08
Decisions Status Board & Phased Roadmap
22 tracked decisions as of Master Architecture v1.7. All 14 original items have a direction; 8 more surfaced when the three standalone platform documents were folded in. 18 of 22 still have real, unstarted next actions — none of the 8 newest have an owner yet.
Decisions 1–14 — the September 1–2 resolution round
| # | Decision | Status | Residual |
|---|---|---|---|
| 1 | Global Risk peril list | Resolved | Communicate to PDD owner for ratification; 3 of 12 perils carry accepted near-term data-coverage risk. |
| 2 | RiskLayer partnership terms | Resolved | None — revisit only if negotiation status changes. |
| 3 | Food Security's EUDR scope | Resolved | Communicate adoption to Food Security PDD owner. |
| 4 | NRM's Protect Nature scope | Resolved | Resourcing for two simultaneous mining tracks needs confirming. |
| 5 | Climate-adjusted crop calendars | Resolved | See Decision #13 for the specific engine choice. |
| 6 | TFFF strategic engagement (NRM) | Resolved | NRM/LCEM platform lead needs to actually initiate the conversation. |
| 7 | Funding disclosure | Resolved | None — disclosure standard, not a funding search. |
| 8 | Hub-sequencing mismatch | Resolved | Food Security's own hub-sequencing mismatch hasn't been separately walked through. |
| 9 | AI build-out ownership & sequencing | Resolved | Contingent on Decision #12; no named engineering owner yet. |
| 10 | Vendor-neutrality commitment | Resolved | None. |
| 11 | Reactivate-vs-build + Mesoamerica/Central America | Resolved | David's status check on six dormant tools hasn't happened yet. |
| 12 | Coordination with SERVIR-AI/global-platform | Resolved | Not yet contacted — blocks Decision #9's agent build from starting. |
| 13 | Crop-model licensing (DSSAT vs. PCSE/WOFOST) | Resolved to a process | Bake-off hasn't been scoped or run yet. |
| 14 | Single shared gateway (SERVIR-AI as target) | Resolved | Hub-org migration inventory (§02) hasn't happened — now upstream of Decision #9's first three agents. |
Decisions 15–22 — surfaced by the v1.7 document consolidation
| # | Decision | Status | Residual |
|---|---|---|---|
| 15 | Compute cost/ownership for Global Risk's annualized engine | Open | No owner, no scoping started. |
| 16 | Direct integration talks — Pyregence/Pyrecast, HydraFloods teams | Part-resolved | Pyregence side reviewed (§04): pyretechnics is on PyPI under EPL-2.0 and co-authored inside SIG, so this is an internal adoption decision, not an external negotiation. The open question moved: not "can we use it" but who builds the non-US fuel-model layer. HydraFloods side still unreviewed, conversation not initiated. |
| 17 | Exposure-data refresh cadence & licensing (WorldPop, Open Buildings, IBAT) | Open | Not yet reviewed. |
| 18 | Regional hub governance model for locally-calibrated overrides | Open | No governance model drafted. |
| 19 | EUDR application-date reconfirmation | Open | Not yet reconfirmed against official EU source. |
| 20 | Crop-type coverage prioritization for servir-aces | Open | No prioritization made — largest build effort alongside #13. |
| 21 | Chain-of-custody/traceability standard (logistics) | Open | No specialist conversation scoped yet. |
| 22 | User-tiering for NRM Decision Support layer | Open | Still open internally per the PDD itself. |
| 23 | Intervention-prioritisation layer — adopt Planscape/ForSys, or leave optimisation to each use case (§06 N7) | New | Raised 10 Sep 2026. Planscape is CC0 and SIG-built, so adoption is unblocked technically. The cost is the regional input layer — the same cost as the non-US fuel-model layer in Decision #16. Scope them together. |
Two decisions gate everything else. Decision #12 (contact SERVIR-AI/global-platform's maintainers) and Decision #14's residual (map which hub-org repos migrate into SERVIR-AI) are both upstream blockers — no domain agent in §04–06 should start building against the gateway until both move.
Four-phase roadmap
Per the GeoAI Strategy document — carried here in full rather than referenced, since it's the sequencing this entire engineering plan builds toward.
| Phase | Window | Focus |
|---|---|---|
| I — Platform Development | 2026 Underway | Stand up the gateway and shared capability layer; wrap the first three agents (Flood Risk, Field & Crop-Type, Land Cover & Change) per Decision #9; resolve the SERVIR-AI migration mapping. |
| II — Case-Study Expansion | 2027–28 | Extend agents into the remaining domain capabilities per §04–06's build tables; run the DSSAT/PCSE bake-off (Decision #13); prioritize servir-aces crop coverage (Decision #20). |
| III — Generalization & Platform Integration | 2028–29 | Cross-platform Agent2Agent federation matures; shared infrastructure (§07) fully consolidated; regional hub overrides governed (Decision #18). |
| IV — Institutionalization & Sustainability | 2029–30 | Tiered capacity-building model (§03) reaches research-lead cohorts across hubs; compute/funding model (Decision #15) settled for long-run operation. |