Bain & Company reported at the close of September that as many as 90 percent of enterprises remain focused on tool deployment and narrow use cases, while those that treat artificial intelligence as a full transformation of the business are reporting EBITDA growth of 10 to 25 percent. In its reading, the decisive work lies in redesigning processes and modernizing the data and application environment.
The distance between these two groups is the distance between the pilot and the production line. A pilot flourishes under conditions no enterprise can carry to scale: a single team, data prepared by hand, a person within reach to correct the system’s errors, and a vocabulary the room already shares. A chef may cook by instinct for a table of four; a hundred kitchens require the recipe to be written down. A restaurant franchise is a recipe to which everyone has agreed.
Ask a company how many customers it has, and the honest answer has always been that it depends on who is counting. Sales counts accounts, finance counts legal entities, and support counts logins. For years, a person stood between the data and the decision, quietly reconciling the three. That labor made disagreement a tolerable economy. An agent removes the person. It will answer with complete assurance, drawing on whichever number it encountered first.
Most boards still decide by the number. The more prudent have begun to ask what stands behind it, and they call the question governance: who answers for a figure, what it includes, and whether it means this quarter what it meant last. They would not say semantic layer, and the vocabulary is unnecessary. The instinct is what matters. A board’s credibility rests on the figures it repeats, and defining them first is how that credibility is preserved. Data teams have practiced precisely this for some thirty years, largely uncelebrated, through budget cycles that rewarded whatever shipped first. The work every budget deferred, the definitions, lineage, and ownership, is the work the leaders are now found to have done.
A pilot is a stage of a strategy, and stages end. Ten pilots resting on ten definitions of a customer amount to ten small successes and no capability. A model may be licensed in an afternoon; a definition takes as long as the disagreement it resolves. Each institution will absorb the frontier’s next release at the speed its definitions permit, and that speed was set in the years when no one was asking.