Strategy brings its semantic layer argument to LEAP with a Riyadh partner

Strategy is at LEAP 2026 with Strategy Mosaic, which the company describes as a universal semantic layer — software that unifies fragmented data systems into a consistent framework so that people and AI agents work from the same definitions. It is going to Saudi buyers alongside AiElements, a Riyadh consultancy that specialises in AI-driven data solutions and business intelligence governance. Semantic layer is an unhelpful term for a genuinely useful thing. In practice it means a single, maintained set of definitions sitting between a company's databases and everything that queries them: what counts as revenue, which customers are active, how a fiscal period is bounded, which of four overlapping product tables is authoritative. Without it, every dashboard and every report re-implements those decisions independently, which is why two teams presenting the same metric so often disagree by a few per cent and cannot say why. That has been an irritation in business intelligence for two decades. What has changed is who is asking the questions. A human analyst who gets an odd number goes and checks the query, notices the definition is wrong, and fixes it. An AI agent asked to summarise performance does not. It answers confidently from whichever table it reached, and the error travels downstream in fluent prose with no visible seam. The argument Strategy is making, and it is not only Strategy making it, is that agents raise the cost of inconsistent definitions rather than removing the problem. It is a version of an argument several vendors have brought to Riyadh this week from different directions. ARIS is pitching process context as the missing layer for enterprise agents, Liferay is pairing its agent tooling with a governance argument , and the data-governance advisory firm APGAR is working its Riyadh base on the same premise. The common claim is that the constraint on enterprise AI is not model quality but the state of the data and definitions underneath it — a claim that is convenient for the people making it and also broadly true. The AiElements pairing is the more interesting operational detail. Semantic layers are not shipped and forgotten; they are built out of a specific organisation's conventions, disagreements and legacy systems, and then maintained as those change. That is delivery work, done locally, by people who understand the customer's reporting culture. A software vendor without a local implementation partner tends not to get past the pilot, which is why so many of the vendors on this floor have arrived with one. What has not been published is anything commercial: no Saudi customer named, no deployment, no contract, and no detail of how the Strategy-AiElements relationship is structured. This is a product and partner presence at a trade show, and it should be read at that level. The test for the category is narrow and easy to state. Agents fail quietly on bad definitions. Whether a semantic layer catches that, or simply adds another place for definitions to drift, is a question a customer answers in production rather than on a stand. TechScoop is reporting from LEAP 2026 in Riyadh.