A roadmap for the Telecom, Media, Entertainment, and Gaming industries
TL;DR
- The data moved. The decisions did not. Customer data now sits in the lakehouse, but bids, offers, and pacing are still decided in outside tools.
- That gap costs money and adds risk. Egress fees, latency, compliance exposure, and an audit trail your CFO cannot follow end to end.
- Running decisions on Databricks addresses all three. Your agents act on full customer context, you see what every AI decision costs, and you keep one record from signal to result.
- DemandPilot runs it for you. Live in four to six weeks, with ROAS uplift around 35% in our production retail media work.
A postpaid customer in Dallas has three dropped calls in a week. On Monday night she compares plans on a rival’s site. By Tuesday morning, your lakehouse knows: churn risk is up, contract ends in 40 days, lifetime value is in the top decile. The retention offer goes out on Friday. She filed her port-out request on Thursday.
Nothing was wrong with the data. The decision was running somewhere else.
That delay is getting more expensive. Deloitte’s 2026 Digital Media Trends survey found 41% of US consumers canceled at least one paid streaming service in the past six months. Budgets offer no cushion. Gartner’s 2026 CMO Spend Survey puts marketing budgets flat at 7.8% of revenue, and 70% of CMOs say their processes are not mature enough to scale AI. As Gartner’s Ewan McIntyre puts it, “AI maturity is beginning to separate marketing leaders from laggards.”
Most telcos, streamers, publishers and game studios are here today. They spent years consolidating customer data into one lakehouse. The decisions did not follow. Bids still change inside a DSP. Offers still fire from an ESP. Make-goods are still worked out in a spreadsheet.
That split is about to get expensive. Gartner expects more than 40% of agentic AI projects to be canceled by the end of 2027, citing escalating costs, unclear business value, and inadequate risk controls. Marketing agents that decide outside your governed data carry all three risks at once.
Where the Decision Runs Is Now a Board Question
Regulation now reaches the decision itself.
GDPR and CCPA already limit where subscriber data can be processed. Since August 2, 2026, the EU AI Act’s transparency rules also require disclosure when people interact with AI and labeling of certain AI-generated content. Every third-party layer in that decision is one more dependency your legal team has to defend.
Signal volume breaks the round trip.
Few sectors produce marketing signal at this scale. Sending it to an external decisioning service and pulling the answer back adds egress fees, latency, and gaps in lineage. At telecom and streaming scale, all three move the ROI math.
The CFO wants the whole chain.
CFOs now ask marketing to show the signal, the decision, and the result behind the last dollar spent. On a third-party stack, the honest answer stops where the vendor’s logging stops. In your own environment, you can show the full chain. That difference decides whether marketing is a line the CFO defends or one the CFO cuts.
What Changes When Decisions Run on Databricks
At Data + AI Summit in June 2026, Databricks argued that agent quality depends on context. It launched CustomerLake, an agentic CDP with Profile and Campaign Agents, and Unity AI Gateway for runtime control of AI spend. Four things change when the decision layer moves inside that perimeter.
Your business definitions stay yours.
Unity Catalog Metrics and Genie Ontology fix what pacing, fill rate, make-good, and sell-through mean inside your governance boundary. No vendor can redefine them.
Your controls govern the agents.
Agent Bricks runs the multi-agent logic under your Unity Catalog permissions. Human approval gates use the same access model as the rest of your data estate. The audit trail flows to your systems instead of a vendor console.
AI cost becomes visible.
Every agent decision is a compute event on a platform you already pay for. It is metered in the same ledger as your other workloads, so your team can see what each decision costs and tune it.
Every decision ties to an outcome.
Each agent action writes to a governed table with full lineage, so the CFO and the CISO read the same record. Our Value & Outcome Ledger ties every agent decision to a business result, inside your own perimeter.
“Your signal already lives in your lakehouse. The next advantage goes to the telecom, media and gaming companies whose marketing decisions run there too.”
Speed of Change Matters More Than the Savings
Moving decisions in-perimeter cuts middleware cost. The bigger gain is how fast you can change. When your decision layer is someone else’s API, your marketing improves at the pace of their roadmap. When it runs in your environment, a new agent on the same governed data takes about a week.
Three Shifts to Plan For in 2027
Procurement is scoring it. In the RFPs we see, where a marketing AI decision is made now appears as a scored item.
Middleware is consolidating toward the platform. Identity, activation and measurement vendors are rebuilding for native Databricks Marketplace listings.
Databricks is moving into marketing. CustomerLake puts Databricks directly into the martech category. That raises the value of whoever operates the decision layer and stands behind the numbers.
Who Runs It Once It Moves
Standing up the decision layer is the easy part. The hard part is running it around the clock, keeping context clean as markets switch on and defending the audit trail in a QBR. Most marketing technology teams are staffed to keep a stack healthy. Running a governed multi-agent operation is a different job.
That is the work DemandPilot does. We operate the agentic marketing decision layer for telecom, media, entertainment and gaming companies directly on Databricks, inside your Unity Catalog perimeter. Our forward deployed engineering pods go live in four to six weeks, and for the first cohort, Databricks Customer Investment Fund credits fund the pilot.
In production across our retail media and advertising work, the model has delivered around 35% ROAS uplift, sustained 99% campaign accuracy at portfolio scale, and scaled measurement from $200 million to $1 billion in annual volume without a matching rise in operating costs. It carries to any telecom, media or gaming company whose subscriber data already lives on Databricks.
Three Questions for Your Next Board Pack
Where do your marketing decisions run today?
For your top campaigns, name the system where bid, audience, pacing, and retention calls are made. Count how many sit inside your governance boundary. That ratio is your exposure.
What does it cost to move, and what does it cost to stay?
Moving is a finite project cost. Staying compounds through middleware fees, compliance overhead, and the speed you give up. In our client conversations, the cost of staying overtakes the cost of moving within two years.
Who will operate it?
If your team is not staffed to run a multi-agent operation around the clock with governed lineage, you need an operating partner.
The Dallas customer did not leave because your data was wrong. She left because the decision arrived a day late. Close that gap and every campaign starts to pay for itself.
Where to go next
See the returns DemandPilot delivers, 35% ROAS uplift, 99% campaign accuracy and measurement scaled from $200M to $1B, run natively on Databricks.
