A processor on a circuit board
← All workshops

Course title

Design and Orchestration of Autonomous AI Agents

Agent architecture, tool integration and multi-agent orchestration

Overview

Agentic AI has progressed from research to enterprise deployment, with agents that plan tasks, invoke tools, and delegate work to one another. This two-day programme examines how to design, connect, evaluate, and govern agents that can be trusted in production environments.

An agent differs fundamentally from a conversational assistant. It plans, selects tools, acts, evaluates the result, and continues, which means that architecture, permissions, and evaluation matter far more than prompt wording.

The programme moves from agent design patterns to integration with enterprise systems through the Model Context Protocol, then to multi-agent coordination, evaluation, and security. Participants develop a working agent progressively across both days.

Learning outcomes

By the end of the workshop, you will be able to

  1. 01

    Apply agentic design patterns, including planning, tool use, reflection, and routing.

  2. 02

    Integrate agents with enterprise systems through the Model Context Protocol under scoped permissions.

  3. 03

    Orchestrate multi-agent systems with defined roles, hand-offs, and supervisory controls.

  4. 04

    Evaluate agent performance and defend deployments against prompt injection and misuse.

Programme

What you will learn

Part 1 · 2 modules

Agent design

1.1What makes a system agentic
  1. (a)From prompts to plans
  2. (b)Reasoning and acting in a loop
  3. (c)When an agent is the wrong answer
1.2Design patterns
  1. (a)Tool use and function calling
  2. (b)Reflection and self-checking
  3. (c)Routing and human-in-the-loop control

Part 2 · 3 modules

Connecting and orchestrating

2.1The Model Context Protocol
  1. (a)MCP servers, clients and tools
  2. (b)Exposing internal systems to agents
  3. (c)Permissions and scoped access
2.2Memory and context
  1. (a)Context engineering
  2. (b)Short-term and long-term memory
  3. (c)Retrieval as a tool
2.3Multi-agent systems
  1. (a)Frameworks such as LangGraph and CrewAI
  2. (b)Agent-to-agent protocols
  3. (c)Supervisors, specialists and hand-offs

Part 3 · 2 modules

Evaluation and security

3.1Evaluating agents
  1. (a)Building an evaluation set
  2. (b)Tracing and observability
  3. (c)Measuring cost, speed and accuracy
3.2Securing agents
  1. (a)Prompt injection and tool impersonation
  2. (b)Least-privilege tool access
  3. (c)Audit trails for agent actions

Who should attend

  • Software developers and AI engineers
  • Solution and enterprise architects
  • Technical product managers
  • IT leads responsible for AI adoption

In-house delivery

Tailored for your organisation.

Delivered at your offices or conducted virtually, each run is customised around your active course portfolio and internal operating procedures.

Book a consultation All workshops