Overview
This programme enables a cross-functional AI Task Force to initiate an AI transformation with practical governance, data controls, security guardrails, and intermediate-to-advanced generative AI solution patterns. The intent is to move beyond SOP-only governance by equipping the task force to design pilot-ready use cases, select appropriate GenAI architectures (e.g., Retrieval-Augmented Generation), apply LLM security controls, and define evaluation/monitoring to manage risk and quality.
Course Objectives
On completing this course, the participants will:
- Establish an AI Task Force operating model (charter, RACI, cadence, decision rights) suitable for rapid AI pilot execution.
- Create a use-case intake and prioritisation process with risk-based approval gates aligned to recognized AI risk management concepts.
- Apply intermediate/advanced GenAI solution patterns (e.g., RAG/grounding, tool-use) to common departmental workflows.
- Identify and mitigate key LLM application risks (e.g., prompt injection and insecure output handling) using recognised guidance such as OWASP LLM Top 10.
- Define a minimum AI data governance package for pilots (classification, access, provenance, retention, quality gates).
- Design evaluation and monitoring plans so pilots have measurable quality, safety, and regression controls.
Target Audience
The AI Taskforce course target audience includes but is not limited to:
AI Task Force members and stakeholders from Sales, Finance, HR, Operations, Sustainability, and other business units.
Risk/Compliance, Security, Data Governance, and IT representatives supporting AI transformation activities.
Non-IT participants who contribute process knowledge, policy ownership, and operational controls.
Pre-requisites
Knowledge: AI fundamentals, basic prompting, and basic data labelling for data governance.
Tooling / access: Organisation-approved GenAI tools and data handling rules.
Module 1: AI Task Force Operating Model for GenAI Transformation
Key topics:
Charter, scope, decision rights, meeting cadence, escalation paths.
AI lifecycle governance orientation (governance as an enabling function for pilots).
Hands-on / Lab:
Draft a Task Force Charter (scope, KPIs, cadence).
Create a RACI covering use-case intake -> build/buy -> deploy -> monitor.
Expected outcomes:
A usable operating model and role clarity for cross-functional AI execution.
Assessment / Checkpoint:
Checkpoint: 10-minute governance walkthrough per group.
Module 2: GenAI Use-Case Design (Intermediate) - From Idea to Buildable Specification
Key topics:
Use-case intake elements: value hypothesis, workflow integration, stakeholders, human oversight points, data needs, and constraints.
Success metrics and acceptance criteria for pilots.
Hands-on / Lab:
Each department group produces a use-case one-pager plus acceptance criteria.
Triage and prioritise use cases into a pilot shortlist.
Expected outcomes:
Pilot-ready use cases with measurable outcomes and clear scope.
Assessment / Checkpoint:
Checkpoint: use-case review board simulation (approve / conditional approve / defer).
Module 3: GenAI Solution Patterns (Intermediate/Advanced) - RAG, Grounding, and Tool-Use
Key topics:
Retrieval-Augmented Generation (RAG) / grounding for enterprise knowledge accuracy.
Pattern selection guidance: when to use RAG vs simple prompting vs tool-augmented workflows.
Data preparation considerations for retrieval (chunking, metadata, access control points).
Hands-on / Lab:
Design a reference architecture for one selected use case: inputs, retrieval sources, access controls, human review, logging/telemetry.
Define a minimal document ingestion + retrieval plan (conceptual, tool-agnostic).
Expected outcomes:
A practical architecture blueprint to hand to IT/security teams or vendors.
Assessment / Checkpoint:
Checkpoint: architecture design review against governance and security requirements.
Module 4: Secure GenAI (Advanced) - Threat Modeling and OWASP LLM Top 10 Controls
Key topics:
LLM application threats (e.g., prompt injection, insecure output handling, data leakage).
Defence-in-depth controls: policy + process + technical controls, with incident response linkage.
Hands-on / Lab:
Tabletop exercise: prompt injection/data leakage scenario with preventive and detective controls.
Create a minimum viable GenAI security control checklist for pilots.
Expected outcomes:
A shared security baseline and response approach for GenAI pilots.
Assessment / Checkpoint:
Checkpoint: tabletop debrief with owners and control actions.
Module 5: Data Governance for GenAI (Intermediate) - Access, Privacy, Quality, Provenance
Key topics:
Data classification and permitted-use rules for GenAI.
Data access control, retention, provenance/lineage, and quality gates for AI pilots.
Hands-on / Lab:
Build an 'AI Data Pack' template: data inventory, owners, retention, access control, quality checks, provenance notes.
Apply the template to the selected pilot use case.
Expected outcomes:
Reusable data governance artefacts that reduce risk and accelerate approvals.
Assessment / Checkpoint:
Checkpoint: peer review of each group’s AI Data Pack.
Module 6: Responsible AI and Release Gates (Intermediate) - Policy-to-Practice Controls
Key topics:
Translate Responsible AI principles into practical release gates.
Human oversight, transparency expectations, and documentation requirements for pilots.
Hands-on / Lab:
Define release gates checklist: approved data sources, security checks, evaluation thresholds, monitoring readiness.
Draft a lightweight AI policy pack (v1) that references the operating model and controls.
Expected outcomes:
A workable governance-and-release process suitable for piloting GenAI use cases.
Assessment / Checkpoint:
Checkpoint: gate review simulation for a candidate pilot.
Module 7: Evaluation and Monitoring (Advanced) - Quality, Safety, and Regression Control
Key topics:
Evaluation dimensions: grounded accuracy, harmful content, privacy leakage signals, operational reliability.
Red-teaming prompts and test set design at a practical, pilot level.
Hands-on / Lab:
Build an evaluation plan for one use case: test set, red-team prompts, pass/fail thresholds, monitoring metrics.
Define a monitoring cadence and escalation criteria for production-like pilots.
Expected outcomes:
A measurable quality & safety plan that prevents uncontrolled pilot behaviour and regressions.
Assessment / Checkpoint:
Checkpoint: evaluation plan review and sign-off criteria.
Module 8: 90-Day Transformation Plan (Capstone) - Pilot Portfolio and Execution Roadmap
Key topics:
Pilot portfolio planning across departments.
Resourcing, communications, and governance rhythm for the first 90 days.
Hands-on / Lab:
Deliverable pack per group: use-case spec, architecture pattern choice, security controls mapping, AI Data Pack, evaluation & monitoring plan, release gates checklist.
Build a 90-day action plan with owners and milestones.
Expected outcomes:
A complete, pilot-ready governance + GenAI blueprint to start AI transformation.
Assessment / Checkpoint:
Final checkpoint: presentation of 90-day plan and deliverables pack.
Programme scope, delivery arrangements and suitability are confirmed during enquiry.