AI orchestration and agents
AI training programme · 8 modules · 4 Skilledge Hyper labs
Agents get the most hype and the most cancellations. We teach teams to build the ones that survive.
Gartner expects cost, unclear value and weak risk controls to sink over 40% of agentic AI projects by the end of 2027. The ground has settled enough to teach properly: MCP and AGENTS.md now sit under the Linux Foundation's Agentic AI Foundation. So we teach the patterns first and the framework second. Two tracks: builders (engineers, architects) and no-code (operations and business teams).
At a glance
- Who it’s for
- Engineers and architects (builder track); operations and business teams (no-code track).
- Tracks
- Builder and no-code tracks
Modules
- When not to build an agentPlenty of "agent" projects are a plain workflow with one AI step. Telling the two apart saves months.
- Tool use and MCPConnecting an agent to your systems with the narrowest access that still does the job.
- Retrieval on company knowledgeHow RAG works, and the ways it quietly returns the wrong answer.
- Multi-agent patternsHandoffs, supervisors, parallel workers, and when one agent is enough.
- Human in the loopApproval steps, and deciding exactly where autonomy stops.
- Guardrails and permissionsLeast privilege, prompt injection, and what happens when an agent reads a hostile document.
- Evals and tracingTesting an agent before release and watching it after.
- Running it in productionCost per task, monitoring, and who gets the call when it misbehaves.
Skilledge Hyper lab exercises
Lab designs: what a cohort does in Skilledge Hyper, on mock data.
- Refund agentOn mock order and ticket data, build an agent that reads a complaint, looks up the order, drafts a decision and escalates above a set value. Then try to break it with an injected instruction.
- Agent on a budgetEvery run draws from the learner's Hyper wallet. Get the same result for half the cost.
- Write the eval setThirty test cases, a pass mark, and a scored report on your own agent.
- Mock first, real laterPrototype an agent for your own workflow on practice data, then leave with a hand-over plan for production.
Tools covered
LangGraph, CrewAI, OpenAI Agents SDK, Claude Agent SDK, Google ADK, Microsoft Agent Framework, Copilot Studio, and n8n or Zapier on the no-code track.
Sources
Related programmes
- Enterprise tools with AIAI training
- AI for software engineersAI training
- AI for QA and testingAI training
Want this for your team? Tell us who it’s for and what they need to do differently after it.