Microsoft AI Transformation Leader

Results depend on what the model is told and what it can see. You learn how prompt wording changes output, how grounding ties answers to company data, and how retrieval-augmented generation works without writing code.

• What a good prompt contains
• System instructions and user prompts
• Grounding answers in company data
• Retrieval-augmented generation explained
• Why the same question gives different answers
• Judging output quality as a business owner

You build A set of prompts for one business process, with notes on how grounding changes the result.

AI projects that ignore cost lose support fast. You learn what makes usage expensive, from model size and token volume to hosting and data preparation, and how per-user and consumption pricing change a business case.

• Tokens, model size and usage patterns
• Per-user licensing against consumption pricing
• The cost of grounding, fine-tuning and hosting
• Hidden costs: data preparation, training and support
• Building a simple cost model for a use case

You build A cost estimate for one AI use case, with its assumptions and the points where cost could grow.

Copilot sits inside the tools people already use. You map real processes to Copilot’s capabilities in documents, mail, meetings and spreadsheets, and judge where it saves effort and where it does not.

• What Copilot does in Word, Excel, Outlook and Teams
• Copilot Chat and work grounded in organizational data
• Mapping a business process to Copilot capabilities
• Spotting tasks that suit AI and tasks that do not
• Prebuilt agents and what they add

You build A process map that marks where Microsoft 365 Copilot helps in one department’s week.

Some work needs an agent made for it. You learn what Copilot Studio lets a team build, how knowledge and actions shape an agent’s ability, and when to build one instead of using what is ready-made.

• What an agent is and what it can do
• Knowledge sources and actions
• Building against using a prebuilt agent
• Who builds agents: makers, IT and developers
• Testing and improving an agent’s answers

You build A specification for a business agent: its purpose, knowledge, actions and test questions.

When ready-made assistants are not enough, teams build on Foundry. You learn what Foundry Tools offer, how models differ in capability, cost and speed, and how to match a model to a need.

• What Foundry Tools provide
• Model families, sizes and specializations
• Capability, cost and latency trade-offs
• Build, buy or configure: a decision frame
• Working with technical teams on a build

You build A model selection brief that compares options for one use case on capability, cost and speed.

Principles are only useful when someone applies them. You study the responsible AI principles, charter an AI council, and set the review and escalation steps that turn policy into decisions about real projects.

• The six responsible AI principles, applied to a project
• Chartering an AI council
• Risk review for an AI use case
• Policy for acceptable use
• Escalation and oversight
• Communicating AI decisions to staff and customers

You build An AI council charter with its membership, decision rights and a review checklist for new use cases.

AI can expose data that was always reachable but rarely found. You see how data quality, access and privacy affect AI results, and which questions to put to security and compliance teams before a rollout.

• Data quality and readiness for AI
• How permissions shape what AI shows people
• Privacy, regulation and data residency questions
• Security risks specific to AI
• Questions to ask before approving a rollout

You build A data and security checklist for approving an AI rollout in a department.

Technology does nothing until people use it. You assemble an adoption team, recruit champions in each area, plan communication and training, and set up the feedback loops that keep use growing.

• Executive sponsors and an adoption team
• Recruiting and supporting champions
• Role-based training and communication
• Feedback loops and office hours
• Celebrating and sharing early wins

You build A champions program plan with roles, training, communication and a feedback routine.

Adoption stalls on trust, skills and cost, not on features. You anticipate barriers, plan budget and Copilot license types, and choose measures that show whether the change is paying back.

• Common barriers: trust, skills, data and habits
• Choosing Copilot license types for different roles
• Budgeting and phasing a rollout
• Measures of adoption and business value
• Reporting results to leadership

You build A phased rollout and licensing plan with success measures for each phase.

One plan joins everything you have learned. You choose use cases, tools and models, set governance, plan adoption and show the cost and value, then present it to a leadership team.

• Choosing and ranking use cases by value and risk
• Matching tools and models to each
• Governance, data and security conditions
• The adoption, cost and measurement plan
• Presenting the plan to leadership

You build A complete AI transformation plan for a sample organization, ready to take to its leadership.