The day before Strategy World 2026 opened in Las Vegas, a research note from Citrini Research titled The 2028 Global Intelligence Crisis cratered software stocks across the board. CrowdStrike, Atlassian, GitLab, Datadog, and Asana all dropped 8–12%. The thesis: aggressive AI agent adoption will gut SaaS pricing and displace traditional systems of record. The note pulled in over six million views and spooked public markets.
Twenty-four hours later, Strategy CEO Phong Le walked on stage at the Wynn and made essentially the same argument. He called it “Three Funerals and a Wedding.” The funerals: Traditional BI, data warehouses, and SaaS itself. The wedding: Mosaic, Strategy’s Universal Semantic Layer, paired with AI agents.
But what Le said on stage wasn’t the real story. What he said in a closed-door analyst briefing afterward was.
The Keynote Thesis: Everything Must Go
Le’s argument boils down to this: Enterprises maintain 50,000 reports (10,000 actually get used), spend billions replicating data into warehouses that add no business value, and run 130 applications that could be replaced by a single intelligent layer that understands what you’re asking and goes and gets the answer.
On the BI front, Le argued that dashboards haven’t meaningfully changed in 30 years. Static reports get built, pinned to a wall, and ignored. The replacement: Proactive, intent-based delivery, where information is pushed to users rather than pulled through dashboards.
On data warehouses, Le made his boldest claim: With modern cloud compute and Mosaic pointing directly at source systems, enterprises can bypass the warehouse entirely. He projected $100 billion in industry savings. He took a direct shot at Snowflake and Databricks, arguing that both companies built growth models on convincing enterprises to replicate everything into centralized repositories—work that modern architecture no longer requires.
On SaaS, Le was blunt. He said 90% of current software companies will not exist and that new ones will rise in their place. He compared the disruption to what Bitcoin is doing to traditional finance.
What Happened Behind Closed Doors
In an analyst briefing with industry researchers from Forrester, IDC, Constellation Research, ESG/Omdia, GigaOm, BARC, Nucleus Research, Radiant Advisors, and Treehive Strategy, Le went further than his keynote.
He announced that Strategy plans to replace Salesforce CRM internally within the next year. And ServiceNow. And Workday within two years. Not with competing SaaS products—with AI-generated applications built on top of Mosaic and an open-source data layer.
Then he asked the room: How many other companies are going to do the same thing?
His answer: Everyone. That, Le said, is the death of SaaS.
The plan is straightforward. Build custom applications using AI coding tools, sit them on top of an open-source database like PostgreSQL, layer Mosaic on top for semantic governance, and let users interact through Claude or Gemini with no traditional UI required.
Le acknowledged that ERP replacement is harder and probably five years out. But CRM, service management, and workflow tools? Those are going away fast. He pointed specifically to ServiceNow, calling it “basically a workflow” that is “mostly replaceable by AI.”
The $10 Million Quarter That Killed a Data Warehouse Project
When I asked the room whether Le’s $100 billion data warehouse savings claim was legitimate, a Strategy executive put it in practical terms. The number comes from sizing the total database market and projecting what happens when enterprises stop moving data from point A to point B to point C just because that’s how it’s always been done.
The executive shared one example that made the room go quiet: A large enterprise customer using Databricks hit $10 million per quarter in consumption costs. The company abandoned the warehouse project entirely.
The technical argument is straightforward. Two decades ago, querying data across multiple sources in real time was unreasonable. Latency was too high. Processing was too expensive. Today, it’s not. Cloud architecture has effectively solved the data gravity problem that made centralized warehouses necessary in the first place.
That doesn’t mean databases go away. Trading desks, quantitative finance, and high-frequency applications still need them. But the practice of replicating data into a warehouse purely for analytics—without adding value along the way—is the part Strategy argues is no longer necessary.
One detail that didn’t make the keynote but came up in the analyst session: Mosaic includes a caching layer built on Strategy’s existing BI engine. Slow-moving, frequently accessed data is cached, so no consumption charges are incurred by the downstream warehouse. Combine that with intelligent personalization—a finance user closing the books needs live data for three days a month, but stale data is fine the other 27—and the cost savings start compounding fast.
The Semantic Layer as a “Gigantic Prompt”
The most thoughtful moment in the briefing came from CPO Saurabh Abhyankar, responding to my question about how long before AI agents can reconcile semantic definitions on their own, making a centralized semantic layer optional.
Abhyankar described two possible futures. In one, agents run the business entirely. They define the semantics themselves and don’t care what the layer looks like. In that world, all software becomes a commodity—including Strategy’s.
But Abhyankar thinks a more likely outcome is one where humans still interface with agents and tell them how they want the business to run. In that world, he argued, you need what he called a “gigantic prompt”—a structured way for humans to express how their business should be unique, how it should operate, and how it should make decisions.
That gigantic prompt, Abhyankar said, is the semantic layer.
One of the analysts in the room—a former VP of innovation at Qlik and Microsoft—picked up the thread and drew an important distinction between descriptive and prescriptive semantics. A traditional semantic layer describes your data. But once you want agents to take action on behalf of the business, you need prescriptive definitions: Rules, verbs, workflows, constraints. That’s the territory Strategy is moving into with its business ontology expansion—encoding not just what entities exist, but how they relate and what rules govern their behavior.
Abhyankar emphasized that this prescriptive layer can’t live inside any single system. It can’t be just for Dynamics or just for SAP. It has to describe the entire business. And he argued that Strategy’s independence—not tied to a warehouse, cloud, or application stack—gives them a unique position to build it.
The Copyright Problem Nobody is Talking About
One of the most unexpected threads in the analyst session had nothing to do with BI or semantic layers. It was about intellectual property.
A Strategy executive argued that AI-generated code is not copyrightable under current case law. The prompts that generate the code aren’t copyrightable either—courts have ruled that prompts constitute ideas, not work products of human effort. If you vibe-code a CRM application and your competitor copies it, there’s no legal violation.
The implication for enterprise IT strategy: If the code is a commodity, and the AI that writes it is available to everyone, then data becomes the only defensible asset a company has. Your proprietary data—how your business actually operates, what your customers actually do, what your supply chain actually looks like—is the one thing a competitor can’t replicate.
That argument directly aligns with Mosaic’s value proposition. If data is the last competitive moat, then the layer that governs, defines, and secures access to that data becomes the most critical piece of the enterprise technology stack.
The Reality Check
For all the bold vision, Le was candid about the adoption gap. When asked how far ahead Strategy is of its own customers, he estimated that roughly 10% of attendees in the conference rooms were actively using Mosaic and AI agents today. He expects that number to change dramatically within a year.
The barriers are familiar ones. Le singled out large financial services customers—without naming names, but making clear he was referring to major banks—as companies that struggle to move on any of this. The resistance doesn’t come from data teams. It comes from risk officers and IT governance structures designed to prevent exactly the kind of rapid change Le is advocating.
But Le also drew a sharp line between banks adopting AI and Mosaic and those that aren’t, saying the early movers will create a competitive gap that late adopters will struggle to close.
Competitive Positioning: Direct and Unfiltered
Le was unusually direct about competitors. On Microsoft Fabric, he said the company is too unfocused to execute, distracted by the LLM battle after Gemini and Claude leapfrogged OpenAI, and that Strategy has effectively ended its Azure partnership because Microsoft’s sales teams consistently undercut partners by pushing Power BI into every account.
On Databricks, Le acknowledged that the technology is strong but called their pricing model unsustainable at enterprise scale. He described the company’s growth strategy as innovative lock-in—the same old playbook in a more modern wrapper. On Unity Catalog specifically, Le said it’s a governance tool, not a semantic fabric, and the two shouldn’t be confused.
On Google and Looker, the relationship is healthier. Out of roughly 3,000 customers, Le estimated that only about 10 accounts compete directly with Looker. Google has been responsive to resolving channel conflicts, a stark contrast to the Microsoft dynamic.
Strategy’s current hyperscaler distribution: Approximately 50% AWS, 35% Azure (declining), and 15% Google (growing).
What This Means for Enterprise IT Leaders
Strategy World 2026 was a conference where the vendor’s thesis and the market’s anxiety arrived at the same conclusion from different directions. The Citrini report argued that AI agents will destroy SaaS economics. Strategy’s leadership agreed—and announced they’re already acting on it internally.
The question enterprise IT leaders need to answer isn’t whether this disruption is coming. The stock market already priced that in. The question is: Which layer of the technology stack becomes the control plane for an agent-driven enterprise? Strategy is betting everything on the semantic layer being the control plane.
Whether they’re right will depend on execution. Encoding an entire business into a semantic ontology—verbs, rules, workflows, digital twins—is enormously complex. Only 10% of their own customers are using it today. And the agents themselves are getting smarter every month, which could eventually make a centralized semantic layer optional rather than essential.
But the thesis itself is sound: In a world where code is a commodity, and AI can build applications on demand, the layer that defines what data means, who can access it, and how the business operates is the one thing that can’t be commoditized.
That’s not a keynote talking point. It’s an architecture decision that enterprise IT leaders need to make this year.
