Kill a popular product, then rebuild: the Lightfield sequence

You have a product with a lot of usage and a hard decision: keep it, or kill it and rebuild. Lightfield's founders made that call after their AI presentation tool, Tome, had 25 million users and $80 million in funding: the product was not a path to a large company. The test has three moves: judge whether usage is mission-critical, shrink the team and story until the old product leaves the room, and build the data layer before the agent demo.

Scale is not a company

A viral hit can look like proof while the business model stays thin. A presentation tool collects millions of users because the cost of clicking is low, and the founder question is blunt: is my product mission-critical? For a company, the answer shows up in revenue concentration, retention, and whether customers expand when the product becomes part of the workflow.

Usage is a weak signal when the product is cheap to try and easy to abandon. A viral product can be a great distribution story and a poor company story. The difference is whether the product protects something the customer cannot easily replace. If it does, the company should invest. If it does not, the founder should treat the popularity as a signal to look for the larger market.

Cut the story before you cut the roadmap

Keith Peiris and Henri Liriani cut headcount from 70 to 7 when they pivoted to an agent-focused CRM, and the company later had about 40 employees. They kept a small engineering core and rebuilt around a gap they had felt while scaling: no AI-native CRM for growing companies. The contraction was a statement about which customer problem the company would own.

The new product was built for agents from the start, using a continuously updated record from calls, emails, and meetings to identify deal risk and likely customer needs. The headcount cut forced a clean break. When a company keeps the old team, the old product keeps a seat at the table. Roadmaps drift, messaging blurs, and the new customer problem gets diluted by the old user base. A small core makes the new product the default.

The story matters because it sets hiring, sales, and engineering priorities. If the company still tells the old story, the new product inherits the old constraints. Lightfield launched publicly in November 2025 after about a year in stealth, and the launch was a reset of the story, the team, and the customer.

Build the data layer before the agent demo

Agents are only as good as the context they can trust. A demo without that context is a trick; a system with it can change how a sales team works. The visible feature is often the agent, but the data layer is what makes it useful.

A sales system that can see the full conversation has a different kind of value than a chatbot that answers questions. The first can surface risk; the second can entertain. For a CRM, the useful output is not a clever answer. It is a reliable record that tells a rep which deal is slipping and what the customer will need next.

Lightfield uses sales-led consultative implementation rather than self-serve onboarding. Technical staff help customers choose the data to capture and the workflows to automate. Self-serve onboarding scales quickly, but it can stay shallow. A consultative motion is slower, but it can make the product part of the customer's operating workflow.

Expansion, not trial, is the test

In September 2026, Lightfield reported about 5,000 companies using its software and 400% net dollar retention. Those numbers show expansion, not just trial. A product that is mission-critical gets used by more people inside the same company.

Net dollar retention is the metric that separates a tool from a dependency. When existing customers expand, the product is becoming part of the daily workflow. When it does not, the usage often stays shallow. The raise gave the rebuilt company room to scale a motion that depends on implementation quality.

On September 9, 2026, Andreessen Horowitz led a $47 million Series A for Lightfield, joined by Lightspeed Venture Partners, Greylock, Alumni Ventures, Audacious, Coatue, and Maverick Capital. The Series A label was notable because many participants had previously backed Tome's earlier rounds. The investor overlap suggests the bet was on the founders' ability to rebuild, not on a single product cycle.