Generative AI for Business Leaders Training Course

Participants can explain core generative AI components and their business implications.

• Transformer models generate tokens from patterns learned during training
• Context windows constrain the information available during each model response
• Temperature and sampling parameters affect output variability and predictability
• Hallucinations arise when generated text lacks reliable grounding evidence
• Multimodal models process combinations of text, images, audio and documents
• AI agents combine model reasoning, tools, memory and workflow actions

Participants can convert business problems into scoped, testable generative AI use cases.

• SIPOC mapping defines suppliers, inputs, process steps, outputs and customers
• User journeys reveal delays, rework, knowledge gaps and service failures
• Task segmentation separates drafting, classification, retrieval and decision support
• Human-in-the-loop design assigns approval points and escalation responsibilities
• Service blueprints connect customer interactions to internal AI-supported activities
• Acceptance criteria define acceptable outputs, exceptions and required evidence

Participants can design prompt patterns and evaluate outputs against business requirements.

• Structured prompts specify role, task, context, constraints and output format
• Few-shot prompting uses labelled examples to guide repeatable responses
• Chain-of-thought exposure is avoided when internal reasoning requires protection
• Prompt templates separate stable instructions from variable business inputs
• Evaluation rubrics score accuracy, relevance, completeness, tone and safety
• Red-team testing probes ambiguity, prompt injection and misleading outputs

Participants can define business requirements for grounded AI responses using organisational knowledge.

• Retrieval-augmented generation grounds responses in approved enterprise source material
• Document chunking balances retrieval precision against context and meaning
• Metadata design supports permissions, retention rules, ownership and traceability
• Vector embeddings represent semantic similarity for knowledge retrieval
• Citation requirements expose sources supporting generated recommendations and answers
• Knowledge governance assigns review cycles, publication authority and content quality

Participants can apply practical governance controls to generative AI use cases.

• NIST AI RMF functions organise govern, map, measure and manage activities
• Risk registers capture harms, likelihood, impact, owners and mitigations
• Data classification determines permitted prompts, storage locations and access controls
• Bias testing examines disparate output quality across relevant user groups
• Model cards document intended use, limitations, evaluation and known risks
• Human oversight thresholds define review requirements for consequential outputs

Participants can identify security, privacy and legal requirements for AI deployment decisions.

• OWASP LLM Top 10 identifies common application security failure modes
• Prompt injection threatens tool access, retrieved content and system instructions
• Data minimisation limits personal and confidential information supplied to models
• Cross-border data transfers require documented processing and hosting decisions
• Intellectual-property review addresses training data, outputs and licence terms
• Records retention rules cover prompts, outputs, approvals and audit evidence

Participants can evaluate AI platforms and suppliers against business, technical and control requirements.

• Build-buy-partner analysis compares strategic control, speed, skills and cost
• Vendor scorecards assess model quality, security, integration and commercial terms
• API architecture determines authentication, logging, rate limits and system integration
• Service-level agreements define availability, support, incident notification and remedies
• Exit planning addresses data portability, model substitution and contract termination
• Proof-of-value criteria establish evidence required before production approval

Participants can define operating arrangements that support responsible AI adoption at scale.

• RACI matrices assign accountabilities across business, technology, risk and legal
• Product operating models organise discovery, delivery, monitoring and improvement
• AI literacy programmes distinguish permitted use from prohibited employee practices
• Change impact assessments identify role changes, skills needs and resistance
• Standard operating procedures document review, escalation and exception handling
• Communications plans explain purpose, controls, user responsibilities and feedback channels

Participants can prepare a measured implementation plan for a priority generative AI initiative.

• Benefits cases quantify time savings, quality gains, risk reduction and revenue
• Total cost of ownership includes licences, integration, data, support and assurance
• Baseline measures establish current cycle times, error rates and service levels
• Leading indicators track adoption, prompt success, review rates and user confidence
• Stage-gate governance links funding decisions to evidence and control readiness
• Ninety-day plans sequence ownership, milestones, dependencies, measures and decisions