Prentis: Hoffman and Pincus Bet $100M on AI Agents

Two Silicon Valley veterans place a $100M bet against coding

Reid Hoffman helped build LinkedIn. Mark Pincus built Zynga into a social-gaming empire. Now the two are backing a new AI lab called Prentis, and they are aiming it at a target most of the industry has quietly agreed to ignore: the boring, repetitive computer work that fills the average office worker’s day.

Prentis is in talks to raise $100 million, according to TechCrunch. That is a serious number for a company that describes itself not as a startup but as a “neolab,” and it signals that Hoffman and Pincus think the moment for this idea has arrived. Their wager is blunt. Automating routine computer tasks, they believe, will soon overtake coding as the single biggest thing people actually use AI to do.

Read that again, because it cuts against the grain of almost everything the AI industry has spent the last two years celebrating. Code generation has been the poster child of the generative AI boom. GitHub Copilot, Cursor, and a dozen well-funded rivals turned programming assistance into the clearest commercial win the field could point to. Prentis is essentially arguing that the party is looking in the wrong room.

Why routine work is the bigger prize

The logic behind the bet is simple once you sit with it. Software engineers are a valuable but relatively small slice of the global workforce. The far larger population is everyone else who spends the day clicking through spreadsheets, reconciling invoices, updating records in some aging internal tool, copying data from one browser tab into another, and filing the same report every Friday. That work is tedious, it is expensive when measured across millions of employees, and until recently it has been almost entirely resistant to automation because it lives inside interfaces built for humans, not machines.

Agents that can operate a computer the way a person does — reading the screen, moving the cursor, filling in fields, navigating software that has no clean API — are the missing piece. If a model can reliably drive the tools a company already uses, it does not need those companies to rebuild anything. It just needs to show up and do the job. That is the frontier Prentis is planting its flag on, and it is a genuinely hard one, which is part of why the founders are treating this as a lab rather than a quick product play.

The framing matters. A “neolab” positioning suggests Prentis wants to be judged on research ambition, not on shipping a slick demo next quarter. Hoffman and Pincus have both lived through enough hype cycles to know the difference between a feature and a foundation. Betting nine figures on the unglamorous half of knowledge work is a contrarian move, and contrarian moves from people with their track records tend to be worth watching.

A crowded frontier and a real question

Prentis is not walking into empty territory. The idea of AI agents that use a computer directly has become one of the most contested areas in the field, with the largest model developers all racing to make their systems reliable enough to trust with real tasks. The gap between an agent that works in a controlled demo and one that works on a messy corporate laptop at nine in the morning is still wide. Anyone selling this has to close it.

What makes Prentis interesting is less the technology, which many are chasing, and more the conviction about where the value pools. Most of the money and attention in AI still flows toward writing software and generating text. Hoffman and Pincus are saying that the center of gravity is about to shift toward doing the work instead of describing it, and they are putting $100 million behind that reading of the market.

If they are right, the winners of the next phase of AI may not be the companies that write the best code. They may be the ones that finally teach a machine to handle the tasks nobody ever wanted to do in the first place. Whether Prentis becomes that company, or simply proves the thesis for someone else, is the open question worth tracking as the funding round closes.

For more coverage of AI agents and computer automation, visit Mylistingo.

Source: Original Article