Certified Chief Agentic AI Officer (CCAAO).

Delivery:Architecture walkthrough of a live agentic system, pattern comparison exercise and a proposal interrogation clinic.

• LO1.1 Evaluate single-agent and multi-agent designs, including orchestrator and worker, sequential pipeline and peer collaboration patterns, against a defined enterprise workflow.
  Assessment criteria:

  • Compares at least two orchestration patterns against the same enterprise workflow and states which the organization should adopt, with the coordination overhead of each made explicit.
  • Identifies where a proposed design concentrates failure risk, naming the long action chains, ambiguous handoffs or unstable external dependencies that would break first.
• LO1.2 Interrogate technical proposals for brittleness, failure modes and cost to run before approving deployment.
  Assessment criteria:

  • Puts questions to a technical proposal that expose its failure modes, recovery behavior and running cost profile, including model calls, tool execution and retries, before any approval is given.
  • Returns a proposal whose capability claims cannot be traced to evaluation evidence, stating what evidence would change the decision.
• LO1.3 Select the architecture pattern that fits the organization's risk appetite and justify the choice in business terms.
  Assessment criteria:

  • Selects an architecture pattern and defends the choice in terms of risk appetite, integration constraints and cost to run rather than technical novelty.
  • States the conditions under which the chosen pattern would no longer be appropriate and what would replace it.

Delivery:Oversight design lab, risk-tiering workshop and a reviewer staffing case discussion.

• LO2.1 Classify agent actions into risk tiers and place human approval gates where the tier demands them.
  Assessment criteria:

  • Classifies a set of agent actions by reversibility, blast radius, financial exposure and regulatory sensitivity, and assigns each to a defined risk tier.
  • Places a mandatory human approval gate on every action in the highest tier, with post-hoc review or sampled audit specified for lower tiers.
• LO2.2 Design escalation paths, review thresholds and reviewer staffing so that oversight remains substantive rather than ceremonial.
  Assessment criteria:

  • Designs escalation triggers covering confidence thresholds, value limits, novel situations and repeated failure, each routed to a named reviewer role.
  • Sizes the review function with workload limits that keep each approval a genuine judgment rather than a rubber stamp.
• LO2.3 Evaluate an oversight function in operation, using intervention and missed-catch evidence.
  Assessment criteria:

  • Reads intervention rates, missed catches and reviewer fatigue signals from oversight data and states what each implies about the health of the function.
  • Recommends a specific change to gate placement or reviewer staffing when the evidence shows oversight has become ceremonial.

Delivery:Permission model design lab, guardrail commissioning exercise and an incident simulation with a live kill-switch decision.

• LO3.1 Define least-privilege permission scopes for agents across systems, data and actions.
  Assessment criteria:

  • Specifies a permission model in which every tool an agent may invoke carries an explicit scope, escalation trigger and revocation path.
  • Assigns each agent an identity separate from any human user, with data boundaries and action ceilings tied to the task rather than the department.
• LO3.2 Commission guardrails, spend caps and kill switches, and direct adversarial testing of each before deployment.
  Assessment criteria:

  • Commissions guardrails, spend caps and a kill switch for a given deployment, stating who may trigger the switch, how quickly it takes effect and what safe state results.
  • Directs adversarial testing that covers prompt injection and tool misuse, and withholds deployment approval until the results are evidenced.
• LO3.3 Decide when to narrow or withdraw agent permissions in response to behavioral drift.
  Assessment criteria:

  • Defines the drift signals that trigger a permission review, from unexpected tool calls to spend anomalies and quality regression.
  • Chooses between narrowing scope, adding gates and suspending a drifting agent, and records the reasoning behind the choice.

Delivery:Economics case clinic, baseline measurement workshop and a sensitivity analysis exercise.

• LO4.1 Build the business case for an agentic workflow against a measured human baseline.
  Assessment criteria:

  • Builds a cost per completed task for an agentic workflow that includes model usage, tool execution, retries, rework and supervision time.
  • Measures the case against a real human baseline on agreed quality and throughput measures, not a convenient or assumed one.
• LO4.2 Identify the genuine cost drivers of agentic work, including oversight effort and orchestration overhead, and test the case's sensitivity to them.
  Assessment criteria:

  • Names the oversight staffing, orchestration overhead, integration and incident costs that headline per-task figures omit, and prices them into the case.
  • Tests the case's sensitivity to changes in model pricing, failure rates and oversight load, and states the point at which the investment no longer repays.

Delivery:Role decomposition workshop, accountability mapping exercise and a workforce transition case discussion.

• LO5.1 Redesign roles and team structures around a combined human and agent workforce, with a named human accountable for every agent's output.
  Assessment criteria:

  • Decomposes an existing role into tasks and reassigns each across humans and agents by comparative strength, with the resulting role designs written down.
  • Names a single accountable human owner for every agent's output and writes that ownership into the role design.
• LO5.2 Direct the workforce transition candidly, protecting morale, capability and skill development.
  Assessment criteria:

  • Plans a transition that sequences the change, tells affected staff the truth early and provides retraining pathways for work that moves to agents.
  • Identifies where skill atrophy will follow the move of routine work to agents and puts a countermeasure in place for each domain affected.

Delivery:Lifecycle governance walkthrough, change control case work and a decommissioning simulation.

• LO6.1 Govern agents from provisioning to decommissioning under a defined control framework, including deployment approval and registry requirements.
  Assessment criteria:

  • Sets provisioning requirements under which no agent enters production without registration, a named owner, a defined permission scope and evaluation evidence.
  • Specifies what the agent registry must record, including owner, scope, permissions, model version and evaluation history.
• LO6.2 Direct production monitoring, change control and re-evaluation when models, prompts or tools change.
  Assessment criteria:

  • Defines production monitoring covering behavioral drift, cost drift, quality regression and security events, with thresholds that trigger review.
  • Requires re-evaluation before any model swap, prompt change or tool change reaches production, rejecting silent rollover.
• LO6.3 Decide when an agent must be retired and direct an orderly decommissioning.
  Assessment criteria:

  • States the conditions under which an agent must be retired, including sustained drift, failed re-evaluation and the loss of its economic case.
  • Directs a decommissioning that revokes credentials, resolves in-flight work, meets data obligations and closes the registry record.