Korva
Startups

Figma Rebuilds Startup Products to Test If the Technology Is Real

Its finance-led deal team uses quick AI rebuilds to gauge how hard a product was to build.

AI image
The a16z Show · The New CFO: How AI Is Rewriting the Finance Function | Figma CFO Praveer Melwani

When Figma looks at buying a startup, its deal team runs an unusual test. It tries to rebuild the startup's product itself, using artificial intelligence.

The method is sometimes called vibe coding, which means describing the software you want in plain words and letting AI write the code. Praveer Melwani, Figma's chief financial officer, said his corporate development team, the group that evaluates potential acquisitions, now uses it as one of its checks in diligence.

The question is simple: can a few finance staffers recreate what the startup built? If they can, that can hint at whether the product is real and how difficult it was to build.

"How deep is your moat? How interesting is the technology there?" Melwani said. A moat, in business language, is an advantage that keeps competitors from copying you. He said the rebuild gives the team what he called a solid intuition, even though many nuances remain.

That is a change from how such reviews used to work. Before, finance would have needed help from the product team to get that kind of technical read. Melwani said his team still partners with product specialists, but it arrives with evidence in hand.

"This is what I built, this is what I've seen," he said of the new approach. Showing a working copy builds a lot of credibility.

The builders are not engineers on loan. They are deal and finance staff with finance backgrounds, not product people. Melwani said the team has gotten good at this kind of quick rebuilding.

For now, the rebuild is a way for finance to bring more to its work alongside product specialists. The team brings what it made, and works through the remaining nuances together.

Questions people ask

What is vibe coding?

Vibe coding means telling an AI coding tool what you want in everyday language and letting it generate working software. It can produce a simple app or website in hours, though the result is usually a rough copy rather than a secure, finished product.

Does this replace review by Figma's product team?

No. Melwani said his team still partners with Figma's product specialists. The rebuild gives finance a starting point and credibility, not a final technical judgment, and deeper questions still go to product experts.

What does the team learn if it succeeds or fails?

Generally speaking, an easy rebuild might raise questions about how hard a product would be to copy, while a hard rebuild might point to deeper technical work. Melwani gave no scoring rules for the test. He said only that the exercise offers a solid intuition and that nuances remain.