testing and explaining. 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
Refreshing forecasts with new monthly data
Show answer
AI-assisted. An automated pipeline can do it, once the forecasting logic is sound.
Sort it+10 XP
Drafting plain-language risk summaries
Show answer
AI-assisted. With every figure and cause checked before it is shared.
Sort it+10 XP
Combining indicator, budget and procurement data
Show answer
AI-assisted. An automated flow, checked each month.
Sort it+10 XP
Back-testing forecasts against previous years
Show answer
New skill. Predictive modelling is the profile's first learning priority.
Sort it+10 XP
Spotting overfitting before it makes forecasts overconfident
Show answer
New skill. Machine learning fundamentals are a learning priority in the profile.
Sort it+10 XP
Deciding whether a forecast is reliable enough to act on
Show answer
Human. Listed in the profile as human judgement.
Sort it+10 XP
Recommending corrective actions programme teams can carry out
Show answer
Human. Listed in the profile as human judgement.
Sort it+10 XP
Communicating risk honestly without causing unnecessary alarm
Show answer
Human. Listed in the profile as human judgement.
Step 4 of 4Your turn: two minutes
Think about your last quarterly review. Which 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 three checks every AI-supported forecast needs, and practise spotting problems in an AI draft.
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
Writing forecasting code
Suggesting drivers of a change
Drafting risk summaries and briefs
Explaining a method in plain words
Can be partly automated
Monthly forecast refreshes
Combining finance, indicator and procurement data
Dashboards with forecast ranges
Flags where spending runs ahead of outputs
New responsibilities
Back-testing forecasts on past years
Showing ranges, not single figures
Checking AI explanations against records
Watching for overfitting
Stays with you
Deciding if a forecast is reliable enough to act on
Understanding the operational reasons behind a trend
Recommending feasible corrective actions
Communicating risk honestly
Step 2 of 4The three checks you must always build in
A forecast that refreshes itself can be wrong every month without anyone noticing. Here are the checks, one at a time.
Test it on the past. Back-test every forecast against previous years before anyone acts on it.
Check causes against records. An AI-suggested driver is a question until procurement records or the programme team confirm it.
Show the uncertainty. Ranges, not single figures, in plain words, without unnecessary alarm.
Step 3 of 4Mini-game: spot the red flags
Case study · Water and Livelihoods Portfolio Illustrative
Microsoft 365 Copilot drafted a note on a grant where spending runs ahead of outputs. Here is the data and the draft.
Click every sentence you would not send.
Grant G-04, end of September
Budget spent
70%
Outputs delivered
45%
Procurement log
Advance payment for pipes in June: 20% of the budget
Construction
Permits approved in August, three months late
Red flags found: 0 of 3+30 XP
Show answer
Three sentences should not go out: "the gap shows that money is being misused" (the procurement log shows an advance payment), "construction will catch up by December" (nothing supports it), and "the partner's management is weak" (a speculative cause that blames the partner).
Step 4 of 4Mission check
Your call+20 XP
The forecast says a target will be missed. What do you do before it goes into a brief?
Show answer
Check the range and how accurate the model was on the past, then agree the causes with the team.
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 forecast performance risks with AI, with a human check at every step.
Case study · Water and Livelihoods Portfolio Illustrative
Every worked example below uses this portfolio, so you can follow one set of forecasts from the model to the quarterly review.
Six real tasks from programme performance work. 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
Forecast year-end target achievement, with ranges
From the profile: a Python model uses three years of monthly delivery data for 24 indicators to forecast year-end achievement with ranges; the specialist reviews forecasts that flag risk and discusses them with programme managers; updating a forecast showing that two of nine outcome targets are likely to be missed.
Step 1 of 6Today, and with AI
How it is usually done today Illustrative
Each forecast takes days in a spreadsheet.
It shows one figure, treated as certain.
Seasonality is ignored, or guessed.
How it works with AI
A Python model forecasts every indicator, with a range.
You check the code against your rule, and test it on past years.
You review every flagged forecast with the programme manager.
Step 2 of 6Step by step
YouWrite the forecasting rule
For example: achieved to date, plus the same remaining months in the past three years, as a range.
AIAsk AI to write the scriptClaude · Python (pandas)
It reads the monthly data and applies your rule (prompt below).
YouRead and test the codeJupyter
Run it on last year's data. Does it match your rule and the actuals?
AutomatedRefresh the forecast every monthPower BI
With the new monthly data.
YouReview every flagged forecast
Is the range below the target? Is the data complete?
Your teamDiscuss with programme managers
What is behind it, and what can still change?
Step 3 of 6Worked example
What you give the AI Illustrative
Indicator: farmers trained
Value
Year-end target
4,000
Achieved, January to September
2,600
Trained October to December, past years
2023: 600 · 2024: 900 · 2025: 750
What would you ask the AI to do with this?
Think about it for a moment, then reveal.
The task prompt (copy and adapt)
Write a Python (pandas) script that forecasts year-end achievement for each indicator.
Rule: achieved to date, plus what was delivered in the same remaining months in each of the past three years. Report the lowest, middle and highest result as a range, and compare them with the target.
List indicators with missing months separately. Do not fill missing values.
Columns: indicator, year, month, value, target.
What comes back
Forecast
Farmers trained
% of target
Low
3,200
80%
Middle
3,350
84%
High
3,500
88%
The whole range is below the target: likely to be missed. You review it with the programme manager before it goes in a brief.
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.
Act as an independent reviewer of a forecasting script.
Compare the code below with the forecasting rule.
List: every line where the calculation differs from the rule, every place the code changes or drops data silently (such as missing months), every place a range is replaced by a single figure, and a small test table, with the right answer, that would show each problem.
Do not rewrite the script. List the problems only.
Rule: [paste]
Code: [paste]
What should the check catch?
Think about it for a moment, then reveal.
What the check finds
fillna(0) counts missing months as zero delivery, which pulls forecasts down. Missing months should be listed, not filled.
The rate uses this year's average month, so it ignores seasonality. The rule uses the same months in the past three years.
rate * 12 adds twelve months, not the three months left in the year.
There is one figure, no range.
Corrected output
# past = October to December in 2023, 2024 and 2025, by indicator
rem = past.groupby(['indicator', 'year'])['value'].sum()
low, mid, high = rem.groupby('indicator').agg(['min', 'median', 'max']).T.values
Forecast = achieved to date plus low, middle and high. Missing months are listed separately. Tested on last year's data before use (guide 2).
Step 5 of 6Your checks
Tick each check as you do it on your own work. Each one earns +5 XP.
Accountability: you decide whether a forecast is reliable enough to act on. The model gives a range; it doesn't decide.
Step 6 of 6Quick check
Quick check+25 XP
The script fills missing months with zero. What happens?
Show answer
List missing months separately. Never let a data gap look like a delivery gap.
Try it this week
Pick five indicators. Forecast their year-end achievement with a range, using the same months in past years, and compare each range with the target.
From the profile: the specialist re-runs last year's forecasts against actual results for 30 indicators, finds that forecasts for training outputs were consistently optimistic and adjusts the model; back-testing forecasts against previous years to measure their accuracy.
Step 1 of 6Today, and with AI
How it is usually done today Illustrative
Nobody checks how accurate past forecasts were.
Managers trust or distrust forecasts on instinct.
The same errors repeat every year.
How it works with AI
You re-run last year's forecasts on last year's data.
You compare them with what actually happened, by type of indicator.
You adjust the model where it was wrong, and say how accurate it is.
Step 2 of 6Step by step
YouRe-run last year's forecastsPython (pandas) · Jupyter
Using only the data available at the time.
YouCompare with the actual resultsExcel
For each indicator: the error, and whether the actual fell inside the range.
AIAsk AI to summarise by type of indicatorClaude
Where were forecasts too high or too low (prompt below)?
YouCheck the summary against the table
Averages can hide one group's errors.
YouAdjust the modelJupyter
Only where the back-test shows a problem. Then test again.
Your teamShare the accuracy with managers
In plain words: how often the range held.
Step 3 of 6Worked example
What you give the AI Illustrative
Type of indicator
Indicators
Average error (direction)
Actual inside the range
Training outputs
8
+18% (too high)
2 of 8
Infrastructure
10
−3%
8 of 10
Services
12
+2%
10 of 12
What would you ask the AI to do with this?
Think about it for a moment, then reveal.
The task prompt (copy and adapt)
Summarise this forecast back-test for managers.
For each type of indicator, say: how far forecasts were from the actual results on average, in which direction, and how often the actual fell inside the forecast range.
Name any type where forecasts were consistently too high or too low.
Do not suggest causes or changes to the model.
[paste back-test table]
What comes back
Training output forecasts were consistently too high: 18% on average, and the actual fell inside the range for only 2 of 8 indicators. Infrastructure and service forecasts were close: the actual fell inside the range for 18 of 22.
The training forecasts need a different method. That decision, and the new test, are yours.
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
Back-test summary (AI): The model is accurate: the average error across 30 indicators is 4%. No changes are needed.
The QA prompt (run it on every draft)
Act as an independent reviewer of a forecast back-test summary.
Compare the summary with the back-test table.
List: every average that hides a group of indicators with large errors, every error averaged so that over- and under-forecasts cancel out, every result compared with the target instead of the actual, and every conclusion the table does not support.
Do not rewrite the summary. List the problems only.
Back-test table: [paste]
Summary: [paste]
What should the check catch?
Think about it for a moment, then reveal.
What the check finds
The 4% average hides training outputs, which were 18% too optimistic.
Over- and under-forecasts cancel out in a signed average. Use the average size of the error, ignoring its direction.
For training outputs, the actual fell inside the forecast range for only 2 of 8 indicators.
"No changes are needed" is not supported: the training forecasts need a different method.
Corrected output
Training output forecasts were 18% too optimistic on average, and the actual fell inside the range for only 2 of 8 indicators. The model now uses the completion rate of past training groups for these indicators. Infrastructure and service forecasts held up: the actual fell inside the range for 18 of 22 indicators.
Step 5 of 6Your checks
Tick each check as you do it on your own work. Each one earns +5 XP.
Professional development: you test the model on past years yourself, and learn where it fails, so your judgement keeps growing.
Step 6 of 6Quick check
Quick check+25 XP
The average error across all indicators is small, but training forecasts were 18% too high. What do you do?
Show answer
Fix the method for the group that was wrong, test again, and report accuracy by type.
Try it this week
Take last year's forecast for ten indicators. Compare each with the actual result, and note how often the actual fell inside the range.
From the profile: an automated flow combines finance and IPTT data monthly; Copilot drafts a note on grants where spending runs ahead of outputs, which the specialist checks against procurement records; comparing burn rate and output delivery for a grant that has spent 70% of its budget but delivered 45% of its outputs.
Step 1 of 6Today, and with AI
How it is usually done today Illustrative
Finance and indicator data are compared by hand, once a quarter.
A gap between spending and delivery is taken as a problem.
Procurement records are not checked.
How it works with AI
An automated flow combines finance and indicator data every month.
Copilot drafts a note on the grants with gaps.
You check every gap against procurement records before anyone acts.
Step 2 of 6Step by step
AutomatedCombine finance and indicator data monthlyPower Automate · Excel
Budget spent and outputs delivered, by grant.
AIAsk Copilot to draft a noteMicrosoft 365 Copilot
Facts only, and questions for procurement (prompt below).
YouCheck every gap against the procurement logExcel
Advance payments, late invoices, delayed deliveries.
YouRemove causes the records don't show
And words like "overspending" the data don't support.
Your teamAsk the programme team
What is behind the gap, and what is planned?
YouUpdate the note and the forecast
Confirmed causes only.
Step 3 of 6Worked example
What you give the AI Illustrative
Grant
Budget spent
Outputs delivered
Gap (points)
G-02
52%
55%
−3
G-04
70%
45%
25
G-07
30%
50%
−20
What would you ask the AI to do with this?
Think about it for a moment, then reveal.
The task prompt (copy and adapt)
From the table below, draft a short note on grants where the share of budget spent and the share of outputs delivered differ by more than 15 points.
For each, state the figures and list questions to check in the procurement log and with the programme team.
Do not suggest causes or decisions.
[paste table]
What comes back
G-04: 70% spent, 45% delivered (25 points). Check: advance payments, deliveries not yet used, construction schedule.
G-07: 30% spent, 50% delivered (20 points the other way). Check: unpaid invoices for delivered outputs.
Two grants with questions, no causes yet. The procurement log answers 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
Copilot note: G-04 is overspending and should be put on hold. G-07 is performing well: it has delivered more than it has spent.
The QA prompt (run it on every draft)
Act as an independent reviewer of a note comparing spending and delivery by grant.
Compare the note with the finance data, the indicator data and the procurement log.
List: every word such as "overspending" the data does not support, every cause or decision the note states that the records do not show, every grant where the gap could have another explanation in the procurement log, and every grant flagged or cleared without a check.
Do not rewrite the note. List the problems only.
Data and procurement log: [attach]
Note: [paste]
What should the check catch?
Think about it for a moment, then reveal.
What the check finds
G-04 has spent 70% of its budget, within the budget: "overspending" is wrong.
The procurement log shows a June advance payment for pipes (20% of the budget), which explains much of the gap.
"Put on hold" is a decision for managers, not a finding.
G-07 spending behind delivery may mean unpaid invoices: it needs a check, not a "performing well".
Corrected output
G-04: 70% spent, 45% of outputs delivered. Much of the gap is a June advance payment for pipes (procurement log); construction started late because permits came in August. To discuss with the programme manager: the revised construction schedule.
G-07: 30% spent, 50% delivered. To check with finance: are invoices for delivered outputs still unpaid?
Step 5 of 6Your checks
Tick each check as you do it on your own work. Each one earns +5 XP.
Independence: a gap between spending and delivery is a question. Procurement records and the programme team answer it, not the AI note.
Step 6 of 6Quick check
Quick check+25 XP
A grant has spent 70% of its budget but delivered 45% of its outputs. What does it show?
Show answer
A question. The procurement log and the programme team give the answer.
Try it this week
Take your grants' latest finance and indicator data. Find the largest gap between spending and delivery, and check it against the procurement log.
From the profile: checking an AI-suggested explanation for a forecast change against the procurement log; using AI to suggest drivers of performance changes and verifying them with programme teams.
Step 1 of 6Today, and with AI
How it is usually done today Illustrative
A forecast change is explained from memory.
The first plausible reason goes into the brief.
Data problems are mistaken for delivery problems.
How it works with AI
AI suggests possible drivers, each with a record to check.
You check the data and the records first.
You confirm the cause with the programme team before reporting it.
Step 2 of 6Step by step
YouNote what changedPower BI
Which forecast, by how much, since when.
AIAsk AI for possible driversClaude
Each with the record that would confirm or rule it out (prompt below).
YouCheck the data firstPython (pandas)
Late or missing data, changed definitions, a model change.
YouCheck the recordsActivityInfo · Excel
Procurement log, partner reports, activity plans.
Your teamConfirm with the programme team
Name who confirmed it, and when.
YouReport confirmed causes only
Say what is still unknown.
Step 3 of 6Worked example
What you give the AI Illustrative
Forecast change: water connections by December fell from 88% to 71% of target (range 64–78%) between the August and September updates.
What would you ask the AI to do with this?
Think about it for a moment, then reveal.
The task prompt (copy and adapt)
The forecast for an indicator changed between two monthly updates. Details below.
List possible drivers, including data or model problems. For each, name the record that would confirm or rule it out (for example the procurement log, partner reports or the data entry log).
Do not choose a driver and do not write a conclusion.
[paste the forecast change, indicator definition and recent data]
What comes back
Data: late or missing September data for a district. Check: the data entry log.
Supplies: delayed pipe delivery. Check: the procurement log.
Plan: activities moved to next year. Check: the revised workplan.
Model: a change in the forecasting code. Check: the code history.
Four drivers to check. None of them is a cause yet.
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
AI explanation: The forecast for water connections fell from 88% to 71% because community demand dropped after the rainy season.
The QA prompt (run it on every draft)
Act as an independent reviewer of an AI explanation of a forecast change.
Compare the explanation with the data, the procurement log and the notes from the programme team.
List: every cause the records do not show, every alternative explanation in the records that was not checked, and every change in the data or the model itself that could explain the change.
Do not rewrite the explanation. List the problems only.
Records: [attach]
Explanation: [paste]
What should the check catch?
Think about it for a moment, then reveal.
What the check finds
Nothing in the data or partner reports measures community demand.
The procurement log shows the pipe delivery delayed by six weeks, and it was not checked.
September's figures for one district were entered late: part of the fall could be a data gap, not a delivery gap.
The explanation was not confirmed with the programme team.
Corrected output
The forecast for water connections fell from 88% to 71% of target (range 64–78%). Confirmed causes: pipe delivery delayed by six weeks (procurement log; confirmed by the programme manager on 3 October). September data for District 2 arrived late and are now included; this explains about 3 points of the fall.
Step 5 of 6Your checks
Tick each check as you do it on your own work. Each one earns +5 XP.
Accountability: every driver in a brief is confirmed by a record or a named programme team member before it is reported.
Step 6 of 6Quick check
Quick check+25 XP
The AI explains a fall with a cause none of your data measures. What do you do?
Show answer
Treat it as a question. Check the records, confirm with the team, and report only what is confirmed.
Try it this week
Take the largest change in your last forecast update. Ask an AI assistant for possible drivers with records to check, then check the first two.
Draft a corrective action brief for a high-risk programme
From the profile: Claude drafts a brief for each high-risk programme from the forecast and partner reports; the specialist removes speculative causes and adds actions agreed with programme teams; meeting a programme manager to agree corrective actions for delayed construction activities.
Step 1 of 6Today, and with AI
How it is usually done today Illustrative
Briefs are written from scratch before each review.
Causes are guessed, and sometimes blame partners.
Actions are recommended without asking who will carry them out.
How it works with AI
Claude drafts the brief from the forecast and partner reports.
You remove every cause the records don't confirm.
You add the actions agreed with the programme team, with owners.
Step 2 of 6Step by step
YouCollect the inputsPower BI
The forecast with its range, confirmed causes, partner reports.
AIAsk Claude to draft the briefClaude
Risk, confirmed causes, questions for the meeting (prompt below).
YouRemove speculative causes
Anything the records don't show, and anything that blames without evidence.
Your teamAgree actions with the programme manager
Feasible this year, with an owner and a date.
YouAdd the agreed actions
And when the forecast will show whether they work.
YouTrack the actionsExcel
Update the forecast next month to see the effect.
Step 3 of 6Worked example
What you give the AI Illustrative
Forecast: water connections in G-04, 64–78% of target by December. Confirmed causes: permits approved three months late; pipe delivery delayed by six weeks. Partner report: two contractor teams available in November.
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 a one-page corrective action brief from the forecast and the confirmed causes below.
State the risk with its range, the confirmed causes with their sources, and three questions to agree actions with the programme manager.
Do not add causes that are not listed. Do not recommend actions; the actions will be agreed in the meeting.
[paste forecast and confirmed causes]
What comes back
Risk: water connections forecast at 64–78% of target by December. Causes: late permits; pipe delivery six weeks late (procurement log).
Questions for the meeting: Can a second contractor team start in November? Which districts first? When will we know if it is working?
The actions come from the meeting, not from the AI.
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
Corrective action brief (AI): Construction in G-04 is delayed because the partner lacks capacity. Recommended action: replace the partner. Construction will be completed by December.
The QA prompt (run it on every draft)
Act as an independent reviewer of a corrective action brief.
Compare the brief with the forecast, the partner reports and the notes of the meeting with the programme team.
List: every cause the records do not show, every action that was not agreed with the people who must carry it out, every date or figure with no source, and every sentence that blames a person or partner without evidence.
Do not rewrite the brief. List the problems only.
Forecast, reports and meeting notes: [attach]
Brief: [paste]
What should the check catch?
Think about it for a moment, then reveal.
What the check finds
"The partner lacks capacity" is a speculative cause that blames the partner. The records show late permits and a late pipe delivery.
"Replace the partner" was not agreed with anyone, and is not feasible this year.
"Completed by December" has no source: the forecast range is below target.
The actions agreed in the meeting with the programme manager are missing.
Corrected output
Risk: water connections in G-04 are forecast at 64–78% of target by December. Causes (confirmed): permits approved three months late; pipe delivery delayed by six weeks.
Actions agreed with the programme manager on 3 October: a second contractor team in Districts 2 and 3 from November (owner: programme manager); weekly delivery check (owner: performance specialist). The forecast will be updated in November to see whether the actions are working.
Step 5 of 6Your checks
Tick each check as you do it on your own work. Each one earns +5 XP.
Ethics: no speculative cause blames a partner or team, and actions are agreed with the people who must carry them out.
Step 6 of 6Quick check
Quick check+25 XP
The AI draft says the delay is because "the partner lacks capacity". Nothing in the records shows it. What do you do?
Show answer
Remove it. Only confirmed causes go into a brief.
Try it this week
Take one high-risk indicator. Draft a brief with AI from the forecast and confirmed causes only, then agree two actions with the programme manager.
From the profile: preparing three slides for the quarterly performance review showing risk levels by programme; showing forecast ranges rather than single figures in dashboards and briefs; generating plain-language performance risk summaries and editing them for accuracy; explaining forecast methods, assumptions and uncertainty to managers in plain language.
Step 1 of 6Today, and with AI
How it is usually done today Illustrative
Slides show single figures, treated as certain.
Risk levels are coloured without explanation.
Managers don't know how far to trust the forecast.
How it works with AI
Dashboards show forecast ranges and risk levels by programme.
AI drafts plain-language summaries; you edit them for accuracy.
You explain the uncertainty and the agreed actions, without alarm.
Step 2 of 6Step by step
AutomatedUpdate the dashboardPower BI
Forecast ranges and risk levels by programme.
YouSet the risk levels
For example: on track, at risk (target inside the range), likely to miss (whole range below).
AIAsk AI for a plain-language summaryClaude · Microsoft 365 Copilot
From the forecast table, with ranges (prompt below).
YouEdit for accuracy and tone
Every figure against the table; no alarm words.
YouAdd the agreed actions
What is being done about each risk.
Your teamPresent and answer questions
Including how accurate past forecasts were (guide 2).
Step 3 of 6Worked example
What you give the AI Illustrative
Programme B target
Forecast range (% of target)
Risk level
Water connections
64–78%
Likely to miss
Latrines
80–90%
Likely to miss
Irrigation plots
92–108%
At risk
Three other targets
100% or more
On track
What would you ask the AI to do with this?
Think about it for a moment, then reveal.
The task prompt (copy and adapt)
Write a plain-language summary of this forecast table for one slide.
For each risk level, say how many targets, name those at risk or likely to be missed, and give their forecast ranges.
Use calm, factual words. Do not add causes or actions; I will add the agreed actions.
[paste forecast table]
What comes back
Programme B: 3 of 6 targets on track; 1 at risk (irrigation plots, 92–108% of target); 2 likely to be missed (water connections, 64–78%; latrines, 80–90%).
Next: add the actions agreed with the programme team, and how accurate past forecasts were.
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
Slide text (AI): Programme B will fail its targets. RED ALERT. Water connections: 71%.
The QA prompt (run it on every draft)
Act as an independent reviewer of a slide summarising performance risk.
Compare the text with the forecast table.
List: every single figure shown where the forecast gives a range, every word that overstates the risk or causes unnecessary alarm, every risk level that does not match the table, and every agreed action that is missing.
Do not rewrite the slide. List the problems only.
Forecast table: [paste]
Slide text: [paste]
What should the check catch?
Think about it for a moment, then reveal.
What the check finds
"Will fail its targets": two of six targets are likely to be missed, one is at risk and three are on track.
"RED ALERT" causes alarm without saying what is being done.
71% is shown as a single figure; the forecast is 64–78%.
The actions agreed in October are missing.
Corrected output
Programme B: 3 of 6 targets on track; 1 at risk (irrigation plots, 92–108% of target); 2 likely to be missed: water connections (64–78%) and latrines (80–90%). Actions agreed in October: a second contractor team from November; the November forecast will show whether it is working.
Step 5 of 6Your checks
Tick each check as you do it on your own work. Each one earns +5 XP.
Quality: risk is shown with ranges and plain words, honestly and without unnecessary alarm.
Step 6 of 6Quick check
Quick check+25 XP
A manager asks for "just one number" instead of the range. What do you do?
Show answer
Give the middle figure with its range, and say how reliable past forecasts were.
Try it this week
Take your next performance slide. Replace every single forecast figure with a range, and add one sentence on how accurate past forecasts were.
Pick a recurring task that is not covered above, such as supporting a target revision or tracking whether corrective actions work, and use this prompt to draft your own step-by-step guide.
I am a programme performance specialist. Here is a task I do regularly: [describe the task, when, who is involved, which tools and data].
Break it into numbered steps. For each step, say who should do it (me, the programme team, finance, an automated tool 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 personal data or confidential partner data into an AI tool that is not approved.
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 13 skills for this role 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
Predictive modelling, followed by Python, machine learning fundamentals, communicating evidence and Power BI.
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 13
A. Existing performance competencies
Transfers from your role
Tracking indicators against targets and the workplan.
Build it: Map one programme's targets to the forecast you will build.
Transfers from your role
The base you need to read a forecast and its errors.
Build it: Calculate the average size of last year's forecast errors.
Transfers from your role
Knowing what each indicator counts, so a change in definition isn't read as a trend.
Build it: Check the definitions of five indicators you forecast.
Transfers from your role
Linking outputs to outcomes, so forecasts make sense together.
Build it: Check that output and outcome forecasts in one programme tell the same story.
Transfers from your role
Turning forecasts into corrective actions teams can carry out.
Build it: Track whether one corrective action changed the forecast.
Transfers from your role
Now including uncertainty: managers need to understand it to decide well (a learning priority).
Build it: Explain one forecast range to a manager in two sentences.
Transfers from your role
Reading spending against delivery, and the procurement behind it.
Build it: Compare spending and delivery for one grant (guide 3).
Step 3 of 5New skills for the AI era
These are the ones most performance specialists still need to learn, or to apply in a new way.
Your skills match
0 of 13
B. New AI and data competencies
Needs new learning
Forecasts are the core output, and their limits must be well understood. The profile's first learning priority.
Build it: Forecast five indicators with ranges (guide 1).
Needs new learning
Reproducible forecasting code is easier to test and update than spreadsheet models (a learning priority).
Build it: Read and test one AI-written forecasting script.
Needs new learning
Knowing about overfitting and validation prevents overconfident predictions (a learning priority).
Build it: Back-test one model on a year it was not built on (guide 2).
Needs new learning
Combining indicator, budget and procurement data automatically.
Build it: Set up a monthly flow for one grant's finance and indicator data.
Needs new learning
Showing ranges and risk levels, not single figures.
Build it: Add a forecast range to one Power BI visual.
Needs new learning
Checking AI code, explanations and drafts against the data and records.
Build it: Check one AI explanation against the procurement log (guide 4).
Step 4 of 5Your five learning priorities
In this order, from the career profile:
Predictive modelling and forecasting. Forecasts are the core output, and their limits must be well understood.
Python for data. Reproducible forecasting code is easier to test and update than spreadsheet models.
Machine learning fundamentals. Knowing about overfitting and validation prevents overconfident predictions.
Communicating evidence. Managers need to understand uncertainty in forecasts to make good decisions.
Power BI. Forecasts and risk ratings are usually shared through dashboards.
The tools change; the method doesn't. Whichever tools you use, write the rule first, test the forecast on the past, check causes against records, and show the range.
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 run AI-driven forecasting. For every Beginner answer, start with the matching action:
Use the routine in mission 4, guide 1.
Run one AI-written script on last year's data and compare.
Use the routine in mission 4, guide 2.
Read a short introduction to validation and overfitting.
Set up one monthly flow for a single grant.
Use the routine in mission 4, guide 3.
Use the routine in mission 4, guide 4.
Use the routine in mission 4, guide 5.
Use the routine in mission 4, guide 6.
Explain one forecast range to a manager in two sentences.
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 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
Write down how you forecast today and learn what AI-written models get wrong.
Loot: written forecasting rules, an error log and the data you may use.
Step 2 of 5Weeks 3–4: Apply AI
Stage 2 · Weeks 3–40 / 4 done
Apply AI
Try four performance tasks with AI, with checks.
Loot: a tested forecast, a back-test and one checked gap.
Step 3 of 5Weeks 5–8: Integrate AI
Stage 3 · Weeks 5–80 / 5 done
Integrate AI
Use AI across a full forecasting cycle, with checks.
Loot: a monthly forecast cycle with checks at each step.
Step 4 of 5Weeks 9–12: Demonstrate competence
Stage 4 · Weeks 9–120 / 5 done
Demonstrate competence
Show the approach is reliable and others can follow it.
Loot: a portfolio project, a methods note and a development plan.
Step 5 of 5Check
Mission check+20 XP
In which stage do you first back-test a forecast on last year's data?
Show answer
Stage 2 (weeks 3–4).
End of Mission 07
MISSION08
Your practical learning project
Boss level · Badge: Forecast Tester
1 challenge + 8 deliverables
Mission goal: produce a portfolio piece that proves you can forecast with AI responsibly.
Step 1 of 6Your boss-level project
Boss level project
Build a tested forecast for one programme, and turn it into agreed actions
Objective: show you can forecast with AI, test the forecast on the past, check causes against records and communicate risk honestly.
Case study · Water and Livelihoods Portfolio Illustrative
No live programme to use? Work through this portfolio, using the examples from mission 4.
Context: one programme or grant, with at least two years of monthly indicator data. Use data you are allowed to use, in approved tools.
Resources: indicator data and targets, finance data, the procurement log, partner reports, Python or Excel, Power BI and an approved AI assistant.
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.
The code, against the written rule.
The back-test, by type of indicator.
Every cause, against a record or a named team member.
Every figure on the slide, against the forecast table.
No confidential partner data in an unapproved tool.
Step 5 of 6How it's judged
Criterion
What good looks like
Sound logic
A written rule, and code that matches it.
Tested
Back-tested, with accuracy by type of indicator.
Honest ranges
Ranges everywhere, never a single certain figure.
Confirmed causes
Every cause checked against a record.
Feasible actions
Agreed with the people who carry them out.
Transparent
Methods, assumptions and AI use written down.
Step 6 of 6Check
Mission check+20 XP
Which deliverable shows managers how far they can trust your forecast?
Show answer
The back-test.
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 send. There are five problems: an invented figure, an unsupported causal claim, an incorrect percentage, a missing indicator and a misleading interpretation.
Case study · Water and Livelihoods Portfolio Illustrative
An AI assistant drafted the summary for the quarterly performance review from the forecast table. The review is tomorrow.
Source (forecast table, nine outcome targets)
Result
On track
5
At risk (target inside the range)
2: irrigation plots, clinic visits
Likely to be missed (whole range below target)
2: water connections (64–78%), latrines (80–90%)
Back-test, training outputs
Forecasts 18% too high last year; method changed
Model accuracy figure
Not calculated for this year
Problems found: 0 of 5+40 XP
Show answer
Five problems: the 85% (7 of 9 is 78%), "exactly 71%" (the forecast is 64–78%), the invented 95% accuracy, the unsupported claim about partners' management, and the claim that every risk is covered (the two targets at risk are missing).
Step 2 of 6Your decision
Your decision+20 XP
You found the problems. Can the summary go to the quarterly review?
Show answer
Correct it first. Fix each problem, recheck every figure, then send it.
Step 3 of 6Five questions before you send it
The summary is corrected. Before you send it, answer the five questions. Tick each one you can honestly answer yes to.
Step 4 of 6Mini-game: delegate or never?
AI can refresh, suggest, combine and draft. It cannot be accountable. For each task: can AI help (with checks), or must you never delegate it?
Delegate?+10 XP
Refreshing forecasts with new monthly data
Show answer
AI can help. Once the logic is tested, with flags reviewed by you.
Delegate?+10 XP
Deciding whether a forecast is reliable enough to act on
Show answer
Never delegate. Listed in the profile as human judgement.
Delegate?+10 XP
Suggesting possible drivers of a change
Show answer
AI can help. With every driver checked against a record.
Delegate?+10 XP
Understanding the operational reasons behind a trend
Show answer
Never delegate. Listed in the profile as human judgement.
Delegate?+10 XP
Drafting a plain-language risk summary
Show answer
AI can help. With every figure checked against the forecast table.
Delegate?+10 XP
Recommending corrective actions programme teams can carry out
Show answer
Never delegate. Listed in the profile as human judgement.
Delegate?+10 XP
Combining finance, indicator and procurement data
Show answer
AI can help. In an automated flow, checked each month.
Delegate?+10 XP
Communicating risk honestly without causing unnecessary alarm
Show answer
Never delegate. Listed in the profile as human judgement.
Step 5 of 6Six risks, six controls
What can go wrong when AI enters your forecasting? Name three risks before you look.
Think about it for a moment, then reveal.
Risk
What it looks like
Your control
Untested forecasts
A model nobody checked on past years
A back-test by type of indicator
Overfitting
A model that fits the past perfectly and predicts badly
Test on years the model was not built on
False certainty
A single figure treated as a promise
Ranges in every dashboard and brief
Speculative causes
"The partner lacks capacity" with no evidence
Causes checked against records and with teams
Data gaps read as delivery gaps
Missing months counted as zero
List missing data separately
Hidden methods
Managers can't tell how a forecast was made
A methods note, including AI use
Step 6 of 6Tell readers how AI was used
Transparency is a control too. Add a short note to the performance review pack, and adapt it to what was actually done.
AI tools supported this performance review. Forecasts were produced by a Python model written with an AI assistant; the code was checked against the forecasting rules and back-tested on last year's results ([x] of [y] actual results fell inside the forecast ranges). An automated flow combined finance and indicator data; gaps were checked against the procurement log. AI suggested possible drivers of forecast changes; only causes confirmed by records or programme teams are reported. An AI assistant drafted this summary, and [name] checked every figure against the forecast table.
The AI-Enabled M&E Manager is covered in Tutorial 01, the AI-Enabled Senior M&E Specialist in Tutorial 03, the AI-Powered Monitoring Specialist in Tutorial 07, the AI-Enabled Results-Based Management Specialist in Tutorial 11 and the AI-Powered Performance Measurement Specialist in Tutorial 12.
Step 2 of 5What you need to show
Strengths you build on
Tracking indicators against targets and budgets
Results frameworks and indicator definitions
Working with programme teams on delays
Quarterly performance reviews
What you need to show
Forecasts with ranges, from code you tested
Back-tests with accuracy by type of indicator
Spending gaps checked against procurement records
Causes confirmed before they are reported
Corrective actions agreed with programme teams
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 a Python forecast of year-end achievement for [24] indicators, with ranges, and reviewed every flagged forecast with programme managers.
Back-tested forecasts against actual results for [30] indicators, found [training] forecasts [18]% too optimistic and adjusted the model.
Set up a monthly flow combining finance and indicator data, and checked spending gaps against procurement records.
Drafted corrective action briefs with AI support, removing unconfirmed causes and adding actions agreed with programme teams.
Replace the brackets with what you have actually done.
Qualifications in the profile
Master's degree in economics, statistics, public policy, development studies or a related quantitative field.
3–6 years in programme performance analysis, M&E or grants management, including quantitative analysis.
Useful certification: Microsoft Power BI Data Analyst (PL-300).
Step 5 of 5Check
Mission check+20 XP
What's the best way to use "AI-Driven Programme Performance 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 forecast from mission 8, starting with a written rule for five indicators.
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 10
Your name
has completed the learning path
From Programme Performance Specialist to AI-Driven Programme Performance 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-Driven Programme Performance Specialist do?
Analyses whether programmes are on track to meet their targets and budgets, and forecasts where they are likely to fall short so managers can correct course in time. Uses predictive analytics on indicator trends, delivery rates and spending to identify performance risks and support corrective action.
How is AI changing the Programme Performance Specialist role?
Forecasting and trend analysis that once took days of spreadsheet work can now be refreshed automatically, with AI suggesting likely drivers of change. The specialist's job shifts to building sound forecasting logic, testing it against past performance, communicating uncertainty and making sure forecasts lead to realistic corrective actions.
Which judgements should an AI-Driven Programme Performance Specialist never hand over to AI?
The career profile lists these as human judgement:
Deciding whether a forecast is reliable enough to act on.
Understanding the operational reasons behind performance trends.
Recommending corrective actions that are feasible for programme teams.
Communicating risk honestly without causing unnecessary alarm.
Which skills should I learn first?
The learning priorities in the career profile, in order:
Predictive modelling and forecasting. Forecasts are the core output, and their limits must be well understood.
Python for data. Reproducible forecasting code is easier to test and update than spreadsheet models.
Machine learning fundamentals. Knowing about overfitting and validation prevents overconfident predictions.
Communicating evidence. Managers need to understand uncertainty in forecasts to make good decisions.
Power BI. Forecasts and risk ratings are usually shared through dashboards.
The profile lists 12 skills in all: Performance and results-based management, Predictive modelling and forecasting, Statistical analysis, Indicator design, Results frameworks and theory of change, Python for data, Data pipelines and integration, Dashboard development, Validating AI outputs, Adaptive management, Communicating evidence and Budget and resource management.
Which tools does this role use?
The career profile names: Python (pandas), Jupyter, Excel, Power BI, Power Automate, Microsoft 365 Copilot, Claude and ActivityInfo. 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 economics, statistics, public policy, development studies or a related quantitative field.
3–6 years in programme performance analysis, M&E or grants management, including quantitative analysis.
Useful certification: Microsoft Power BI Data Analyst (PL-300).
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-Driven Programme Performance 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.
