evidoria

← Back to browse

Good practice Imported

Tractable AI Damage Assessment — UNCDF's Computer-Vision Tool for Fiji's Climate Insurance Claims

Fiji · Suva · See the Fiji profile · See the Suva profile

Evidence: Descriptive / self-reported Top 83% 40/100 · Ask Evidence Copilot about this practice

UNCDF and AI firm Tractable piloted a Fiji tool letting residents photograph cyclone/flood damage for AI-verified insurance claims, aiming to cut settlement from months to days; it sits inside Fiji's PICAP scheme, which paid 1,700+ farmers after a March 2024 rainfall trigger.

535 beneficiaries
PICAP parametric payout beneficiaries (January 2023 rainfall trigger) (January 2023)
1,700+ beneficiaries
PICAP parametric payout beneficiaries (March 2024 rainfall trigger) (March 2024)
10,000+ people
Anticipatory Action pilot reach (multi-year, across four cooperatives)
Tractable AI Damage Assessment — UNCDF's Computer-Vision Tool for Fiji's Climate Insurance Claims

Details

Maturity
Pilot
Promoter
UN Capital Development Fund (UNCDF) — Pacific Insurance and Climate Adaptation Programme (PICAP), with Tractable
Period
October 2022 – ongoing (initial 12-month phase from launch)
Keywords
AI computer vision, disaster recovery, parametric insurance, climate adaptation, claims verification

Context

On 31 October 2022 in Suva, the UN Capital Development Fund (UNCDF) and applied-AI company Tractable announced a partnership bringing computer-vision damage assessment to disaster recovery in Fiji. Residents photograph storm- or flood-damaged homes on a smartphone app; Tractable's AI (already used by insurers such as Japan's MS&AD, which settled a Typhoon Mindulle claim in a single day in October 2021) assesses the damage and verifies it with partner insurers, aiming to cut claims settlement from months to days. The tool is embedded in UNCDF's Pacific Insurance and Climate Adaptation Programme (PICAP), developed with the Reserve Bank of Fiji and the Fiji Meteorological Office.

Results

PICAP's separate parametric (trigger-based) insurance product has a track record of fast, measured payouts: 535 insured beneficiaries were paid within two weeks of a January 2023 rainfall event, and a March 2024 rainfall trigger released payments to more than 1,700 beneficiaries, including two smallholder farmers who each received a historic FJ$1,000 payout; an associated Anticipatory Action pilot reached 10,000+ people across four cooperatives with payouts of up to FJ$50,000 based on windspeed forecasts. These figures relate to PICAP's parametric product line, not to the Tractable computer-vision tool itself, for which no Fiji-specific usage or accuracy results have yet been published — the AI-assessment layer remains at pilot/early-deployment stage over three years after its announcement.

Conclusions

Included as a real, credibly sourced government-linked AI pilot with an honest gap between its stated ambition and its published, tool-specific evidence.

Implementation

Indicative cost
Medium (€50k–€500k) — No published budget for the AI-assessment layer; the broader PICAP programme is multi-year and government-linked.
Time to results
Long (> 3 years) — Announced 31 October 2022 with an initial 12-month phase; described as ongoing but still pilot/early-deployment stage as of latest reporting (three years post-launch).
Staffing & skills
UN Capital Development Fund (UNCDF) partnering with Tractable (applied-AI firm) and Fiji's Reserve Bank and Meteorological Office under the PICAP programme

Conditions for success

  • Embedded within an established, multi-year, government-linked programme (PICAP) rather than a standalone tool
  • Underlying Tractable computer-vision technology already proven and deployed by insurers elsewhere (e.g. Japan's MS&AD)

Common failure modes

  • No Fiji-specific usage, accuracy or settlement-time data published for the AI tool itself over three years after announcement
  • No independent audit of the computer-vision damage assessments has been found

Where it fits

Governance type
UN agency + national regulator/insurer partnership
Scale
pilot/early deployment
Income level
lower-middle-income (Pacific SIDS)

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful