Korva is a scroll of the internet’s best ideas.
Easy for one person
A single agent is straightforward to build by prompting, and productivity on personal work jumped as much as a hundred X. Everyone assumed the enterprise would be just as easy. It turned out to be hard.
Who does the implementing
The proof is who is doing the implementing. Microsoft is hiring 6,000 people, and Anthropic and OpenAI have both said they'll deploy their own staff into companies. Palantir is built around forward deployed engineers as well. If AI were what it's described as, a company could simply ask it to do what it needs and to explain how to implement it.
The jobs prediction
The prediction was that half of white-collar workers would lose their jobs within two years. Employment is still growing, people are hiring, and companies need more AI-literate people.
Why is enterprise AI harder than personal use?
A company's data sits in many separate systems, each with its own permissions, formats, and compliance rules. An agent needs credentials and connections to all of them, and its mistakes can touch customers or financial records. A single person working alone faces none of that. MIT's NANDA initiative reported in 2025 that about 95 percent of enterprise generative AI pilots showed no measurable effect on profit and loss.
Why do engineers embed inside customers?
Palantir popularized the forward deployed engineer role in the years after its 2003 founding. Engineers worked on site with government and corporate customers to fit the software to messy data and unwritten workflows. What they learned flowed back into the product team. Other software companies have since copied the model for complex deployments.
Why do new technologies take years to lift productivity?
Economist Paul David's 1990 paper on the electric dynamo found that factories took decades after electric motors arrived to rebuild their floor plans around them. The same lag showed up with computers. Economist Robert Solow remarked in a 1987 book review that the computer age was visible everywhere except in the productivity statistics.
More from All-In with Chamath, Jason, Sacks & Friedberg
