Learning Framework 2026

AI Proficiency Roadmap

Pick your role. Pick your level. Start building real AI skills today.

Four Levels
Your path from Beginner to Expert. Level up your AI skills.
Tier 1

Beginner

1-2 months

Daily Use

Draft emails, generate content, summarize docs, brainstorm ideas.

Tier 2

Intermediate

2-4 months

Daily Use

Automate reports, build content pipelines, analyze data.

Tier 3

Advanced

4-6 months

Daily Use

Build AI-powered automations, create agents, integrate into systems.

Tier 4

Expert

6-12+ months

Daily Use

Drive transformation, build platforms, mentor teams.

Learning Cadence
Stay current through consistent practice.

Weekly

Every Week

Review new tools, experiment with capabilities, share learnings with team.

Monthly

Every Month

Deep-dive into one capability, evaluate ROI on current use cases.

Quarterly

Every Quarter

Skills assessment, technology trend review, strategic planning.

Non-Developer Roles

Non-Developer Roles

Anyone who doesn't write code but wants to leverage AI effectively.

Microsoft Copilot Ecosystem

Web · Teams · Office Apps · SharePoint · Power Platform · Copilot Agents

Tier 1 · Beginner

AI User

1-2 months

  • Prompt engineering
    • Ask Copilot to draft an email to a client
    • Ask for a rewrite in a different tone (formal, casual)
    • Ask a yes/no question and verify the answer
    • Ask for 3 alternatives to consider
  • Tool literacy
    • Use copilot.microsoft.com for general questions
    • Use Copilot in Word to summarize a long document
    • Use Copilot in Teams to recap a meeting
    • Use Bing Chat for web research
  • Meeting AI tools
    • Use Copilot in Teams to generate meeting minutes automatically
    • Ask Copilot to extract action items from a meeting and format as a list
    • Upload meeting notes and ask "what decisions were made?"
    • Ask Copilot to summarize decisions and next steps for absentees
  • Hallucination awareness
    • Check AI facts against a source before sharing
    • Ask "are you sure?" and re-ask the question differently
    • Recognize when AI says "based on my training" — that's a signal to verify
  • Safety & ethics
    • Never paste sensitive data (PII, passwords, financials) into AI tools
    • Don't share confidential company information
    • Know your company's AI usage policy

Daily Leverage

Draft emails, summarize articles, generate copy, brainstorm ideas.

Copilot Web Word Teams Chat SharePoint
Tier 2 · Intermediate

AI Workflow Designer

2-4 months

  • Advanced prompting
    • Give AI context ("I work in HR and need to...")
    • Use chain-of-thought: ask AI to reason step by step before answering
    • Use follow-up prompts to refine output through multiple rounds
    • Ask AI to output in a specific format (bullet points, table, JSON)
  • File-based work
    • Upload a spreadsheet and ask "what are the top 5 trends?"
    • Upload a contract and ask "summarize the key obligations"
    • Upload meeting transcripts and ask "what decisions were made?"
    • Analyze a PDF report and extract key numbers
    • Ask Copilot to search and summarize files in a SharePoint library
  • Context engineering
    • Provide role and audience in every prompt ("as a manager reviewing a report...")
    • Give AI background context and constraints before asking
    • Attach relevant documents so AI has source material to work from
    • Structure multi-part prompts so AI knows what's most important
  • Output evaluation
    • Read AI output critically and ask "does this match the source?"
    • Compare AI output across 2-3 tools and pick the best
    • Know when AI output is good enough vs. needs human review

Daily Leverage

Automate reports, build content pipelines, create email sequences.

Copilot Web Excel PowerPoint Teams Meetings Outlook SharePoint
Tier 3 · Advanced

AI Power User

4-6 months

  • Tool integration
    • Use Power Apps with AI Builder to create a form that auto-classifies submissions
    • Use Power Automate to trigger a Copilot action when a new email arrives
    • Connect Copilot Studio to a SharePoint site so the bot answers questions about your internal docs
    • Build a Power App that surfaces Copilot for employee self-service
  • Custom automation
    • Build a Power Automate flow that pulls data from a table, sends it to Copilot, then writes the response to a report
    • Create an AI-assisted approval workflow where Copilot summarizes a request and routes it
    • Automate weekly report generation — pull data → Copilot writes summary → sends email
  • Data manipulation
    • Use Copilot in Excel to identify trends and generate charts from raw data
    • Ask Copilot to reformat messy data into a clean table
    • Use Power Query + Copilot to explain what a transformation does
    • Generate sample data for testing when real data isn't available
  • Agent creation
    • Create a Copilot Agent in Microsoft 365 that answers common HR policy questions
    • Build an agent that searches your SharePoint and answers questions from company docs
    • Configure an agent to handle recurring IT support questions autonomously
    • Set up an agent that summarizes your team's weekly Slack activity
    • Use Copilot Agents to draft and maintain ADRs, SADs, and AALs autonomously
    • Set up an agent to monitor architecture decisions and flag when docs need updating
  • Task management
    • Draft JIRA ticket descriptions from meeting action items using Copilot
    • Ask AI to break down a feature request into subtasks for your sprint board
    • Use AI to draft ticket descriptions from a client email or request
    • Write acceptance criteria for a ticket using AI
  • Team enablement
    • Train a Copilot Studio bot on your team's internal wiki
    • Document AI workflows and share with teammates
    • Identify repetitive tasks in your team and propose AI solutions
    • Create prompt templates the team reuses for common tasks

Daily Leverage

Build custom agents, connect AI tools, automate team workflows, train others.

Power Apps Power Automate Copilot Agents Excel + Copilot
Tier 4 · Expert

AI Innovator

6-12+ months

  • Strategic planning
    • Identify 3-5 AI use cases with the highest ROI for your department
    • Build a business case to propose AI tools to leadership
    • Plan a phased rollout — start with one team, measure impact, expand
    • Align AI initiatives with company goals and priorities
  • Opportunity identification
    • Audit existing processes and flag where AI can reduce time or errors
    • Spot new product or service ideas powered by AI
    • Identify bottlenecks in knowledge work that AI can solve
    • Recognize when AI is the wrong tool and flag it
  • Governance & compliance
    • Define guidelines for what data can and can't go into AI tools
    • Create an approval process for new AI workflows in your team
    • Ensure AI outputs meet regulatory or policy requirements
    • Document AI use cases for audit and accountability
  • Leadership
    • Mentor other teams on AI adoption
    • Present AI ROI results to stakeholders
    • Drive cultural change around AI-assisted work
    • Build internal AI champions network across departments

Daily Leverage

Identify breakthroughs, establish governance, scale successes.

Power Apps Power Automate Copilot Agents Copilot Studio Custom Copilots
Developer Roles

Developer Roles

Software engineers, data engineers, DevOps, Technical Architects — also need all non-developer AI skills, plus technical depth.

Developer & Architect AI Toolkit

GitHub Copilot · Claude Code · AMP

Tier 1 · Beginner

AI-Augmented Coder

1-2 months

  • Code task prompting
    • Ask GitHub Copilot to generate a GET request handler
    • Ask to write a unit test for a specific function
    • Ask to explain what a regex or algorithm does
    • Ask to convert code from one language to another
  • AI pattern recognition
    • Use Copilot autocomplete to generate boilerplate as you type
    • Accept suggested code and review before using
    • Recognize when AI suggestion doesn't match the codebase context
    • Use tab to accept, tab to reject, and learn the difference
  • Debugging assistance
    • Paste an error message and ask "what's causing this?"
    • Ask "how do I fix a null reference in JavaScript?"
    • Ask Copilot to add console logs to trace a bug
    • Ask for common fix patterns for a specific error type
  • Agentic dev tools
    • Use Claude Code to autonomously build and ship a complete feature branch
    • Use AMP or similar tools to automate repetitive coding tasks
    • Set up an AI agent to run tests, review PRs, or update documentation
    • Use AI to automate CI/CD pipeline monitoring and fix failures
    • Architects: use agents to draft and maintain ADRs, SADs, and AALs autonomously
    • Set up an AI agent to monitor architecture decisions across a codebase and flag drift
  • Security practices
    • Never ask AI to generate API keys, passwords, or credentials
    • Review AI-generated SQL queries for injection risks
    • Ask "is this code secure?" before using AI suggestions for auth
    • Understand what AI knows — it knows patterns up to its training date

Daily Leverage

Generate boilerplate, write tests, explain code, debug errors.

GitHub Copilot
Tier 2 · Intermediate

AI-Assisted Engineer

2-4 months

  • Advanced code review
    • Use Claude Code to review a PR and identify issues
    • Ask "are there any edge cases I'm missing in this function?"
    • Ask AI to check for resource leaks, unhandled errors, or missing null checks
    • Use AI to verify code matches the original intent after changes
  • System design assistance
    • Describe a requirement and ask "what's a good architecture for this?"
    • Ask "draw a diagram" (mermaid, PlantUML) for a flow
    • Ask AI to evaluate trade-offs between two approaches
    • Use AI to draft a technical design doc and iterate on it
    • Generate ADRs (Architecture Decision Records) with context and trade-offs
    • Draft SADs (System Architecture Documents) or AALs (Application Architecture Letters)
    • Use AI to keep architecture documentation up to date as systems evolve
  • Technical writing at scale
    • Generate README files, API docs, and changelogs from code
    • Ask AI to document a complex function in plain English
    • Generate migration guides or upgrade checklists
    • Write code comments and docstrings automatically
  • Project & task management
    • Ask AI to draft JIRA ticket descriptions from meeting notes
    • Use AI to break down a feature request into subtasks
    • Ask AI to summarize sprint progress from a board view
    • Use AI to write acceptance criteria for a ticket
  • Context engineering
    • Give AI your codebase context before asking architecture questions
    • Share relevant files, error logs, or docs before asking to debug
    • Use system prompts to set AI's role and constraints
    • Include examples of desired output format in the prompt
  • Security review patterns
    • Ask "find potential security issues in this code"
    • Ask AI to check for OWASP Top 10 patterns
    • Use AI to generate sample attack inputs to test your code
    • Verify AI security suggestions against known secure patterns

Daily Leverage

Architecture diagrams, comprehensive tests, security reviews.

GitHub Copilot Claude Code
Tier 3 · Advanced

AI-Powered Engineer

4-6 months

  • LLM integration & APIs
    • Call OpenAI or Azure OpenAI API from your application
    • Implement tool calling / function calling so AI can trigger your code
    • Build a chat interface that retrieves context and sends it to an LLM
    • Manage API keys, rate limits, and token usage in production
  • Agent & workflow orchestration
    • Build an agent that can use tools (search, calculator, API calls) to complete tasks
    • Create a multi-step workflow: fetch data → analyze → write response → send email
    • Implement a loop that lets AI retry a task with different prompts
    • Add human-in-the-loop checkpoints for high-stakes decisions
  • RAG pipelines
    • Split documents into chunks and store in a vector database
    • Build a search that retrieves relevant docs and sends them to the LLM
    • Evaluate whether RAG output is accurate and complete
    • Connect your internal knowledge base so AI can answer company-specific questions
  • Production AI systems
    • Build API endpoints that serve LLM responses reliably
    • Add caching so the same prompt doesn't re-call the LLM every time
    • Implement retry logic and fallback when the LLM is unavailable
    • Monitor cost, latency, and quality in production

Daily Leverage

Build LLM-powered features, agents, RAG systems, and production integrations.

Claude Code AMP GitHub Copilot
Tier 4 · Expert

AI Architect

6-12+ months

  • Enterprise strategy
    • Decide which AI capability lives in which tool (Copilot vs. custom vs. third-party)
    • Build a platform that lets multiple teams use shared AI infrastructure
    • Define standards for how AI is used across the organization
    • Evaluate build vs. buy for AI capabilities
  • Multi-model orchestration
    • Route different tasks to different models (fast cheap model vs. slow capable one)
    • Orchestrate multiple AI agents that each handle a subtask
    • Combine copilot-style assistance with autonomous agent workflows
    • Manage model version upgrades with minimal disruption
  • Security & compliance
    • Define policies for data residency, PII handling, and AI outputs
    • Ensure AI systems meet SOC2, GDPR, or other compliance requirements
    • Build guardrails so AI can't be used to generate harmful content
    • Implement audit logs for AI decisions in high-stakes workflows
  • Cost optimization
    • Measure cost per task and optimize prompts to use fewer tokens
    • Set up budget alerts and usage caps per team or project
    • Choose the right model for each task — not every task needs GPT-4
    • Negotiate vendor contracts based on actual usage patterns

Daily Leverage

Platform design, governance, cost optimization, emerging tech evaluation.

Claude Code AMP Custom LLMs
Skills Reference

Cross-Cutting Competencies

Skills that matter across all roles and levels. Click any skill to see examples.

Skills Matrix

Skill Beginner Intermediate Advanced Expert
Prompt Engineering
AI Tool Literacy
Output Evaluation
Workflow Design
Safety & Ethics
Cost Awareness
Team Enablement

1 Non-Developer First Steps

  1. Open copilot.microsoft.com and try basic prompts
  2. Use Copilot in Word to summarize a document
  3. Use Copilot in Teams to summarize a meeting
  4. Pick 1-2 tools to use daily

2 Developer First Steps

  1. Enable GitHub Copilot in your IDE
  2. Use inline completions for boilerplate code
  3. Try Claude Code for a small feature or refactor
  4. Review and refine AI outputs before accepting