← Back to UK Department for Education tests Copilot Studio agents to help delivery officers answer policy guidance questions in a 3-month experiment

Source proof for UK Department for Education tests Copilot Studio agents to help delivery officers answer policy guidance questions in a 3-month experiment

Source-bound proof

Verified source excerpts for every supported field

Each colour maps a published value to the exact source passage used to support it. Only bounded excerpts are public; administrators can inspect the complete captured source.

13 fields supported

Title

UK Department for Education tests Copilot Studio agents to help delivery officers answer policy guidance questions in a 3-month experiment

derived · high
…t a binary; there’s times where an assistant is useful, and times when it isn’t My most recent project was a 3 month experiment at the Department for Education (DfE) where we were looking at the role generative AI might play in supporting internal delivery officers. A multidisciplinary team of dxw and DfE staff worked together to understand if…
…s, we developed our first prototypes. We gave ourselves a technical constraint: our assistant was to be an agent, built in Copilot Studio. This meant we could move fast, but were restricted by the features of the softw…

Description

dxw and DfE staff built and tested two Copilot Studio agents in Teams that answer questions from regional officers using central-team guidance documents, evaluated through moderated sessions and a fortnight-long diary study. Source concludes agents help in some situations and not others; no quantified results.

derived · high
…o the team, in a fortnight-long unmoderated diary study. Learning and iterating By the end of the project, we’d designed, built and tested 2 agents. Both answered queries on the basis of a set of knowledge, documents and intrane…
…er, it was perfectly adequate for prototyping, and for some sorts of wider use. Importantly, it got us an assistant in Teams very quickly, and Teams is where our users spend much of their time. Throughout the project, we deepened our knowledge of the world in which the cen…
…ld in which the central team operated. Our experimental design was in 2 stages. First we ran moderated sessions where officers commented on the answers given by a prototype assistant, using it for historical queries. Then we improved the assistant, and asked officers to use it on queries as they…
…lp expert team members respond accurately to questions from generalists faster. The answer wasn’t a binary; there’s times where an assistant is useful, and times when it isn’t. This has deepened our real-world understanding of when and why an agent can hel…

Company

Department for Education (UK) / dxw

derived · high
…at the role generative AI might play in supporting internal delivery officers. A multidisciplinary team of dxw and DfE staff worked together to understand if an AI assistant would help expert team members respond accurately to questions from generalists faster. The answer wasn’t a binary; there’s times where an assistant is useful, and tim…

Country

United Kingdom

derived · medium
…t a binary; there’s times where an assistant is useful, and times when it isn’t My most recent project was a 3 month experiment at the Department for Education (DfE) where we were looking at the role generative AI might play in supporting intern…

Region

Europe

derived · medium
…t a binary; there’s times where an assistant is useful, and times when it isn’t My most recent project was a 3 month experiment at the Department for Education (DfE) where we were looking at the role generative AI might play in supporting intern…

Industry

Government & Public Sector

classification · high
…evelop a testable hypothesis design and build the experiment evaluate and learn The Department identified a use case which would make a real impact; a central team is responsible for guidance in delivering policy. The team writes documentation, and spends a substantial amount of time answerin…

Problem

A central DfE team writes policy guidance and spends substantial time answering questions from regional officers; the guidance is extensive and traditional search struggles to find answers.

derived · high
…a real impact; a central team is responsible for guidance in delivering policy. The team writes documentation, and spends a substantial amount of time answering questions from regional officers. Members of the team each have deep expertise in a specific subject area, and qu…
…wer them. Some of these queries are answered in the guidance documentation, but there is a substantial amount of it, and traditional search isn’t up to the job of finding answers. Some answers may depend on nuance or local context which isn’t in the guidance…

Solution

In a 3-month experiment, dxw and DfE staff designed, built and tested two agents in Copilot Studio, delivered in Teams, that answer queries from configured documents and intranet pages using CRISPE-framework instructions. They were evaluated in moderated sessions on historical queries and a fortnight-long unmoderated diary study.

derived · high
…iterating By the end of the project, we’d designed, built and tested 2 agents. Both answered queries on the basis of a set of knowledge, documents and intranet pages we’d identified and configured. They followed a set of instructions based on the CRISPE framework to define the…
…set of knowledge, documents and intranet pages we’d identified and configured. They followed a set of instructions based on the CRISPE framework to define the agent’s Capacity and Role, the Insight it needed, the Statement t…
…assistant, using it for historical queries. Then we improved the assistant, and asked officers to use it on queries as they came into the team, in a fortnight-long unmoderated diary study. Learning and iterating By the end of the project, we’d designed, built and test…

Business value

This was an experiment, not a production rollout. The source concludes an assistant is useful in some situations and not others, and that the time saving is the time a human would spend combing the guidance; no figures are given. The next step is to roll out what worked.

derived · high
…the detail. The human is still the decision maker, this AI is their researcher. The time saving is the length of time a human would need to spend combing the guidance for the specific detail they needed. In other cases, we found an AI couldn’t grasp the national guidance in sufficie…
…e. I’d like to thank DfE for giving us the opportunity to follow this approach. The next step is to roll out what worked, and scope another experiment. Next Unseen architecture: Leadership’s role in creating space for service…

AI capabilities

Conversational AI, Retrieval-Augmented Generation

classification · high
…iterating By the end of the project, we’d designed, built and tested 2 agents. Both answered queries on the basis of a set of knowledge, documents and intranet pages we’d identified and configured. They followed a set of instructions based on the CRISPE framework to define the…

Technology

Copilot Studio, Microsoft Teams

classification · high
…er, it was perfectly adequate for prototyping, and for some sorts of wider use. Importantly, it got us an assistant in Teams very quickly, and Teams is where our users spend much of their time. Throughout the project, we deepened our knowledge of the world in which the cen…
…s, we developed our first prototypes. We gave ourselves a technical constraint: our assistant was to be an agent, built in Copilot Studio. This meant we could move fast, but were restricted by the features of the softw…

Headline outcome

derived · high
…lp expert team members respond accurately to questions from generalists faster. The answer wasn’t a binary; there’s times where an assistant is useful, and times when it isn’t. This has deepened our real-world understanding of when and why an agent can hel…

Use case type

Knowledge search

classification · high
…iterating By the end of the project, we’d designed, built and tested 2 agents. Both answered queries on the basis of a set of knowledge, documents and intranet pages we’d identified and configured. They followed a set of instructions based on the CRISPE framework to define the…
Capture details
Captured
20 Nov 2026, 06:03 UTC
Extractor
fetch-strip@1
Snapshot hash
e7bbfcedbf6f1054070e7fce264d92882a33e545c2bcfcb464eea3ea4f6fc18f