Strategy World 2026: Three Customers, Three Industries, One Honest Gap Between AI Vision and Enterprise Reality

Three enterprise customers at Strategy World 2026 described what Mosaic adoption actually looks like. The gap between keynote vision and production reality is the story.

Hours before this customer panel began, Strategy CEO Phong Le told a room full of industry analysts that traditional BI is dead, data warehouses are unnecessary costs, and 90% of software companies won’t exist in the near future. CPO Saurabh Abhyankar demonstrated AI coding tools building complete analytics applications on top of Mosaic in minutes.

Then three enterprise customers sat down in a packed room and described what it actually looks like to implement this technology inside organizations that process toll transactions, manufacture beverages, and fly airplanes.

The gap between the keynote and the customer panel wasn’t a contradiction. It was the most useful information at the entire conference.

300 Billion Transactions and a Decision to Keep the Data Warehouse

Sachin Bhatta, BI Director at the North Texas Tollway Authority, is the closest thing Strategy has to a Mosaic power user. NTTA was one of the earliest pilot customers, and they’re now live in production with Mosaic running in dark mode alongside their existing infrastructure.

The numbers are striking. NTTA processes over 300 billion toll transactions and moves 1.7 terabytes of data daily across approximately 80,000 log files. They serve 40 million customers across 150 managed lanes in North Texas. They’ve been running their data architecture on Strategy’s platform for 15 years.

Bhatta described the immediate value of Mosaic in practical terms: the ability to combine traffic data with transaction data. These are two fundamentally different data streams — one captures revenue continuously, the other tracks vehicle speed, distance, location, and timing across road sensors to optimize the driving experience. Before Mosaic, combining them for analysis required a lengthy pipeline through ETL, staging, and the enterprise data warehouse. With Mosaic, Bhatta’s team can model and analyze the combined data directly.

But here’s what made Bhatta’s session stand out from the keynote messaging: he’s not bypassing the data warehouse. At least not yet.

“I’m still planning to keep the BigQuery Google database for now, until I see the prototype can replace that,” Bhatta told the audience. He acknowledged the vision that Abhyankar outlined in the morning keynote — that Mosaic will eventually replace the need for data warehouse construction and modeling — but he’s waiting for proof at his scale before making that transition.

What Bhatta did endorse without hesitation was Mosaic’s impact on new projects. For any net-new data source — pulling ServiceNow data, combining external data with an existing data lake — he recommended starting in Mosaic rather than running it through the traditional pipeline. The speed difference, he said, is dramatic: what used to take weeks of data engineering can now be modeled in five to ten minutes.

NTTA also gave the audience a concrete AI use case. Bhatta’s team had been running customer clustering models in Google Vertex AI — 32 distinct processes involving knowledge graphs in BigQuery and ML models. He noted that Mosaic’s built-in clustering feature could have accomplished much of that work within a single tool, saving significant cost. It’s the kind of detail that doesn’t show up in keynotes but matters to practitioners making budget decisions.

18 Plants, 6 Million Cases a Month, and a Lot of Excel

Anthony Harnisch, Supply Chain Analytics and Reporting Manager at Keurig Dr. Pepper, brought a different perspective. KDP has been a Strategy customer since 2012, but they’re in the infancy stage with Mosaic. His value wasn’t in Mosaic implementation stories. It was in describing the problems that Mosaic needs to solve.

The most vivid example: KDP’s urban plant produces roughly 6 million product cases per month. Throughout the month, the forecast for that volume changes — switching between a 12-pack of cans and a 24-pack of PET bottles has a real financial impact, sometimes $100,000 or more on a single packaging change. Today, the planning team meets with site production teams and manages those forecast adjustments in Excel. That data never touches the company’s master data or actuals systems. The financial impact of each change is either invisible or takes too long to calculate.

Harnisch sees Mosaic as the path to bringing that offline data into the same semantic layer as actuals, so the financial impact of a packaging change shows up in real time. It’s not a flashy AI use case. It’s closing the gap between what production teams know and what finance teams can see.

He also gave what might have been the most grounded endorsement of Strategy’s semantic layer in the entire conference. KDP pulls data from more than 10 different sources, and over 11 years, nearly every one of those sources has changed at least once — including a migration from SQL Server to Snowflake. Their end users had no idea any of it happened. The semantic layer absorbed the changes. That’s not AI. That’s not Mosaic. That’s the core architecture doing exactly what it was designed to do.

The company is also navigating a corporate split. Keurig is separating from Dr. Pepper, and while Harnisch said the current split won’t heavily impact reporting infrastructure, he pointed to KDP’s 2017 merger — when Keurig acquired Dr. Pepper Snapple Group — as proof of concept. When the coffee side and the beverage side needed to report financials together, they had no efficient way to mesh the data. A semantic layer that can absorb new data sources on the fly would have made that transition significantly easier.

Where Harnisch sees the biggest AI opportunity is with site directors at KDP’s 18 beverage manufacturing plants. These are experienced operators who understand their financials deeply but spend maybe 10 minutes a day at a computer. Their questions currently go through other teams and take time to be answered. Conversational AI on top of a governed semantic layer would let them ask questions and get answers directly — no dashboard training required.

800 Million Fares, Safety-Critical Data, and Why Trust Comes Before AI

Faarina Memon, Data & Analytics Product Portfolio Manager at Emirates, brought a regulatory perspective. Before Emirates, Memon worked at the NFL, and she drew a sharp comparison between the two environments: both demand that the right information reaches the right person at the right time, but Emirates raises the stakes because the data is often safety-critical. Trust and clarity aren’t optional.

The data scale at Emirates is enormous. The airline processes approximately 2 million shopping requests per day. Their global pricing system captures around 800 million fares from roughly 500 airlines worldwide. All of that information currently flows through Strategy dashboards, and Memon’s team is actively piloting conversational BI to make the experience more seamless for business units.

But Memon was deliberate about where Emirates stands in the adoption curve. They’re in the early stages with Mosaic, and her framing consistently focused on the foundation before acceleration. The real challenge, she said, isn’t a lack of information — it’s potential inconsistency across data sources. With data flowing from staffing, fuel, customer, and commercial systems, the risk is that different teams make decisions from different understandings of the same data.

Her strongest point landed on governance. In a highly regulated industry, governance has shifted from reviewing decisions after the fact to building guardrails into the decision-making process itself. Explainability, she argued, matters more than model sophistication. Leaders need to know where a specific insight came from before they act on it. That’s where semantic clarity becomes essential — and why Memon sees the semantic layer as a prerequisite for AI, not a complement to it.

Looking ahead, Memon connected this directly to Dubai’s planned tech-first airport. As scale and complexity increase, the tolerance for fragmented or manual data interpretation drops to zero. Her vision is augmented decision-making at scale — not full automation, but AI-assisted decisions built on a foundation that humans can verify and trust.

The Pattern: Trust Before Adoption, Foundation Before AI

The most striking thing about this panel wasn’t any single customer story. It was that all three, independently, said the same thing in different words: trust drives adoption, and trust requires a solid semantic foundation before AI can deliver value.

Harnisch put it most directly: there’s a correlation between adoption and trust. If users trust the data, they’ll adopt the tools. If they don’t, they’ll go back to pulling data from SAP or their source systems manually.

Memon framed it through governance: build the trust before you put AI on top. Bhatta gave the tactical version: get your data dictionary and data catalog in order first. He also offered a useful corrective to AI hype — AI reads textual information well, he said, but still struggles with mathematical calculations. That’s why the metrics layer in Mosaic matters. It’s not just about defining what “revenue” means. It’s about ensuring that calculated metrics are computed correctly, not hallucinated by a language model.

The Questions the Keynote Didn’t Answer

The audience Q&A raised several practical concerns that enterprise teams evaluating Mosaic should address.

The first is integration with existing Strategy deployments. Organizations that have spent years building semantic layers with thousands of attributes and metrics on Strategy’s traditional platform want to know how those investments carry forward into Mosaic. Bhatta acknowledged this is a gap. New projects work well in Mosaic. Integrating the existing semantic layer from an Intelligence Server environment with Mosaic is still pending. He’s hopeful for a future release that bridges the two.

The second is the memory and compute cost. When one audience member asked about the limitations of in-memory processing, Bhatta didn’t sugarcoat it: expanding memory capacity costs money. But he argued that Strategy’s caching through intelligent cubes is still significantly cheaper than hitting a data warehouse with every query. The economics shift when you factor in the compute costs that cloud data platforms charge per query.

The third is preparatory work. An audience member asked about all the foundational work — KPI definitions, ontologies, taxonomies — that a tool can’t do for you. Bhatta’s answer was direct: you have to do that work before Mosaic delivers value. Data governance, data catalogs, and naming conventions need to be in place so that when Mosaic reads your metadata and an LLM answers questions against it, the answers are consistent.

What This Means for Enterprise Analytics Teams

Strategy’s keynote painted a future in which data warehouses are unnecessary, BI tools are dead, and AI agents handle everything via a Universal Semantic Layer. The customer panel painted a picture in which a toll authority keeps its data warehouse while testing Mosaic in parallel, a beverage manufacturer still manages production forecasts in spreadsheets, and an airline methodically builds semantic foundations before turning on AI capabilities.

Neither picture is wrong. The keynote describes where the technology is heading. The customers describe where enterprises actually are. The distance between those two points is measured in data governance maturity, integration complexity, and organizational trust — not in product features.

For analytics leaders evaluating Mosaic, the practical takeaway from this panel is clear: start new projects in Mosaic to capture the immediate speed benefits. Keep your existing data infrastructure running until integration with legacy Strategy environments matures. Invest in your data dictionary and catalog before expecting AI to deliver consistent answers. And understand that the path from traditional BI to an AI-driven semantic layer isn’t a migration — it’s a transition that runs both environments in parallel until trust is established.

The customers who are furthest along aren’t the ones who moved fastest. They’re the ones who built the foundation first.