Four tracks, each with working demos. We start with the transformation vision, prove the platform can be trusted, show agents doing real work, and finish with the insights and the roadmap that turn it into a programme.
End to end AI transformation vision with Microsoft Fabric and Azure AI Foundry.
TRACK 026 demosFabric data governance and lineage, AI Foundry safety and observability, agent approval workflows, and shadow AI discovery and controls.
TRACK 034 demosEnterprise knowledge agents with citations, RAG architecture, multi agent orchestration and process automation scenarios.
TRACK 044 demosNatural language analytics in Fabric, executive insights and forecasting, recommendation engine, and end to end Fabric + AI Foundry architecture from data to business outcome.
What you'll experience. End to end AI transformation vision with Microsoft Fabric and Azure AI Foundry.
A short walkthrough of the target state: data unified in OneLake, agents on Azure AI Foundry, outcomes measured on the same numbers.
Talk track · Anchor on outcomes, not features. Every demo that follows is one link in this chain.
The one-page architecture that every later demo lands on: sources, OneLake, semantic layer, agents, guardrails, consumption.
Talk track · Point out that governance and agents sit on the same estate. This is what makes it incremental.
A Fabric report tying AI initiatives to revenue, cost, risk and cycle-time metrics, so the programme is tracked like any other investment.
Talk track · CXO takeaway: AI value is reported on the same governed model as the business.
What you'll experience. Fabric data governance and lineage, AI Foundry safety and observability, agent approval workflows, and shadow AI discovery and controls.
AI scans Purview metadata and drafts business terms and definitions. Data stewards review and approve them, and the approved terms are published back to Purview.
Talk track · AI writes the definitions and people approve them. Everyone ends up using the same meaning for "customer"
AI traces data lineage across Fabric, Databricks, Snowflake, SAP and more, down to column level, with the reasoning behind each link. Approved lineage is published to Microsoft Purview.
Talk track · Pick a table, and the agent maps where every number comes from, up to the Power BI report. What used to take weeks of manual work takes minutes.
Write one access policy and it's enforced across every Fabric engine: rows, columns, files and AI agents. It also finds sensitive data and audits every change.
Talk track · Same query, different users, different results. Then kill the agent and its access drops to zero.
Content safety filters, evaluation runs and end-to-end tracing on a live agent: what it was asked, what it retrieved, what it answered, and how it scored.
Talk track · Show a blocked prompt and a traced answer side by side.
An agent proposes an action; a human approves or rejects in Teams before anything executes. The audit trail records who, what and when.
Talk track · This is the answer to 'what stops the agent from doing something wrong'.
Discover unsanctioned AI usage across the tenant, classify the risk, and apply policy so approved tools are the easy path.
Talk track · Frame as enablement with guardrails, not a lockdown.
What you'll experience. Enterprise knowledge agents with citations, RAG architecture, multi agent orchestration and process automation scenarios.
Ask a policy, contract or product question in plain language. The answer arrives with the source passages cited, from documents the agent is permitted to see.
Talk track · Click a citation. The trust comes from the source, not the fluency.
How the answer was built: ingestion, chunking, vector and hybrid search in Azure AI Search, grounding, and the guardrails from track 02 wrapped around it.
Talk track · Two minutes, one diagram. Then back to the live agent.
An orchestrator agent delegates to specialist agents (data, documents, actions) and assembles one answer, with each hand-off visible in the trace.
Talk track · Show the trace: the room should see that nothing here is a black box.
An end-to-end business process run by agents with humans in the loop: intake, validation, decision, system update, notification.
Talk track · Pick the scenario closest to the audience's industry before the session.
What you'll experience. Natural language analytics in Fabric, executive insights and forecasting, recommendation engine, and end to end Fabric + AI Foundry architecture from data to business outcome.
A Fabric data agent answers business questions in plain language from the governed semantic model, in Fabric and inside Microsoft 365 Copilot.
Talk track · Take a question from the room. Same measures as the report, so the answers cannot drift apart.
An executive view with automated narrative, anomaly callouts and forward forecasts driven by ML models running in Fabric.
Talk track · Lead with the one number that changed and the reason the model gives for it.
Next-best-action recommendations generated from the same estate: who to contact, what to offer, what to fix, ranked by expected impact.
Talk track · Tie one recommendation back to the outcomes scorecard from track 01.
The complete Fabric + AI Foundry architecture and the phased roadmap: what to stand up first, what it unlocks, and how value is measured at each phase.
Talk track · Close on the roadmap and the first 90 days.
Nothing shown today is a point solution. OneLake holds the data once, the semantic layer defines each measure once, AI Foundry agents reason over the same definitions, and governance wraps the whole chain. That is what makes the roadmap incremental rather than a rebuild.
Every source lands once, governed from the first byte.
One semantic model, lineage and certified measures.
Power BI, data agents and natural-language analytics.
Knowledge agents, orchestration, automation.
Forecasts, recommendations, a roadmap you can commit to.