Your complete guide to
AI — 360°
Structured learning for every role — from first prompt to enterprise architect.
The major AI tools — and when to use each
Before choosing a platform, understand what's available today. An honest side-by-side of the six leading AI tools, their strengths, and the right use case for each.
The most widely used AI assistant globally. Best all-round capability — writing, coding, reasoning, image generation, voice, data analysis. Largest ecosystem of custom GPTs and integrations.
Leading model for nuanced writing, long-document analysis, and coding benchmarks. 200K context window. Claude Code excels at agentic terminal-based development.
Deeply integrated with Google Workspace — Docs, Gmail, Sheets, Meet. 1M token context window. Best for multilingual tasks and teams already running on Google infrastructure.
AI embedded in Word, Excel, PowerPoint, Outlook, and Teams. Only tool with direct access to your actual files, emails, and calendar. Copilot Studio extends it with custom agents.
A research engine, not a chatbot. Every answer comes with cited sources from live web data. Best when you need verifiable, up-to-date information fast. Not designed for writing or coding.
The developer's AI companion inside VS Code, JetBrains, and the terminal. Writes, explains, and debugs code inline. Copilot Workspace enables full agentic feature development from issue to PR.
When to use which AI tool
Nuanced prose, editing, or analysis of large documents?
→ Use ClaudeCode completion, reviews, or agentic feature building?
→ GitHub Copilot + Claude CodeNeed current, cited, verifiable answers?
→ Use PerplexityWorking in Word, Excel, Outlook, Teams daily?
→ Use Microsoft CopilotCustom no-code agent on Microsoft infrastructure?
→ Copilot Studio (see MS track)Foundation models, RAG, agents on AWS infrastructure?
→ Amazon Bedrock (see AWS track)General productivity, image generation, voice, custom GPTs?
→ Use ChatGPT PlusProduction AI with fine-tuned models and vector search?
→ Azure AI Foundry or AWS BedrockWhat best describes you?
Jump straight to the content that's most relevant for your role and goal.
The Microsoft AI platform — what does what
Microsoft offers multiple AI products. Understanding how they fit together is the first step to building on the right foundation.
AI embedded in Word, Excel, PowerPoint, Outlook, Teams, and Edge. Reactive — you ask, it helps. GPT-4o powered. Best starting point for business users.
Build custom AI agents without code. Design, configure, connect to SharePoint or APIs, and deploy to Teams, websites, or M365 Copilot.
Enterprise access to GPT-4o, o3, DALL·E, Whisper, and Embeddings via Azure. Private, compliant, and integrated with Azure security and governance.
200+ model catalog, prompt flow, RAG pipelines, evaluation, fine-tuning, and hosted agent runtime in one place. Massively updated at Build 2026.
Vector indexing, semantic ranking, and hybrid retrieval. The primary grounding engine for RAG implementations in Copilot Studio and AI Foundry applications.
Microsoft's open-source AI orchestration framework. Plug in any LLM, build agents, chain prompts. The developer-first framework in the Microsoft ecosystem.
Which Microsoft AI tool should I use?
Match your role and goal to the right starting point
AI in Word, Excel, Outlook, Teams?
→ Microsoft CopilotBuild a team chatbot using your data, no code?
→ Copilot StudioBuild AI applications or APIs on Azure?
→ Azure AI Foundry + Azure OpenAIMulti-agent, RAG, governance-grade AI at scale?
→ AI Foundry + Semantic Kernel + AI SearchThree Microsoft AI learning tracks
Each track builds directly on the previous. Start at Foundations even if you have some experience — it establishes the Copilot Studio mental model.
- 🧠AI concepts & ecosystemML, GenAI, Agentic AI explained
- 💬Microsoft Copilot for daily workM365, Teams, Word, Excel
- 🤖Your first Copilot Studio agentCreate, test, publish in 30 min
- 📖Knowledge sources & groundingSharePoint, websites, documents
- 📡Publishing & channelsTeams, web, M365 Copilot
- ✏️Prompt engineering & system promptsPersona, instructions, LLM config
- ⚡Actions: Power Automate & APIsFlows, REST, dynamic outputs
- 🧠RAG & grounding patternsRetrieval-augmented generation
- 🔐Auth & Entra ID SSOSecure, personalised agents
- 📊Analytics & optimisationTelemetry, CSAT, iteration
- 🌐Autonomous agent designProactive triggers, multi-step reasoning
- 🤝Multi-agent orchestrationAgent networks, delegation, context
- 🏗️Azure AI Foundry integrationCustom models, vector search, RAG
- 🛡️Governance & Responsible AIDLP, audit trails, compliance
- 🔄ALM & CI/CD for agentsDevOps, versioning, promotion
All Microsoft modules — filter by level
Every module combines theory, a practical walkthrough, and a hands-on lab exercise.
What is AI, machine learning, generative AI, and agentic AI? Understand the full landscape before touching any tool.
Use Copilot in Teams, Outlook, Word, Excel, PowerPoint — summarise, draft, analyse, automate without writing a line of code.
What Copilot Studio is, how it fits within Power Platform, and the difference between Copilot, agents, topics, and knowledge sources.
Create an agent with natural language, configure identity, add a knowledge source, test, and publish to Teams in one session.
Build trigger phrases, conditional branches, and use entities to extract key information from user messages.
Connect SharePoint, public websites, uploaded documents. Understand how generative answers retrieve and cite from your knowledge base.
Publish to Teams, embed in a website, surface via M365 Copilot. Understand channel settings, authentication, and testing.
Master system prompt design, configure AI topic generation, tune response style and length, understand LLM orchestration inside Copilot Studio.
Trigger Power Automate flows from conversations, call REST APIs directly, handle dynamic outputs, build connector-based actions end to end.
Implement RAG patterns — combine generative AI with grounded knowledge. Configure search indexes, chunking strategies, citation formatting.
Configure SSO with Entra ID, scope knowledge access by user context, enable secure personalised data operations in your agents.
Built-in dashboards, session transcripts, satisfaction scores. Use conversation telemetry to improve topics and reduce escalation rates.
Build custom connectors and extend agents with OpenAPI definitions to integrate third-party services and internal APIs.
Design agents that proactively trigger on events, reason across multiple steps, and take autonomous actions without user prompts.
Build agent networks where a master orchestrator delegates to specialist agents. Implement handoff protocols and context passing.
Connect Copilot Studio to Azure AI Foundry for custom models, fine-tuned embeddings, and enterprise-scale vector search with Azure AI Search.
Implement content moderation, DLP policies, audit logging, and responsible AI frameworks for enterprise Copilot Studio deployments at scale.
Manage agent lifecycle with Power Platform ALM: export solutions, configure environments, build Azure DevOps pipelines, promote to production.
Build something real and deployable
Each lab is a complete, real-world project. Follow the guided steps or use them as a blueprint for your own variation.
- 1Create a new agent in Copilot Studio
- 2Add a public website as knowledge source
- 3Configure generative answers & citations
- 4Test with 10 realistic user questions
- 5Publish to Microsoft Teams
- 1Connect SharePoint list as knowledge
- 2Create topics for status queries
- 3Extract project name as entity
- 4Format responses as adaptive cards
- 5Deploy and test in Teams
- 1Build leave request conversation flow
- 2Create Power Automate approval flow
- 3Integrate Outlook calendar API
- 4Add manager approval adaptive card
- 5Send confirmation with calendar invite
- 1Upload PDF policy documents
- 2Configure Azure AI Search index
- 3Tune chunk size and overlap
- 4Enable citations in agent responses
- 5Evaluate accuracy with a test set
- 1Register an app in Microsoft Entra ID
- 2Configure SSO in Copilot Studio
- 3Use user context in the system prompt
- 4Filter Dataverse data by user claims
- 5Test with multiple user personas
- 1Design orchestrator with routing logic
- 2Build HR, IT & Finance specialist agents
- 3Configure agent-to-agent handoffs
- 4Pass context across agent boundaries
- 5Aggregate and return unified responses
Official Microsoft learning resources
Official MS Learn video series — from conversational to fully autonomous agents.
Watch on Microsoft LearnComplete official docs — concepts, how-tos, API reference, troubleshooting.
Browse docsOfficial self-paced path covering AI and ML fundamentals on Azure. Aligned to AI-900/AI-901.
Start pathAsk questions and connect with thousands of Copilot Studio builders in Microsoft's official forum.
Join communityOfficial sample solutions, starter templates, and reusable patterns from Microsoft's GitHub repo.
View samplesEnterprise AI platform for model deployment, prompt flow, and RAG — integrates with Copilot Studio.
Explore FoundryThe AWS AI platform — what does what
AWS offers a comprehensive suite of AI services from foundation models to production-grade agent infrastructure. Here's how the key services fit together.
AWS's managed service for accessing 40+ foundation models via a single serverless API — Claude, Llama, Titan, Mistral, Stable Diffusion, and more. No infrastructure to manage. The equivalent of Azure OpenAI on AWS.
Build fully managed AI agents that break down tasks, call APIs (Action Groups), query knowledge bases (RAG), and take multi-step actions. The AWS equivalent of Copilot Studio's agentic capabilities.
AWS's new production agent runtime (2026). Manages agent state, memory, tool execution, observability, and security. Equivalent of Microsoft Agent Framework on AWS. Included in the Generative AI Developer Professional exam.
Fully managed RAG — connect S3 documents, sync to an OpenSearch or Amazon Aurora vector store, and ground agent responses in your enterprise data. The AWS equivalent of Copilot Studio's knowledge sources + Azure AI Search.
AWS's end-to-end ML platform. Build, train, fine-tune, and deploy custom models. SageMaker JumpStart provides a foundation model catalog. SageMaker Studio is the IDE for ML engineers and data scientists on AWS.
AWS's enterprise AI assistant — the M365 Copilot equivalent on AWS. Connects to 40+ enterprise data sources (Salesforce, Jira, SharePoint) and lets employees ask questions in natural language. Q Developer assists developers in the IDE.
Apply content filters, topic restrictions, PII redaction, and grounding checks across all Bedrock models and agents. The responsible AI governance layer — equivalent to Microsoft's DLP and content moderation for Copilot Studio.
Visually chain prompts, agents, knowledge bases, and Lambda functions into end-to-end AI workflows — no code required. Think Power Automate for AI pipelines on AWS.
Which AWS AI service should I use?
Match your goal to the right AWS starting point
Use multiple FMs via a single API without managing infrastructure?
→ Amazon BedrockAgent that queries documents, calls APIs, takes multi-step actions?
→ Agents for BedrockRAG over S3 documents with automatic indexing and retrieval?
→ Bedrock Knowledge BasesCustom ML models, fine-tuning on your data, model hosting?
→ Amazon SageMakerEmployees ask questions across Jira, Salesforce, SharePoint?
→ Amazon Q BusinessContent moderation, PII protection across all AI interactions?
→ Bedrock GuardrailsThree AWS AI learning tracks
From first API call to production-grade agentic systems. The AWS path is more developer-oriented than Microsoft's low-code approach — some coding experience is helpful from Intermediate onwards.
- 🧠AI & ML concepts on AWSLLMs, generative AI, responsible AI
- 🪨Introduction to Amazon BedrockConsole walkthrough, model catalog
- ✏️Prompt engineering on BedrockZero-shot, few-shot, chain-of-thought
- 💬Amazon Q Business fundamentalsEnterprise assistant, data connectors
- 🛡️Responsible AI & GuardrailsContent safety, PII, compliance
- 🗄️RAG with Bedrock Knowledge BasesS3, OpenSearch, chunking, retrieval
- 🤖Agents for Amazon BedrockAction groups, Lambda tools
- 🔄Bedrock Flows & Prompt ChainingVisual workflow design
- 🧑🔬SageMaker JumpStart & Fine-tuningFoundation models, custom datasets
- 📊Evaluating & monitoring AI appsBedrock Evaluations, CloudWatch
- 🚀Bedrock AgentCore & agent runtimeState, memory, observability
- 🤝Multi-agent patterns on AWSSupervisor agents, sub-agents, Strands
- ⚡Serverless AI with Lambda & Step FunctionsEvent-driven agentic workflows
- 🛡️Advanced security & governanceIAM, VPC, audit, cost governance
- 🔬Advanced RAG & vector optimisationS3 Vectors, embeddings, hybrid search
All AWS modules — filter by level
Every module includes concept explanation, AWS console walkthrough, and a hands-on lab using real AWS services.
LLMs, foundation models, generative AI, the AWS AI/ML stack, and responsible AI principles including fairness, explainability, and safety.
Navigate the Bedrock console, explore the model catalog (Claude, Llama, Titan, Mistral), run your first inference, understand pricing and inference parameters.
Zero-shot and few-shot prompting, chain-of-thought, system prompts, and inference configuration. Practical exercises using the Bedrock Playground.
Set up an Amazon Q Business application, connect enterprise data sources (S3, Confluence, Salesforce), configure permissions, and test natural language queries.
Configure content filters, topic restrictions, PII redaction, and grounding checks. Apply Guardrails across models and agents. AWS responsible AI principles.
Connect S3 documents to a managed vector store (OpenSearch or Aurora). Configure chunking, embedding models, metadata filtering, and retrieval tuning.
Build agents with Action Groups (Lambda functions) and Knowledge Bases. Configure agent instructions, test traces, deploy via alias. End-to-end agent project.
Build visual AI pipelines by chaining prompts, agents, knowledge bases, and Lambda nodes in the Flows designer. Low-code AI workflow automation on AWS.
Explore the SageMaker JumpStart model catalog, fine-tune a Hugging Face model on a custom dataset, deploy as a real-time endpoint, evaluate performance.
Use Bedrock Evaluations for automated and human-based quality scoring. CloudWatch metrics for agent performance. Cost governance and quota management.
Use LangChain's Bedrock integrations to build chains, agents, and RAG pipelines in Python. Connect to OpenSearch, DynamoDB, and external APIs as tools.
Production agent runtime (2026). Manage agent state, persistent memory, tool execution, observability, and fine-grained security for enterprise multi-step agents.
Supervisor agent + sub-agent patterns. Agent Squad and Strands frameworks. Context passing, inter-agent communication, result aggregation in multi-agent systems.
Build event-driven agentic workflows using Lambda for tool execution and Step Functions for orchestrating multi-step AI pipelines. Production patterns.
IAM policies for Bedrock, VPC endpoints, PrivateLink, CloudTrail audit logging, cost governance, and data residency for enterprise AI deployments on AWS.
S3 Vectors, OpenSearch Serverless, hybrid search, reranking, metadata filters, query expansion, and GraphRAG patterns for enterprise knowledge retrieval.
Build real AI apps on AWS
Each lab builds a complete, deployable application using real AWS services. AWS free tier or a sandbox account covers most labs.
- 1Set up Bedrock model access in the console
- 2Run prompts in the Bedrock Playground
- 3Call Bedrock API with Python (boto3)
- 4Build a simple conversational loop
- 5Add Guardrails for content filtering
- 1Create an Amazon Q Business application
- 2Connect an S3 bucket as data source
- 3Configure user access with IAM Identity Center
- 4Test natural language queries
- 5Embed as a web widget
- 1Upload PDFs to S3, create Knowledge Base
- 2Configure OpenSearch Serverless vector store
- 3Sync and verify the embedding pipeline
- 4Query with Retrieve & Generate API
- 5Display citations in a Streamlit app
- 1Create Bedrock Agent with instructions
- 2Build Lambda Action Group functions
- 3Define OpenAPI schema for actions
- 4Connect Knowledge Base for FAQs
- 5Test traces, deploy via alias
- 1Prepare training dataset in JSONL format
- 2Upload to S3, configure SageMaker training job
- 3Fine-tune Llama via SageMaker JumpStart
- 4Deploy as real-time inference endpoint
- 5Evaluate and compare vs base model
- 1Create HR, IT, Finance sub-agents with action groups
- 2Build supervisor agent with routing instructions
- 3Enable multi-agent collaboration in Bedrock
- 4Pass context across agent boundaries
- 5Test, trace, deploy with AgentCore runtime
Official AWS learning resources
Official AWS training platform — free and paid courses, labs, practice exams, and the AI Practitioner learning path.
Access Skill BuilderComplete official docs — user guide, API reference, SDK examples, and best practice guides for Bedrock and agents.
Browse Bedrock docsOfficial AWS samples for Bedrock, agents, knowledge bases, and RAG patterns — ready-to-run notebooks and applications.
View on GitHubFree hands-on workshops in sandbox AWS environments — Bedrock, agents, RAG, and SageMaker labs with step-by-step guidance.
Browse workshopsAWS's official Q&A community. Ask questions about Bedrock, SageMaker, and Amazon Q — answered by AWS experts and community members.
Visit re:PostLatest announcements, architecture patterns, and tutorials from the AWS AI/ML team covering Bedrock, SageMaker, and Amazon Q.
Read the blogmultiple platforms AI certifications
All current AI certifications mapped across both platforms — with retirement warnings and replacement exam IDs where applicable.
The AI entry point. Covers ML concepts, computer vision, NLP, generative AI, and Azure AI services. Lifetime validity. Sit before June 30 or wait for AI-901 (same scope).
- AI workloads & responsible AI
- ML principles on Azure
- Computer vision, NLP, GenAI
For M365 Copilot admins. Covers Copilot licensing, governance, administration, security, and deployment. Generally available now.
- M365 Copilot administration
- Agent creation basics
- Governance & compliance
Validates Power Platform knowledge including Copilot Studio basics. First cert for the Maker path. Covers Power Apps, Power Automate, Power BI, and Copilot Studio.
- Power Platform overview
- Copilot Studio basics
- Dataverse & connectors
For engineers building AI solutions on Azure. Retiring June 2026 — replaced by AI-103 (Azure AI App & Agent Developer) with expanded agentic AI scope.
- Azure AI Foundry & Azure OpenAI
- RAG with Azure AI Search
- Responsible AI on Azure
Validates Copilot Studio configuration, Power Automate design, and Dataverse modelling. Required for the Expert path (PL-600).
- Copilot Studio configuration
- Power Automate solution design
- Dataverse data modelling
New 2026 certification for Power Platform professionals building AI-first solutions. Apps, agents, automation, and AI models together. Beta from April 2026.
- AI agent design & embedding
- Copilot-driven solution design
- Responsible AI governance
AWS's entry-level AI certification. Covers AI/ML concepts, generative AI fundamentals, Amazon Bedrock, Amazon Q, SageMaker, and responsible AI. No coding required. Equivalent to Microsoft's AI-901.
- AI & ML concepts on AWS
- Generative AI & foundation models
- Amazon Bedrock overview
- Responsible AI principles
For ML engineers deploying and maintaining AI/ML solutions on AWS. Covers SageMaker, model training and deployment, MLOps, Bedrock integration, and production pipelines.
- Amazon SageMaker AI end-to-end
- Model training, tuning, deployment
- MLOps and CI/CD for ML
- Bedrock model integration
Deep specialisation in ML on AWS for data scientists and ML engineers. Covers feature engineering, model selection, training, tuning, and production deployment. The established ML certification path.
- Data engineering for ML
- Exploratory data analysis
- ML model training & tuning
- ML implementation & operations
AWS's flagship AI certification. Validates ability to build production GenAI apps using Bedrock, Knowledge Bases, Agents, AgentCore, RAG pipelines, and multi-agent systems. Most challenging AWS AI exam.
- Bedrock, Knowledge Bases & Agents
- RAG optimisation & vector search
- Multi-agent with AgentCore
- Security, governance & cost
AI use cases for your role
Real, specific use cases mapped to your role — not generic art of the possible. Each card shows the task, expected output, and the exact AI tool to use. Filter by your role or browse all.
Paste a client requirements doc — Copilot extracts scope, recommends rate card, flags risk items, and drafts a pricing summary ready for review.
Upload the RFP. A Copilot Studio agent maps each question to existing proposal content, drafts answers, and highlights gaps needing human input.
Ask Copilot to summarise latest competitor moves, pricing changes, and win/loss patterns from news, earnings calls, and Battlecard documents.
Describe the deal in plain language — Copilot drafts a tailored PowerPoint with exec summary, value proposition, pricing table, and next steps.
After a client call, Copilot reads the Teams transcript and drafts a personalised follow-up email with action items and pricing recap — sent in minutes, not hours.
Prompt Copilot with margin thresholds — it generates a scenario table showing impact of 5%, 10%, 15% discount options on gross profit and revenue.
Upload a contract PDF. The agent highlights non-standard clauses, unusual liability caps, missing SLA definitions, and jurisdiction risks against your standard template.
Describe what a clause needs to achieve — the agent drafts standard-aligned language, explains trade-offs, and suggests alternatives at different risk levels.
Upload v1 and v2 of a contract. Copilot produces a plain-English summary of every change, flagging which changes are material and which are cosmetic.
Paste a regulatory framework (GDPR, DPDP, SOC2). The agent produces a contract review checklist specific to that regulation and jurisdiction.
Before a negotiation call, Copilot summarises open points, walk-away positions, and the other party's likely priorities based on prior correspondence.
A Copilot Studio agent trained on your signed contract library. Team members ask "what is the SLA for client X?" and get an instant, cited answer from the contract.
A Teams chatbot collects joiner details, triggers the AD ID creation flow, sends the welcome email, and tracks task completion — all without manual PMO intervention.
Triggered by a leaver request, the agent generates a role-specific checklist, assigns tasks to IT and HR, tracks completion, and confirms account deactivation.
Resource submits an extension request via Teams. Agent validates, routes to manager for approval, and on approval extends the account automatically — no email chain.
Ask "how many resources are due for offboarding this month?" or "which AD IDs expire in 30 days?" — instant answers from live SharePoint data, no report needed.
Pending requests older than the SLA threshold automatically trigger a Teams nudge to the approver. No manual chasing from the PMO team — ever.
"What is the notice period for offboarding a contractor?" A Copilot Studio agent answers PMO process questions from the policy docs, available 24/7 in Teams.
Paste a code block — GitHub Copilot explains what it does, flags security issues, suggests performance improvements, and adds inline documentation.
Paste error logs and stack traces. The AI identifies the root cause, explains it in plain English, and suggests a fix with code. Cuts debug time by 60-70% on complex issues.
Select a function — Copilot generates a full unit test suite covering happy path, edge cases, and error conditions. Reduces test-writing time by up to 80%.
Describe a design decision in plain English — Azure AI Foundry generates a structured ADR with context, options, trade-offs, and decision rationale in your template.
Upload a BRD or workshop transcript. Copilot extracts user stories in structured format with acceptance criteria, gaps flagged, and ambiguities highlighted.
Paste a user story — the agent generates a test plan with positive, negative, and edge-case scenarios in your format, ready for import into Azure DevOps or Jira.
Walk through a process in a Teams meeting. Copilot produces a structured process document with roles, steps, decision points, and exceptions from the transcript.
Paste the sprint log or project tracker update — Copilot rewrites it as a client-friendly status update: clear language, no jargon, RAG status, and next steps.
Before a C-suite client meeting, Copilot summarises all recent project updates, open risks, commercial positions, and action items into a one-page executive brief.
A morning summary of the client's latest news, earnings signals, industry moves, and regulatory changes — every client conversation informed by current context.
A Copilot Studio agent monitors project status reports across all workstreams, surfaces RAG-red items or trends towards risk, and sends a weekly leadership summary.
Upload the account plan and recent client conversations — Copilot identifies whitespace, matches to service offerings, and drafts talking points for the growth conversation.
After every client meeting, Copilot extracts action items from the Teams transcript, assigns owners, sets due dates, and posts to the shared project tracker automatically.
Paste client emails, survey responses, and meeting notes. Copilot extracts sentiment trends, recurring themes, and relationship health signals for QBR preparation.
6-month path to AI mastery
A structured progression covering both multiple platforms, with certification milestones at each stage. Pick one platform to go deep, or follow both in parallel.
Month 1 — AI Foundations (Both platforms)
Explore the AI tools landscape. Sign up for free tiers: Copilot Studio trial + AWS Bedrock free tier. Complete beginner modules on both platforms. Build your first agent (Copilot Studio) and first chatbot (Bedrock Playground).
Month 2 — First Certifications: AI-900 + AIF-C01
Study and sit Azure AI Fundamentals (AI-900, or wait for AI-901) and AWS AI Practitioner (AIF-C01). Both are entry-level, free to study on Microsoft Learn and AWS Skill Builder. Lifetime validity — no renewal needed.
Months 3–4 — Builder / Associate Track
Dive into intermediate content on your chosen primary platform. Build a RAG policy agent (MS) or a Bedrock Knowledge Base app (AWS). Complete authentication, analytics, and connector labs. SageMaker fine-tuning if AWS-focused.
Month 4 — Associate Certifications: PL-200 or MLA-C01
Microsoft path: sit PL-200 (Power Platform Functional Consultant). AWS path: sit MLA-C01 (Machine Learning Engineer Associate). These open the advanced/professional tracks and significantly increase employability.
Months 5–6 — Architect / Professional Track
Complete advanced modules. Design multi-agent systems (both platforms), integrate Azure AI Foundry or Bedrock AgentCore, implement governance, build CI/CD pipelines for AI. Build the most complex labs.
Month 6+ — Expert & Professional Certifications
Microsoft: AB-410 (Intelligent Applications Builder) + AI-103 (replaces AI-102 Jul 2026). AWS: Generative AI Developer Professional. These are the top-tier AI credentials on each platform — significant study investment but maximum career value.
Common questions
If your organisation uses Microsoft 365, start with Microsoft — Copilot Studio is immediately applicable to your daily work with zero infrastructure costs. If you're a developer on AWS, start with Bedrock. Both paths lead to the same agentic AI capabilities, different tools.
No. Copilot Studio is genuinely low-code/no-code — you can build and deploy agents using only the graphical interface. Advanced features (custom connectors, Azure AI Foundry) benefit from coding but aren't required for most scenarios.
Basic Bedrock usage (Playground, Q Business) requires no coding. Building production agents with Action Groups, Lambda integrations, and RAG pipelines requires Python. The Associate track onwards assumes basic Python familiarity.
Yes, if you can sit it before June 30, 2026. It has lifetime validity and the replacement AI-901 has the same scope. If you're reading this after June 2026, target AI-901 directly — the same Microsoft Learn study materials apply.
Bedrock = access pre-built foundation models via API, no infrastructure. SageMaker = build, train, fine-tune, and host your own custom ML models. Most organisations start with Bedrock for generative AI and use SageMaker when they need custom model training.
Google Cloud Vertex AI content is in the roadmap for the next major release — covering Gemini on Vertex, Agent Builder, Model Garden, and the relevant GCP AI certifications. The portal is structured to accommodate a third platform seamlessly.
More platforms coming
AI 360 covers multiple platforms today. Google Cloud Vertex AI is in development — Gemini on Vertex, Agent Builder, Model Garden, and RAG Engine.
Host AI Learning Hub on Azure
Deploy this portal to Azure Static Web Apps in minutes — global CDN, automatic SSL, and a free tier. Share a live URL with every training participant instantly.
az staticwebapp create