evidoria

← Back to browse

Good practice Imported

OpenEval — Finland's AI-Assisted Search Platform for Development Evaluation Evidence

Finland · Helsinki · See the Finland profile · See the Helsinki profile

Evidence: Descriptive / self-reported Top 66% 53/100 · Ask Evidence Copilot about this practice

Finland's Ministry for Foreign Affairs launched OpenEval, an Azure-based AI tool built with CGI that lets civil servants and the public search, summarise and cross-reference hundreds of development-cooperation evaluation reports at once, supporting evidence-based policy design.

~9 months
Development time from proof-of-concept to launch

Details

Promoter
Ministry for Foreign Affairs of Finland (Development Evaluation Unit), with CGI
Period
June 2025–present
Keywords
development cooperation evaluation, AI-assisted policy research, semantic search, natural language processing, evidence-based decision-making

Context

OpenEval is an AI-assisted search platform built by Finland's Ministry for Foreign Affairs (Development Evaluation Unit) with technology partner CGI, letting civil servants and the public search and synthesize hundreds of pages of development-cooperation evaluation reports in English and Finnish. It uses Azure Document Intelligence, Azure AI Search and GPT-4o, and launched on 18 June 2025 after roughly nine months of development from proof-of-concept, including three rounds of internal and external feedback testing.

Objectives

Support evidence-based development-policy design through AI-assisted semantic search and synthesis of evaluation evidence, while keeping information-use decisions a human responsibility rather than delegating them to the AI.

Activities

Development of a semantic search and synthesis platform on Azure (Document Intelligence, Azure AI Search, GPT-4o) covering bilingual (English/Finnish) evaluation reports; three rounds of internal and external user testing; public launch on 18 June 2025 under project lead Nea-Mari Heinonen and Under-Secretary of State Pasi Hellman.

Results

The tool has been operational since its 18 June 2025 launch and generated significant international interest from the UN, multilateral development banks, the OECD and EU networks, with several organizations reportedly piloting similar approaches. However, no quantified before/after usage metrics have been published. The practice scored 53/100 in the catalogue's evaluation, with 'Evidence of Impact' rated only 1/3 for lacking quantified metrics.

Conclusions

OpenEval demonstrates a replicable model for AI-assisted access to evaluation evidence with an explicit human-oversight principle, but its own evidence base is limited — quantified usage or impact metrics have not been published, transferability beyond the pilot ministry is not yet documented, and scalability remains rated low (1/3) pending planned iterative expansion.

Implementation
Indicative cost
Medium (€50k–€500k) — Built on Microsoft Azure Document Intelligence, Azure AI Search and GPT-4o with vendor partner CGI over roughly nine months from proof-of-concept to launch — a mid-scale managed-cloud AI platform rather than a low-cost internal script or a very large national infrastructure programme.
Time to results
Medium (1–3 years) — Approximately nine months from proof-of-concept to launch (three rounds of testing), officially launched 18 June 2025 and operational since; no fixed end date reported.
Staffing & skills
Ministry for Foreign Affairs of Finland, Development Evaluation Unit, CGI (technology vendor/implementation partner), Project lead Nea-Mari Heinonen

Conditions for success

  • Three rounds of internal and external user feedback before release
  • Human-oversight design principle embedded (AI supports, does not replace, civil servants' judgment)
  • Bilingual (English/Finnish) semantic search built into the platform from the start

Common failure modes

  • No quantified before/after usage metrics have been published to demonstrate impact
  • Currently limited to a single ministry's evaluation reports, with scalability rated low (1/3) pending iterative expansion

Where it fits

Governance type
national ministry unit
Scale
single ministry, with reported international interest from UN/MDBs/OECD/EU
Income level
high income

Commonly funded by

National / regional programmes Digital Europe Programme

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

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.

Similar practices you may find useful