IntelligentHQ
IntelligentHQ
Skip to content

Media and entertainment

How broadcasters, sports bodies and publishers use Microsoft technology: grounded assistants over owned archives, personalization that holds up at live-event peaks, and AI in production, within copyright and GDPR limits.

Try the Agent Platform Chooser

The problem

Media owners hold decades of content that is hard to find and reuse. The Premier League needed to bring 30 seasons of match data, 300,000 editorial articles and 9,000 videos into one environment that fans and editors could query, refreshed during live matches.

Demand is spiky: interest peaks around live events, so any AI feature must answer fast under heavy load and keep personal data out of places it should not be.

Generative AI raises rights questions specific to the sector. Outputs can reproduce third-party material, and audience data in Europe falls under GDPR.1

The approach

Microsoft's media and entertainment page organizes its offer around content creation and production, reliable delivery across platforms, and audience and commercial intelligence. It names Microsoft Foundry, Microsoft 365 Copilot, Azure OpenAI in Foundry Models, Microsoft Fabric, Azure AI Video Indexer, Copilot Studio and Dynamics 365. Media is not one of the Microsoft Industry Solutions on Learn.

The Premier League Companion is a documented design. Azure Cosmos DB holds match and player data, prompts and anonymized conversation logs; Azure AI Search provides agentic retrieval over articles, data and video; Semantic Kernel orchestrates agents and partner systems; and Azure Managed Redis caches frequent answers to keep live interactions under 500 milliseconds. The League operates under GDPR, logs every data interaction and red-teams open prompting features. Microsoft's story reports engagement up 20 percent year over year, a vendor figure from early in the season. Semantic Kernel now has a direct successor in Microsoft Agent Framework.

For many publishers, the search and recommendation features of their CMS or video platform are enough. A grounded assistant is worth building when the archive is large, owned and well tagged, and when the organization can staff the editorial and safety review it needs.123

Conceptual architecture

Media and entertainment: conceptual architectureCopilot & businessappsAgents & automationAI models & platformData & analyticsSecurity, identity &governanceCloud & hybridinfrastructureFan app and website: Where audiences ask questionsFan app and websiteAgent orchestration (Semantic Kernel; Microsoft Agent Framework for new builds): Plans retrieval and calls partner systemsAgent orchestration(Semantic Kernel; Microsoft…Retrieval index (Azure AI Search): Agentic retrieval over text, data and videoRetrieval index (Azure AISearch)Generation (Azure OpenAI in Foundry Models): Writes grounded answersGeneration (Azure OpenAI inFoundry Models)Match data, articles, video and photos from partner APIs: Owned and licensed contentMatch data, articles, videoand photos from partner APIsOperational store (Azure Cosmos DB): Structured data, prompts and anonymized logsOperational store (AzureCosmos DB)Guardrails, logging and red teaming: Safety, traceability and GDPR controlsGuardrails, logging and redteamingAnswer cache for live peaks (Azure Managed Redis): Keeps frequent answers fastAnswer cache for live peaks(Azure Managed Redis)
Diagram as a list

  1. Copilot & business apps

    • Fan app and websiteWhere audiences ask questions

  2. Agents & automation

    • Agent orchestration (Semantic Kernel; Microsoft Agent Framework for new builds)Plans retrieval and calls partner systems

  3. AI models & platform

    • Retrieval index (Azure AI Search)Agentic retrieval over text, data and videoConnects to Agent orchestration (Semantic Kernel; Microsoft Agent Framework for new builds)
    • Generation (Azure OpenAI in Foundry Models)Writes grounded answers

  4. Data & analytics

    • Match data, articles, video and photos from partner APIsOwned and licensed content
    • Operational store (Azure Cosmos DB)Structured data, prompts and anonymized logsConnects to Agent orchestration (Semantic Kernel; Microsoft Agent Framework for new builds)

  5. Security, identity & governance

    • Guardrails, logging and red teamingSafety, traceability and GDPR controlsConnects to Agent orchestration (Semantic Kernel; Microsoft Agent Framework for new builds)

  6. Cloud & hybrid infrastructure

    • Answer cache for live peaks (Azure Managed Redis)Keeps frequent answers fastConnects to Fan app and website
Conceptual: one common way to arrange the parts, not a required design.

Technologies and their roles

  • Microsoft Foundry1

    AI platform

    The Premier League built its Companion with Microsoft Foundry.

  • Microsoft Foundry Models1

    Generation

    Azure OpenAI in Foundry Models generates Companion content.

  • Azure AI Search1

    Grounded retrieval

    Powers agentic retrieval so fans and editors can find articles, match data and videos.

  • Azure Cosmos DB1

    Operational data hub

    Central store for match and player data, prompts and anonymized conversation logs.

What you need first1

  • Rights clearance for every archive and partner feed used for retrieval and generation
  • Consistent tagging of articles, video and match data
  • A consent model for any personal data, agreed with the data protection officer
  • Editorial review and red teaming capacity before and after launch
  • Load testing against real live-event peaks

Risks and how to reduce them

Copyright in generated output4
Microsoft's Customer Copyright Commitment for Azure OpenAI requires mitigations such as a protective metaprompt, evaluation for third-party content and protected material filters; implement and document them.
Personal data in fan interactions1
Follow the Premier League pattern: no personal data without consent, anonymized logs, fine-grained access controls and audit trails.
Design capacity mistaken for real load1
The 8,000 transactions per second figure is described as capacity; test your own peak traffic.

Documented examples

Related

Sources

  1. Customer story: Premier League (Microsoft Foundry) (opens in a new tab)Microsoft · Customer-reported, Vendor-reported
  2. Media and Entertainment Solutions | Microsoft AI (opens in a new tab)Microsoft · Vendor-reported
  3. Microsoft Agent Framework overview (opens in a new tab)Microsoft · Vendor-reported
  4. Customer Copyright Commitment Required Mitigations - Microsoft Foundry (opens in a new tab)Microsoft · Vendor-reported

Fill out the form below to request your copy.

Name(Required)