Case study · Energy · October 2026 workshop

Prime Energy Built and Tested a Handbook-Based AI Assistant

During an October workshop, Prime Energy participants worked with Microsoft Copilot to build a knowledge assistant around a sample employee handbook. They checked whether it could locate policy information, cite the source and handle difficult questions.

Handbook assistant

Built with a sample reference document

Source citations

Answer linked to the relevant handbook section

Agent testing

Exercises covered unclear, missing and out-of-scope information

Work completed during the October 2026 hands-on workshop.

Prime Energy participants during the AI agents and automation workshop, October 2026.

The problem they addressed

Policy and procedure questions come up repeatedly in large organizations. Employees need to find the relevant passage quickly, and the answer must match the approved document.

What the team worked on

Building the assistant

The exercise started with agent instructions and a sample handbook uploaded as its knowledge source. The facilitator restricted unrelated information sources. The assistant answered a policy question and cited the handbook section.

Checking accuracy and boundaries

Participants tested their assistants with different wording, missing details and attempts to override the instructions. They practiced reviewing the answers and refining the agent rules.

Applying the method elsewhere

The workshop included examples involving business requirements drafted from meeting notes, training-completion checks and structured forms. Each process relied on existing documents and clear review steps.

Takeaway for business leaders

A handbook or SOP assistant is a manageable first agent project. Give it a defined source, clear rules and a set of questions it must answer correctly before wider use.

That gives the team a clear set of documents and questions for checking each version of the assistant.

Explore AI agents workshops → Discuss a knowledge-agent pilot with Tellix →