[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"article-body-everyone-has-ai-few-get-value":3},"\nEvery enterprise now owns AI (e.g. copilots in the office suite and a\nmachine learning team in the data function). Ask the same enterprises\nto name one decision that improved because of it, and the room goes\nquiet.\n\nThe numbers say the quiet is deserved. The MIT NANDA initiative's\nreport, The GenAI Divide: State of AI in Business 2025, found that\nabout 95% of enterprise generative AI pilots deliver no measurable\nimpact on profit and loss. S&P Global Market Intelligence found that\n42% of companies abandoned most of their AI initiatives during 2025,\nup from 17% a year earlier. Spending rose. Results did not.\n\nThe models were never the constraint. The distance between a model's\noutput and a decision someone commits to is.\n\n## Where the value leaks out\n\nWatch a typical AI output travel through a large company. A model\nproduces a forecast or a risk score, and it lands in a dashboard or an\nemail. A person, if they see it at all, weighs it against their own\njudgement and decides the way they always have.\n\nThe value leaks three times on that short journey:\n\n- **Nobody has to act on the output.** It informs a decision at best.\n  It does not carry one.\n- **Nobody owns it.** No named person is accountable for acting, so\n  acting is optional, and optional actions lose to the day job.\n- **Nobody scores it.** No one records what was predicted, what was\n  decided and what happened next, so the pilot that saved nothing and\n  the pilot that saved millions produce the same slide.\n\nThe MIT researchers place the stall at the workflow boundary, not the\nmodel boundary. People were handed fluent output that did not fit how\nwork got decided, and they quietly routed around it.\n\n## What the 5% do differently\n\nThe minority getting value do not deploy AI at the organisation. They\nwire it into specific, named decisions.\n\nA decision, concretely, is four things. It is a prediction of what\nhappens if nothing changes, a threshold at which someone must act, a\nnamed owner who acts, and a record of whether the action worked. The\ncompanies extracting value put AI inside that loop, not next to it.\n\nTheir questions are worth borrowing. Which recurring decisions in this\nfunction are made late, blind or not at all? What would a prediction\nhave to say, and how far ahead, for the owner to act on it? Where is\nthe evidence, six months later, that acting beat not acting? None of\nthese are data science questions. They are operating questions, and\nmost AI teams are empowered to build models, not to change who decides\nwhat.\n\n## The debt that builds while you wait\n\nBuying more intelligence does not close a decisional gap. The upgraded\nmodel lands in the same dashboard its predecessor did, which is why\nabandonment climbed in the very year model capability jumped. What\npiles up in the meantime deserves a name. Call it decision debt. Every\nrecurring decision made late, blind or unowned adds a little interest,\nand the balance compounds quietly while the pilots multiply.\n\nPaying it down is unglamorous work. Pick the decisions. Name the\nowners. Set the thresholds, route the prediction to the person rather\nthan the portal, and record the outcome so the record settles which\npredictions deserve trust.\n\n*AI creates value at exactly one moment, when a person decides\nsomething differently because of it. Everything else is cost.*\n\nThis is the problem Prophesee was built for. It is a decision layer\nthat turns enterprise data into predictions, exceptions, plans and\nanswers, wired to named owners with outcomes recorded. And it is built\nas a dozen shared problem classes rather than thirty modules, so every\ndecision one team wires up makes the next one cheaper to wire. Built\nwith and proven inside global enterprises. To see it on your own data,\n[start here](/contact).\n",1786984936836]