Model Online / On-Site
Duration
Modules 5 modules

Course Objective & Scope

Allowing agents to act with full autonomy on irreversible actions—such as issuing payments, deleting customer records, sending contracts or modifying production infrastructure—creates unacceptable enterprise risk. This course teaches how to embed human-in-the-loop (HITL) controls at critical decision points without sacrificing automation speed on routine tasks.

Hands-on labs cover risk-tiered approval gates, LangGraph interrupt() and MCP/OpenAI approval primitives, editing tool arguments before resuming, SLA-based escalation chains and immutable audit trails. Participants build an agent with asynchronous approval routing adapted to an internal workflow.

Who Should Attend?

  • Software and AI engineers building transactional enterprise agents
  • Solution architects translating compliance, risk and internal control rules into system design
  • Technical product teams automating operations centres and customer support workflows
  • Prerequisite: Python/TypeScript and basic agent or workflow development experience

Key Takeaways

  • Classify tools and decisions into risk tiers: auto-execute, notify, require approval or block
  • Suspend and serialise agent state cleanly so waiting for human input consumes zero compute
  • Implement approve, edit-and-resume and reject-with-feedback flows
  • Configure SLA timeouts, multi-tier escalation chains and role-based approval routing
  • Capture tamper-evident audit logs recording who approved what action and with what context

Curriculum

Risk Tiers and Approval Policy Design
  • Classifying read-only, reversible and irreversible tool calls
  • Static policy thresholds (monetary limits, permissions) versus dynamic triggers (uncertainty, sentiment)
  • Designing thresholds that prevent operator approval fatigue
Interrupts, State Suspension and Resumption
  • Asynchronous suspension with LangGraph interrupt() and checkpointers
  • Handling the three human outcomes: Approve, Edit arguments and Reject with feedback
  • Managing partial approvals across parallel tool calls
Escalation Channels and Operator Experience
  • Dispatching interactive approval cards via webhooks to Slack, Microsoft Teams or custom UIs
  • Presenting concise evidence summaries and action previews (diffs) to reviewers
  • SLA timeouts, escalation to backup approvers and full human conversation takeover
Authorisation, Security and Audit Trails
  • Verifying approver identity and enforcing maker-checker separation of duties
  • Preventing prompt injection from spoofing approval state
  • Structuring audit logs for EU AI Act human-oversight compliance
Workshop: Transactional Agent with Approval & Escalation
  • Coding an enterprise agent that pauses for manager sign-off above policy thresholds
  • End-to-end testing of approve, edit, reject and timeout escalation paths

Other Courses in This Category

  • ReAct Pattern for LLM Agents: Reasoning and Acting

    Hands-on training for developers on building ReAct agents from scratch: the Thought–Action–Observation loop, native tool calling, stop conditions, iteration limits and failure analysis.

    AI Agent Engineering Corporate
    Live Course
    5 modules
    View
  • Plan-and-Execute Agents: Planning, Execution and Replanning

    Technical course on agents that split multi-step tasks into planner, executor and replanner roles, covering ReWOO and LLMCompiler-style parallel task graphs, plan validation and cost optimisation.

    AI Agent Engineering Corporate
    Live Course
    5 modules
    View
  • State and Workflow Management for AI Agents

    Course for teams building long-running agents: state schemas, reducers, checkpoints, thread management and state-machine workflows, illustrated with LangGraph and workflow engine examples.

    AI Agent Engineering Corporate
    Live Course
    5 modules
    View
  • Tool Design and Tool Contracts for LLM Agents

    Hands-on course on designing tools agents can use reliably: naming, descriptions, JSON Schema parameters, error messages, tool contracts, versioning and permission scopes.

    AI Agent Engineering Corporate
    Live Course
    5 modules
    View
  • Building Agent Systems with LangGraph

    Hands-on technical course on building production-ready agent systems with LangGraph: StateGraph, conditional edges, checkpointers, interrupt-based human-in-the-loop, subgraphs and the Send API.

    AI Agent Engineering Corporate
    Live Course
    6 modules
    View
  • OpenAI Agents SDK and Tool-Based Agent Systems

    Course on building tool-based agents with the OpenAI Agents SDK and Responses API: function tools, hosted tools such as web search and file search, handoffs, guardrails, sessions and built-in tracing.

    AI Agent Engineering Corporate
    Live Course
    5 modules
    View

Get in touch