He turned off a product with 20M users and went from 70 people to 7—then raised a $47M Series A. | Keith Peiris, Co-Founder & CEO of Lightfield

Lightfield shut off Tome to rebuild the CRM around AI

Lightfield is Keith Peiris's bet that an AI-native system of record replaces the CRM. Tome, his previous product, had 20m users and cheap organic acquisition, but expensive inference made gross margins terrible and retention near zero, so the growth was worth little. Pilots with ten companies showed demand was for customer data assembled across call recorders, email and Slack, not decks. He turned off Tome, cut 70 people to seven, and rebuilt around that.

Takeaways

• Tome's user growth hid a weak business. The product reached about 20m users and was adding one to two million a month, with low acquisition cost from organic demand. But inference made cost of goods sold very high, and repeat usage and revenue retention were low.
• Ten pilots pointed away from decks. Keith ran pilots with about ten companies building decks for sales and marketing teams. Within weeks all ten lost interest and pulled toward account research, proposals and RFP responses. The data agents need sits in call recorders, email, Slack and warehouses, not Salesforce.
• They shut off Tome before finding a replacement. After speaking to their board, Keith and his team turned off the product with no successor and cut from 70 people to seven. Board member Dan argued technology super cycles come once a decade or two, and a soft landing would leave them stuck for years.
• They bet on rebuilding the system of record. The team concluded the only play they would all be excited to spend a decade on was reinventing the system of record, and they wanted the technical challenge of an agentic one. Companies already run about 30 go-to-market tools offering roughly the same thing.
• The wedge differs by company stage. For seed-stage companies the first job is pipeline generation, which means building lists and sending outreach. Larger ones need AI agents and a clean sales-to-implementation handoff. Seed founders said they would pay more than HubSpot or Attio for a system of record solved on day one.
• Metrics are inherited from the core product. Customer acquisition cost moves only about 10 to 20 percent, and the same product cannot go from 20 to 40 net dollar retention. Pablo Srugo adds that a business at $10m in revenue growing 20 percent a year is worth about $30m.

Landing the first YC customers

1. Reached YC founders on LinkedIn rather than email and never led with a pitch.
2. Hosted dinners and talks, invited HubSpot's CRO and the author of Predictable Revenue, and bought 100 copies of the book.
3. Gave the first ten YC customers the software free in a shared Slack channel.