section 4 — target customer & market
The buying unit is typically the planning ministry's M&E or digital transformation directorate. Funding in year one is more often a multilateral or bilateral donor grant than the ministry's own discretionary budget.
| Layer | Definition | Estimate |
|---|---|---|
| TAM | Countries with a formal national development plan | ~134 countries |
| SAM | Subset with RPJMN-equivalent multi-tier structure and a mandated mid-term review | 35–45 countriesworking estimate — needs primary validation |
| SOM (3-yr) | Reachable via Indonesia case study + multilateral procurement channel | 3–6 government contractsworking estimate — needs primary validation |
Flag: SAM and SOM are working estimates, not validated figures. Before this is used externally, build an actual candidate-country list — planning structure, digital-government budget line, no existing incumbent vendor — rather than citing a round number.
section 6 — competitive landscape
Three categories of alternative exist, and the honest positioning against each is different.
McKinsey / BCG / Deloitte public-sector practices already doing parts of this analysis manually
why PETA-AI is different
Consulting is a one-off, expensive, non-continuous engagement. PETA-AI is a standing system that re-checks coherence every time a new document is published, at a fraction of the cost of repeat engagements.
Any team standing up an LLM + vector database over the same government documents
why PETA-AI is different
Can replicate Layers 1–3 in months. Cannot replicate Layer 4 without the historical outcomes dataset and the calibration discipline — that data asset is the barrier to entry, not the code.
KRISNA, One Data Indonesia and similar centralised platforms already contracted to Bappenas
why PETA-AI is different
Data-aggregation platforms, not reasoning systems — structurally unable to do program-goal-level anomaly detection. PETA-AI sits on top of this infrastructure rather than replacing it, which lowers the political barrier to adoption.
Still needed before this goes in front of an investor: a direct check for whether any startup, anywhere, already sells AI-based policy coherence detection to a national government. If one exists, name it and state the differentiation directly.