Frameworks for building new ventures inside big companies
New ventures inside large companies tend to die in one of two ways. Some are suffocated. They go through the same approvals, margin tests and shared teams as the core business, and the core always wins. Others are left alone so completely that nobody ever decides anything. They keep a small budget and a hopeful slide deck for years. They become zombies.
The fix for both is the same. A new venture needs oxygen and it needs milestones. Room to move, and a clear point at which someone decides whether it lives or not.
There are a few ways to set that up. Which one fits depends on the company and the idea. And AI is changing the math on all of them.
1. Decide what evidence should kill your new venture before you start
At Google X, Alphabet’s moonshot lab, the teams have a saying: monkey first. If your goal is to teach a monkey to recite Shakespeare on a pedestal, don’t build the pedestal first. The pedestal is easy and it feels like progress. Train the monkey. Work on the hardest problem first, because that is where the project will live or die.
Astro Teller, who runs X, says the teams start each day asking how they will try to kill their project. X has ended more than 100 projects in a single year. It has paid bonuses to every person on teams that ended their own work. Ending a project on good evidence is treated as success, not failure.
The lesson for a large company is simple. Before the venture gets money, write down what result would end it, and by when. Most companies only define what success would look like. That is how zombies are made. Nothing is ever bad enough to stop.
2. Give it one owner
Amazon calls it the single-threaded leader. One person owns the new thing, and it is the only thing they own. Dave Limp, who ran Amazon’s devices business, put it this way: “The best way to fail at inventing something is by making it somebody’s part-time job.”
This answers one of the oldest problems in corporate venturing. A $10 million idea is a rounding error inside a $10 billion company. Senior people will always be pulled back to the core, because that is where the numbers are. You don’t fix that with culture or speeches. You fix it by giving one capable person a single job and a motivated team that reports to them.
3. Build it partly outside the walls
Some ideas can’t survive inside the parent at all. They conflict with the brand, the sales force, the core channel dynamics or the margin structure. Or they need a partner the parent doesn’t have. For those, the answer is to set the venture up with its own structure: a joint venture, a spinout, or a venture studio that builds it alongside the company.
I helped build OWN, the Oprah Winfrey Network, as a joint venture between Discovery and Harpo. Neither company could have built it alone. Discovery had distribution and the know-how to run a cable network. Harpo had the brand and the talent. The joint venture gave the network its own leadership, its own board and its own decisions, while it drew on what each parent did best. Many traditional cable networks from Discovery to Viacom to Disney and NBC Universal in the 2010s all suffered from the conflict between streaming and their linear revenue streams. While they controlled all the studio production and TV content, this conflict gave Netflix the oxygen to become the king of streaming right underneath them, and in many cases using their own licensed content.
A separate structure gives the new thing room to make its own choices. It also gives it a board and milestones of its own, which makes the zombie problem harder to hide. The cost is complexity, and the risk that the parent stops caring once the venture is out of sight.
4. Use AI to make prototypes quickly and test
The hardest part of corporate venturing has always been that early bets are too big. Testing an idea once took a team, a budget and a year. That made every new venture a large decision, and large decisions get the full weight of corporate review.
That is changing. In a 2025 study at Procter & Gamble, researchers from Harvard and Wharton gave 776 professionals real product challenges. Individuals working with AI matched the performance of two-person teams working without it. The AI users also produced more balanced ideas, less tied to whether they came from R&D or the commercial side.
If one person with AI can do what used to take a team, the first test of an idea gets much cheaper. A cheap test can move fast without a committee. It can also be killed without a fight. That makes the first two ideas above easier to put into practice. A single owner can cover more ground, and kill criteria can be tested in weeks, not years.
Choosing a path
There is no single right structure. If the idea needs the parent’s customers and assets, and doesn’t threaten the core, keep it inside with one owner and clear kill criteria. If it competes with the core or needs a partner, build it outside the walls. In either case, leverage AI to prototype and test much more quickly to get you real world customer feedback to inform the feed or kill decision.