Delivery pathways

AI Agent Build and Handover

Build one governed agent for a frequent, document-heavy task, pilot it and hand it to your team.

Fixed price or quoted, 8 to 12 weeks to the end of the pilot.

On this page

Overview

AI Agent Build and Handover delivers one working agent for a task that costs hours every week, such as checking progress claims or drafting RFIs. DMS tests it against 20 to 50 real cases, pilots it with one team for six weeks and hands it to a named owner on your side. A configured agent is a fixed price, and an agent with a custom connection to your systems is quoted.

The problem

A frequent, rule-based, document-heavy task costs hours every week, such as drafting RFIs, checking progress claims, tracking contract notices or chasing expired insurance certificates. A governed agent can do the first draft and leave the decision with a person.

What you get

One working agent with its own identity and least-access permissions
An agent card setting out its purpose, scope, knowledge, access, action limits, monitoring and owner
A test set of 20 to 50 real cases and a six-week pilot report
A handover so your own owner runs the agent, with any connector it needs to reach your systems kept by you

Methodology

Each agent is built in eight steps with two approval gates.

Choose the job: frequent, rule-based and document-heavy.
Map about 20 real cases to see how the work is actually done.
Design the agent and draft its agent card. This is the first approval gate.
Prepare the data and connections so the agent can reach the systems it needs, using a controlled connector such as Model Context Protocol (MCP).
Build version 1.
Test and tune against 20 to 50 real cases. This is the second approval gate, before go-live.
Pilot with one team for six weeks.
Run and improve, with a named owner on your side.

Every agent starts by assisting a person or working under supervision, and agents that touch money, contracts, safety or decisions about people stay supervised whatever the results. Sourcing follows the ladder: switch on built-in AI first, configure in Copilot Studio or an equivalent second, and build on an API only where the process sets you apart.

Cost and timing

How it is bought: Two tiers. A configured agent, built in Copilot Studio or an equivalent on one knowledge source with no custom connector, is a fixed price to the end of the pilot. A built agent, with a custom connection to your systems through Model Context Protocol (MCP) or an application programming interface (API), is quoted to the end of the pilot. Both include the handover. Monitoring and retesting after model changes then sit under a support agreement.
How long it takes: 8 to 12 weeks to the end of the pilot.
Your team's time: An agent owner for about 2 hours a week, and one pilot team for six weeks.

Talk through your need

Tell us what your team is trying to do. We can discuss whether this service is the right starting point.

Discuss AI Agent Build and Handover