Media & Entertainment Industry Forum

Across two panels, speakers unpacked how teams are moving beyond demos to production, where agents assist instead of autonomously roam, personalization is targeted to moments that matter, and identity and governance make cross-company collaboration safe and valuable.

First principles for AI bets: context, control, choice and collaboration

• Context separates flashy demos from durable systems. Teams emphasized early investment in enterprise semantics so agents and models understand business meaning, highlighting Genie as connective tissue for context.
• Control, through built-in governance, accelerates delivery. Front-loading infosec, legal and policy into apps let teams ship faster with fewer rewrite cycles.
• Choice hedges model volatility. Treating models as swappable components moves organizations from token-maxing to value-maxing.
• Collaboration unlocks growth outside your four walls. Delta Sharing, Databricks Clean Rooms, and Databricks Marketplace capabilities were positioned as the starting point for governed data collaboration across brands, publishers, agencies and identity partners.

Agents that deliver require a trust layer

Production stories focused on measurable value, not autonomy theater. Examples included agent-to-agent audience integrations that let marketers describe an audience in natural language, unify it with first-party signals, then push it to activation, and internal agents that collapsed quarterly business reviews from weeks to hours by unifying data in Unity Catalog and layering agents on top. Teams then extended the same pattern to external customer-facing agents.

Hard-won lessons were consistent: fully autonomous, end-to-end agents are overhyped. Underrated were prompt quality, safety and suitability measurement for AI outputs, human-in-the-loop rollout, and treating migrations as modernizations so you do not just move a data swamp. Governance is a first-class citizen, never a version two.

Personalization that moves the business, from streams to stadiums

When the foundation is in place, personalization becomes an ecosystem rather than a one-off model. A major streaming service builds its homepage from coordinated models for relevance, collection curation, artwork and explanations, with the Databricks Platform powering training and data pipelines. A global game publisher shifted from nightly batch to near real time so it can detect when a player is struggling or affected by cheating and divert them into better experiences in the moment, an investment justified by engagement gains. A sports and entertainment franchise is stitching 60 plus sources into a unified fan record to compress the time from segment construction to activation across teams, arena and a growing district. An identity provider sits beneath much of this as the connective ID spine that enables controlled collaboration across the ecosystem.

Real time emerged as a choice, not a default. It raises costs, engineering complexity, and SLA constraints, so teams advocated targeting use cases where the moment truly changes outcomes, proving ROI, then scaling. A parallel theme ran through every example: clean, governed, fresh data living close to the models that use it.

Operating globally without fragmenting identity

Global personalization faces uneven identity signals and shifting rules. The guidance was to think globally and act locally: build scalable, shared experimentation and ML systems, then tailor content, offers and consent handling to each market. Dedicated trust and safety functions working with legal help set policy proactively. Teams favored global experimentation frameworks that benefit all markets while allowing intentional forks only when local conditions require them.

The quiet groundwork determines the upside

Organizations pulling ahead are not chasing the flashiest agents. They did the patient work of unifying data, hardening governance, and wiring business context into their systems. With that base, they can pick the right model for each task, choose real time only where it pays, and collaborate safely across the ecosystem.