Lightfield Raises $47M to Rebuild CRM for AI Agents

Keith Peiris and Henri Liriani reached roughly 25M users with Tome, raised about $80M, and still decided the company needed a different future. They shut down the presentation product, reduced the team, and rebuilt around a harder problem: giving a business and its AI agents a customer record they can actually trust.

Lightfield is now financing that second act with a $47M Series A led by Andreessen Horowitz. The September 9, 2026 round also includes Maverick Capital, Coatue, Audacious, Alumni Ventures, Greylock, and Lightspeed Venture Partners. The wager reaches beyond a more convenient CRM: if agents are going to work deals and manage customer relationships, the system beneath them has to preserve what happened, who said it, and how the account changed.

What Happened

Lightfield announced the Series A less than a year after launching publicly in November 2025. The San Francisco company says more than 5,000 companies have signed up, ranging from early-stage startups to organizations with hundreds of people using the CRM. That is company-reported adoption, not a count of paying customers or independently audited retention.

Andreessen Horowitz led the round through an investment case centered on Lightfield's data architecture and the founders' decision to pivot. Several participating investors, including Coatue, Greylock, and Lightspeed, also backed Tome. The announcement does not disclose Lightfield's valuation, ownership terms, investor check sizes, or a reconciled lifetime-funding total.

That last boundary matters because Lightfield is both a new product and the continuation of an existing company. Forbes reported that Tome had raised about $80M and still had roughly half of that capital when the founders committed to the pivot. Adding the new $47M to every historical Tome round would create a neat number without explaining how much prior capital remained, was spent, or now sits behind Lightfield.

The Pivot Behind the Round

Tome had the kind of surface success that usually makes a shutdown difficult to explain. The AI presentation tool reached about 25M users, yet Forbes reported that annual revenue plateaued near $3M as much of the audience stayed on free or low-priced plans. By late 2024, Keith Peiris had concluded that the product lacked the customer context needed to become a durable professional tool.

In March 2025, Tome shut down and cut a team of roughly 70 people to six, according to Forbes. VentureBeat's account of the pivot described another year of stealth work before Lightfield launched for go-to-market teams. The founders moved from helping people assemble a presentation to rebuilding the system that should already know the customer, the deal, and the history behind both.

That is the consequential choice inside the financing. Lightfield is not the product of two founders searching for their first market signal; it comes from founders who watched scale arrive without the economics or context they wanted, then chose to make the company smaller before making the problem larger.

How Lightfield Rebuilds the Customer Record

Traditional CRM systems usually store the latest state of a relationship in fields, stages, and notes. The record is only as current as the person who updated it, which becomes a serious constraint when an AI agent is expected to prepare a meeting, draft a follow-up, find a stalled deal, or explain why an opportunity changed.

Lightfield says it captures email, calendar events, calls, Slack, and LinkedIn activity as those interactions occur. It connects the material to people, accounts, and deals while preserving chronology and attribution, creating what the company calls a world model of the business. The goal is to let a person and an agent inspect the same customer history instead of asking the model to reason over partial summaries and stale fields.

The company also describes an agent harness built around a standardized SDK, a code sandbox, and evaluations that monitor output quality. Its open API approach gives humans, internal agents, and external systems access to the same core objects, while the funding announcement says the record is readable and writable through API, MCP, and CLI. These are company-described capabilities; the announcement does not provide an independent audit of agent accuracy or consistency.

Why Andreessen Horowitz Led

The a16z thesis treats CRM as a system-of-record category that changes when a new computing platform changes the customer. Client-server software produced Siebel, cloud delivery produced Salesforce, and inbound software helped HubSpot become a default for smaller teams. Andreessen Horowitz is betting that companies adopting agents early will choose a customer system built for agents before they inherit the workflows and migration costs of an incumbent stack.

The investment also follows the founders, not only the category. Keith Peiris and Henri Liriani met at Meta and built Tome together before making the reset. a16z framed their willingness to shrink to a core engineering team and rebuild around a problem they had experienced firsthand as part of the underwriting case.

What the $47M Changes

Lightfield has said the investment will expand engineering, while its current careers board shows hiring across engineering, machine learning, customer success, sales, marketing, recruiting, delivery, and new-market development. That hiring map suggests the next phase involves more than product construction. A CRM company also has to migrate data, support implementation, teach new operating habits, and earn the right to become the place where revenue teams keep customer truth.

The company-reported growth is moving quickly. Forbes said Lightfield had 1,000 paying customers in July 2026 and reported 80% monthly revenue growth, while the September announcement cites more than 5,000 signups. The figures describe different populations and remain unaudited, so they should not be collapsed into a single conversion or retention claim.

The Larger CRM Test

The AI software market has spent years adding assistants to systems designed around human data entry. Lightfield starts one layer lower by arguing that the record itself must change before agents can do dependable customer work. That makes context, chronology, permissions, and evaluation part of the CRM product rather than features bolted onto it later.

The $47M Series A gives Lightfield more room to build that architecture and recruit the teams required to sell it. The founders have already shown they are willing to abandon scale that does not lead where they want; now they have to persuade customers to hand the rebuilt company the memory behind their pipeline, service, and product decisions, one migration and one verified agent action at a time.