Start from a template
Select one to pre-fill the form, then customise
1
Interpret — Role & Scope
Define the AI’s expert identity and analysis scope
2
Retrieve — Extract from Your Data
AI retrieves only what you provide — the anti-hallucination safeguard
SRS principle: AI extracts only what exists here. Missing data → “NO DATA: [indicator]”
0 / 50,000
Do not paste confidential or personally identifiable data. Use anonymised programme information only. Processed client-side.
3
Generate Analysis
Define analytical operations on retrieved data only
4
Generate Conclusions & Output Format
Define output structure and mandate explicit limitation statements
SRS always includes a LIMITATION clause to flag unsupported conclusions.
16 structured generators — one for every phase of the M&E cycle. Select a phase, complete the fields, generate and copy.
M
Select a phase above
Choose an M&E phase to load its prompt generator
300+ expert-curated prompts. Select a category, then a subcategory to explore.
Select a category above
Choose a category and subcategory to explore expert prompts for that M&E function.
300+
Prompts in library
16
M&E phases covered
4
SRS steps
4
AI platforms
SRS vs. basic prompting
Basic Prompt
“Write a Theory of Change for a health programme.”
- Generic, non-contextual output
- AI may fabricate data
- No assumption testing
- No limitations flagged
SRS Prompt
“Retrieve only from context. Map pathways. Mark testable assumptions. State LIMITATION where data is absent.”
- Grounded in your programme data
- Retrieval-constrained — no fabrication
- Structured causal analysis
- Transparent limitations
Real-world impact
Mid-sized INGO · 8 MEAL staff · 4 countries · 6-month SRS adoption
Before
9 hrs
per lessons learned report
After SRS
25 min
per report · 8.7/10 quality
Annual saving
~$40-70k
67% fewer donor feedback loops
Practitioner Guide — PDF
Full SRS guide with worked examples, templates, and field notes.
01
Step 1
Interpret — Role & Scope
Assign a specific expert identity. Constrains the AI to specialist output, eliminating generic responses.
Role: Senior M&E specialist, ToC development, fragile contexts.
Scope: Girls’ education, rural Kenya. Analysis: Pathway mapping.
Scope: Girls’ education, rural Kenya. Analysis: Pathway mapping.
02
Step 2
Retrieve — From Context Only
“Do not add” is the anti-hallucination safeguard. Absent data returns “NO DATA” — not an invented value.
Extract only: (a) explicit activities, (b) outputs, (c) outcomes.
Do not add outcomes not explicitly listed.
Do not add outcomes not explicitly listed.
03
Step 3
Generate Analysis
Calculate changes, identify trends, group findings — from retrieved data only. No external knowledge added.
Map causal pathways. For each: identify assumptions.
Mark testable vs. untestable. Flag logic gaps.
Mark testable vs. untestable. Flag logic gaps.
04
Step 4
Generate Conclusions
Mandate explicit limitations. “LIMITATION: [missing info]” separates SRS from ordinary prompting.
Present as: (1) narrative, (2) pathway, (3) assumptions table.
LIMITATION: state all unsupported conclusions.
LIMITATION: state all unsupported conclusions.
The core SRS principle
SRS separates what the AI knows from what your data says. Step 2 (Retrieve) is the critical gate: by prohibiting additions and mandating “NO DATA” flags, you eliminate confabulation and produce outputs traceable to your actual programme evidence.
Full example — Theory of Change prompt
