All case studies

UNDP Embeds Foresight Into a 5-Year Green Transition Programme in 6 Weeks

Two citation-backed radars and an AI assistant now feed quarterly portfolio reviews across five pillars of UNDP's Green Transition programme in Bosnia and Herzegovina.

UNDP Bosnia and Herzegovina logo
Public sectorHorizon scanTechnology scan
6 weeks
kickoff to live radars
5 models
in the scanning ensemble
5 pillars
covered in one engagement
2 radars
Green Transition and Circular Economy

The challenge

UNDP's five-year programme "Catalysing the Green Transition in Bosnia and Herzegovina" aims to enable a just shift to a green, zero-carbon economy. It runs as a portfolio across five pillars — decarbonisation, depollution, nature and biodiversity, circularity, and environmental justice — underpinned by continuous learning through Monitoring, Evaluation and Learning (MEL) and Dynamic Portfolio Management (DPM).

The team needed a horizon scanning capability that could surface signals of change across all five pillars and feed them into regular sensemaking and programming decisions, without becoming a one-off report that aged out before the next portfolio review.

What we did

Envisioning used the Signals platform to establish a lightweight, repeatable foresight-to-MEL workflow: surface signals, organise them around the portfolio's pillars, and feed the intelligence into quarterly reviews.

  • Built two purpose-built interactive radars (Green Transition and Circular Economy) containing citation-backed signals of change.
  • Applied a six-stage AI-enhanced methodology: ensemble generation across five frontier models, semantic deduplication, impact assessment, narrative synthesis, source verification, and interactive visualisation.
  • Orchestrated five AI models in parallel (Gemini 3 Pro, GPT-5 Mini, Claude Sonnet 4.5, Mistral Large, Sonar Pro) to increase signal diversity and reduce single-model bias.
  • Integrated a five-level verification system (Grounded → Fabricated) with linked external sources for every signal.
  • Delivered full downloadable datasets and an AI assistant for natural-language exploration of the signal data.
  • Ran a series of working sessions with UNDP to configure the scans and plug the workflow directly into their existing MEL and DPM processes.

What changed

  • A repeatable foresight-to-MEL workflow embedded in a five-year programme: foresight outputs feed directly into quarterly portfolio reviews and programming decisions.
  • All five portfolio pillars covered in a single scanning engagement.
  • Delivered within a six-week timeline, including model configuration, results refinement, and two stakeholder webinars.
  • Six months of hosted access to the interactive radars for ongoing use.
  • The delivery is designed to outlast the engagement: UNDP teams can reuse the datasets, run follow-up scans at set intervals, and embed the radar in portfolio reviews, closing the loop between foresight and action.
Free scan

Bring the decision your team is facing.

We will scope a scan around it and show you what the method surfaces.