For the people who buy media and answer for the results.
Tuesday's report says partner A won last month. It cannot say what to do Thursday. The same rows can, read while the money is still moving: which partner, segment, and path is earning its price right now, with every recommendation written down with its prediction before you act, and graded after. You approve every move.
A month in, the question stops being whether to trust it. You just read its record.
For developers and data teams inside a company.
On his first day, a new engineer pointed it at a public taxicab dataset. Trips, fares, tips. Nothing to do with media, nothing configured. Ranked contribution answer in seconds.
It never cared about media. It cares about a cost column and an outcome column.
A closed box with typed inputs and outputs. Send logs, a warehouse share, or a file; get back contribution values, predictions with confidence bands, and a decision record. API or MCP, so your agents call it, open sessions, and keep long-running analyses warm. It runs in your cloud, ours, or your customer's, and nothing else leaves.
Which customers renew next quarter.
Which jobs run on which machine tomorrow.
What a square foot of flooring should sell for this week.
Which piece of an AI system earned the right answer.
Cost column, outcome column, same box.
Every time the plumbing of an industry changed, the winners made the same move. Not selling the new capability. Putting their workflow where their customers already worked, and taking a position in what flowed through it.
| When | Who | The move | What happened |
|---|---|---|---|
| 1976 | American Airlines | Put SABRE booking terminals inside travel agencies; agents booked every airline through American's workflow | agents booked from the first screen 92% of the time; the spinoff was worth $5.9B, rivaling the airline itself |
| 1967 | American Hospital Supply | Ordering terminals in hospital purchasing departments, wired to each hospital's own stock numbers | Became the nation's largest hospital supplier; a $3.8B merger |
| 1975 | McKesson | Handheld order devices in the pharmacy aisle, with the back office wrapped around them | Nearly 100% of orders became customer-entered; the independents ran on McKesson |
| 1979 | Malone / TCI | Traded carriage on the cable graph for equity in the channels that ran on it: BET, Discovery, QVC, Turner | The owner of the distribution graph ended up owning the capabilities too |
| 1982 | Bloomberg | An anchor customer financed the network; data, analytics, and chat kept in one seat | The workflow lives in the terminal; 320,000 seats that nobody cancels |
| 1991 | Walmart | Pulled suppliers onto Retail Link: item-by-store-by-day data, supplier-run replenishment | Suppliers still run their business on Walmart's system today |
| 2006 | Amazon | Externalized its internal infrastructure as metered APIs anyone could call | Every startup's operating workflow formed inside Amazon; $21M to $100B+ |
| 1975 | Kodak | Invented the digital camera and buried it to protect film and prints | Bankrupt in 2012, the year Instagram sold for $1B |
"The preferential display of our flights, and the corresponding increase in our market share, is the competitive raison d'être for having created the system in the first place."
Robert Crandall, American Airlines, defending SABRE in the 1983–84 CAB proceedingsThe capability was never the business. The business was moving the customer's daily work onto your system while the capability was new. The interface changes every generation: a terminal, a handheld, a data feed, an API.
This generation, the interface is a capability an agent calls.
For leadership at companies that sit at the center of an industry's workflows.
Two companies that have never met. One has a capability the other's agent needs in the middle of a workflow. Today that deal takes a contract, a pilot, and a quarter of meetings, so mostly it does not happen.
Here, every action carries a prediction written down first and an outcome graded after, signed. The other side does not have to trust anybody. They read the record, and the money splits by what the record shows.
That is the network. Not held together by a platform in the middle; held together by the record.
If you own a customer graph, a platform, a supply chain, an industry's workflows, you are holding what the airline, the wholesaler, and the cable operator held the day their plumbing changed. Put your products, and your customers' capabilities, on a governed network. Your customers activate them into their systems; their customers do the same. Each activation takes you deeper into a workflow. Only abstracted evals cross the wire: everyone keeps their own data, context, and authority.
The big platforms are building capability catalogs: a human browses, picks a tool, installs it. Notice what a catalog is for. It exists so a human can inspect a tool before trusting it.
When every capability publishes its own graded record, the inspection already happened. Agents check the record and transact in the moment. Nothing installed, no toll to anyone in the middle.
The marketplace does not get beaten. It gets unnecessary.
None of this asks for new behavior. The logs exist. The outcomes exist. The one new thing is that decisions carry their record with them, and once they do, the rest follows: reports that read live, a box any team can point at any dataset, business between companies that never met.
Now it's possible.
| The math | Contribution value in the context of an outcome: what earned what, priced, predicted, graded. Precise owns it. |
| The runtime | Runs any capability, new code or wrapped legacy software, in your cloud, ours, or your customer's, governed, with only abstracted evals visible to anyone else. |
| The context | Your company's own knowledge, captured and delivered to your people and agents by selection, never leaving your control. |
| The record | Every prediction written down first, every outcome graded after, everything auditable and re-runnable. |