A practical, hands-on programme for building Microsoft Copilot Studio agents that access organisational knowledge, complete tasks, and run reliable business workflows, shaped around Changi Airport Group's real work, not a fixed syllabus.
We don't arrive with a predefined training and read from slides. We work with the Changi Airport Group team first, to find the workflows that matter, surface where agents create real value, and build the sessions around what your people actually do.
Before we finalise the agenda, we run a short pre-survey and sense-check so the training targets the group's genuine needs, and we demonstrate real, working agentic workflows on Changi Airport Group scenarios, not generic examples.
A short survey and working conversation up front to learn the group's roles, tools, confidence levels, and where AI could genuinely help.
We sit with the Changi Airport Group team to map the processes worth automating and agree where an agent adds real, measurable value.
We showcase working agents built on Changi Airport Group's own scenarios, so participants see the method applied to their world, not a toy demo.
The 16 modules below are the full scope. We select, sequence, and deepen them based on what the sense-check tells us the group needs.
This practical training is designed for participants who already understand the basics of Microsoft Copilot and prompt engineering. We focus on how to build agents that access organisational knowledge, complete tasks, manage structured workflows, and operate reliably in a business environment.
We teach participants how to identify processes suitable for an agent and convert a broad business requirement into a clearly defined use case.
Participants define the agent's users, responsibilities, knowledge requirements, expected outputs, actions, limitations, and escalation points before starting the build.
We cover how to move beyond basic prompts and write operating instructions for an agent, defining its role, decision rules, tone, boundaries, response format, source priorities, and conditions for requesting more information.
Participants also learn how instructions influence the agent's choice of knowledge, tools, and actions.
Participants learn how to connect an agent to sources such as SharePoint, OneDrive, Dataverse, websites, and uploaded documents, and how to organise and describe those sources so the agent selects the most relevant information.
They also learn how to restrict the agent to approved organisational sources when public web information should not be used.
We teach how authentication affects what an agent and its users can access.
Participants learn how to design an agent that respects existing user permissions and prevents people from accessing information they would not normally be authorised to view.
Participants learn when to allow the agent to respond dynamically and when to use a more controlled topic or workflow, combining generative orchestration with guided questions, variables, conditions, and structured conversation paths.
This helps them avoid building agents that are either too rigid or too unpredictable.
We teach participants how to create reusable prompts for tasks such as:
Participants also learn how to control the format, length, tone, and structure of the agent's output.
Participants learn how to connect agents to tools such as Outlook, Microsoft Forms, SharePoint, Dataverse, and other services through connectors.
We show how an agent can retrieve information, create records, prepare emails, update systems, and initiate business processes.
We cover how to create workflows that combine AI steps with business rules. Participants learn how to:
Participants learn how to build agents that respond to events without waiting for a user to start a chat, activating when a form is submitted, a document is added, or a new request is received.
We also teach how to define limits, approval stages, and monitoring requirements for autonomous processes.
We teach participants how to identify tasks that should not be completed entirely by AI.
They learn how to add approval stages, human review, exception handling, and escalation rules for sensitive or uncertain situations.
Participants learn how to divide a complex solution into specialised agents, for example, a main agent coordinating separate agents for tender review, finance, customer service, or organisational policy.
We cover how to define each agent's responsibility and prevent unnecessary duplication or conflicting responses.
We teach participants how to create realistic test cases rather than relying only on informal testing, including expected questions, incomplete information, ambiguous requests, incorrect assumptions, restricted information, and unusual user behaviour.
Participants learn to test whether the agent selects the correct knowledge, uses the appropriate action, and produces an acceptable response.
Participants learn how to create repeatable evaluations using prepared conversations and test cases.
They compare agent performance before and after changes to instructions, knowledge, or tools, improving the agent systematically rather than relying on personal judgement.
We cover how to use analytics to understand agent usage, successful interactions, failures, workflow performance, and areas requiring improvement.
Participants learn to identify patterns in real user conversations and use those findings to improve instructions, knowledge sources, topics, and workflows.
Participants learn the basic governance considerations required before deploying an organisational agent, environments, permissions, authentication, data policies, approved connectors, public web access, and responsible ownership.
We also discuss how agents move through development, testing, and production environments in a controlled manner.
Participants complete the training by preparing an agent for deployment through Microsoft Teams or Microsoft 365 Copilot.
They review the agent's instructions, knowledge, permissions, workflows, test results, ownership, and ongoing maintenance requirements before publishing it.
Book a discovery session and we'll run the pre-survey, map your highest-value workflows, and tailor this curriculum to your team.