Using AI to promote wildlife-rich woodlands in England 

The Tree Action Plan Delivery (TAP-D) Impact Evaluation project broke new ground in using artificial intelligence for environmental policy evaluation.

Challenge

Natural England wanted to ensure tree planting in England contributes to nature recovery by promoting wildlife-rich woodlands.

A major challenge was the quality and consistency of the documentation. The evaluation team had to work with a large volume of unstructured, inconsistently formatted case files. These ranged from emails and pro formas to handwritten notes and scanned PDFs. Many documents lacked clear metadata, and terminology varied widely across sources. This made manual review slow, error-prone, and difficult to scale.

Solution

Our team deployed generative AI large language models to analyse 117 Natural England and 67 Forestry Commission case files. Using flexible data pipelines, the AI tools were trained to extract meaning from messy, unstructured content. This allowed us to identify patterns and quantify changes in application content linked to TAP-D guidance.

Flexible data pipelines are like a smart sorting system in a busy mailroom. Imagine receiving hundreds of letters and parcels: some typed, some handwritten, some in envelopes, others just loose pages. A flexible pipeline is like a team of assistants who can read, sort, and organise all of that incoming mail, no matter the format, and route it to the right place.

Results

Our use of flexible data pipelines resulted in Natural England’s increased ability to effectively analyse even messy or inconsistent documents.

The project followed Responsible AI principles to ensure transparency, data protection, and ethical safeguards. Importantly, it also showed how using gen AI can unlock value from even the most fragmented evidence base—transforming scattered case files into actionable insights and setting a precedent for future environmental evaluations.