AIHA Office · PiR2-IT

Agentic AI & Hyperautomation Implementation Office

A working model of the office that takes agentic AI and hyperautomation from idea to production in a large organisation: executive dashboard, governance, guardrails, risk, full reference architecture for cloud, hybrid, on-prem and air-gapped, agentic workflows with human-in-the-loop, hallucination control, domain AI engines, costs, V&V, e-learning and lessons learned.

Inside the demo

Executive dashboard — Portfolio of 20 agentic AI and automation initiatives across six business units: value planned and realised, hours…
Executive summary generator — Generate an executive summary for the board, the steering committee or a business unit from the live portfolio data,…
Portfolio: idea → retire — Every initiative on one board from idea to retirement, with gate criteria G0–G4, risk tier, autonomy level, value and…
Policy register — Ten AI and automation policies with owner, status, version, review date and the standards they implement (EU AI Act,…
Governance & gates — Governance bodies with chair, cadence and decision rights, lifecycle gates G0–G4 with entry criteria, and the…
RACI — Who is responsible, accountable, consulted and informed for each activity, on the application's eight roles and on the…
Guardrails catalogue — Sixteen input, output, action, runtime and data guardrails with enforcement point and the risk tiers they apply to.
Risk register — Twelve agentic AI and automation risks on a likelihood × impact heatmap, with owner, mitigation and status.
Complexity & costs — Answer six questions to get the recommended approach, autonomy, deployment model and AI engine, with indicative…
V&V — V-model for agentic AI: requirements to benefits on one side, tests and evaluations on the other, linked to gates…
Reference architecture — Animated reference architecture with eight layers and a governance plane, two enclaves and four deployment models…
AI engines & model choice — Domain engines for legal, HR, operations, strategy and risk, compliance engines for real-time audit, sovereignty and…
Training on institutional culture — Ten steps from inventory of the culture to continuous learning, a decision tree for RAG versus fine-tuning, and what…
Agentic workflow & HITL — Three animated agentic workflows (P2P invoice exceptions, customer complaints, KYC onboarding) with confidence routing…
Hallucination control — Concept and policy: three verification gates (grounding ≥ 0.90, independent critic model, deterministic validators),…
AI security & threats — AI-specific threats — prompt injection and prompt poisoning, RAG and training-data poisoning, raw-data corruption,…
Lessons learned — Eleven anonymised lessons from PIR2-IT and prior engagements — banks, AML, international organisations, national CERTs,…
e-Learning — Nine courses with lessons, a five-question quiz, an 80% pass mark and a certificate, with a recommended path for each…