Version 1.0
27 activities across 4 stages. A complete system for adopting agentic AI.
Processes
Visualize end-to-end workflows. Document inputs, outputs, decision points, and handoffs.
Knowledge
Map where tacit and explicit knowledge lives. Identify documentation gaps and at-risk knowledge.
Stakeholders
Identify who owns, performs, and depends on each process. Define RACI accountability.
Bottlenecks
Surface inefficiencies, repetitive tasks, and error-prone steps using value stream analysis.
SOPs
Document standard operating procedures: triggers, inputs, steps, decisions, exceptions, outputs.
Task Journal
Document daily work patterns. Compare work-as-imagined to work-as-done.
Classification
Categorize tasks by automation suitability: repetitive, judgment-heavy, creative, collaborative.
Redesign
Challenge whether the workflow itself should change, not just individual tasks.
Tools
Match tasks to tool classes: deterministic automation, AI copilot, or autonomous agent.
Autonomy
Assign human-AI collaboration level based on consequence, reversibility, and regulation.
Value
Plot opportunities on impact-vs-effort. Identify quick wins and strategic bets.
Risk
Assess hallucination exposure, data sensitivity, compliance, and failure reversibility.
Pilot
Design time-boxed proofs-of-concept with hypothesis, KPIs, and go/no-go criteria.
Knowledge Share
Promote effective workflows to team standards. Maintain shared skill libraries.
Composition
Design multi-step workflows: sequential, parallel, conditional, hierarchical patterns.
Handoffs
Define human entry/exit points. Configure approval gates and escalation triggers.
Quality
Establish validation: automated checks, human review gates, and acceptance criteria.
Context
Assemble the right context: files, data, instructions, constraints, and examples.
Direction
Direct agents using zero-shot, few-shot, or chain-of-thought patterns.
Skills
Save effective interactions as reusable, parameterized skill templates.
Policy
Define acceptable use, workflow approvals, data classification, and incident response.
Roles
Establish RACI accountability. AI is Consulted — a human is always Accountable.
Literacy
Build AI capability through structured training. Required under EU AI Act Article 4.
Ethics
Mandatory disclosure, verification obligations, bias checking, attribution honesty.
Audit Trail
Timestamped logs of AI interactions, decisions, and outputs for compliance.
Impact
Track usage, efficiency, and business outcomes. Measurement increases ROI 3×.
Improvement
PDCA for AI workflows: retrospectives, versioning, benchmarking, playbooks.
Map the work before you automate it. — orientic.ca