Supporting Citizen-Service Teams with AI Answers from Official Sources

Image
case study

An AI assistant designed to support helpline staff with source-cited information, currently undergoing evaluation.

Relevance Digital has developed an AI assistant for a citizen-service helpline. Staff can ask questions in natural language and receive answers drawn from official sources, with citations that allow them to inspect the supporting information.

The objective

Support staff with information they can verify

Citizen-service enquiries can depend on specific procedures, conditions and dates. Staff need to find relevant official guidance and assess whether it applies to the question they are handling.

The objective is to make that information easier to access and review, while keeping the person handling the enquiry responsible for how it is used.
 

The solution

Answers grounded in official sources

The assistant is designed to draw its answers exclusively from approved official websites and documents. It retrieves relevant information and includes citations so staff can check the material behind a response.

The source collection can be updated as official guidance changes. Citations support verification, evaluation assesses whether answers accurately reflect that material and address the question.

An assist-only approach

The assistant is intended for helpline staff and provides information for human review. It does not operate as an autonomous citizen-facing service or make decisions about eligibility or entitlements.

Evaluation informed by human feedback

Feedback collected during evaluation helps identify responses that need further investigation. Reported issues can be revisited, and proposed changes can be tested and reviewed before release.

Evaluation considers whether responses answer the question, are supported by the cited sources and use information relevant to the enquiry’s date.

EU-hosted deployment

The solution is hosted in the European Union. Hosting location forms part of the deployment approach, alongside the restriction to official source material and the assist-only design.
 

Current status

Evaluation before operational deployment

The project is in an advanced evaluation and proof-of-concept phase. Live rollout has not yet taken place and depends on the organisation’s planning cycles.

No measured operational outcomes are available for publication at this stage. Improvements in response time, answer accuracy or staff workload remain to be established.