Lightfield Raises $47 Million to Build a CRM for AI Agents

Lightfield, an AI-native customer relationship management startup, raised $47 million in a Series A round led by Andreessen Horowitz, the company said Sept. 9.

The round included Lightspeed Venture Partners, Coatue, Greylock, Maverick Capital, Audacious and Alumni Ventures. Lightfield did not disclose a valuation.

What $47 million buys

Lightfield launched in November 2025, and the company says more than 5,000 companies have signed up since, with dozens replacing Salesforce outright. Chief Executive Keith Peiris said in a statement announcing the round that "agents don't fail because models lack capability. They fail because data is incomplete, inaccurate, and lacks structure for comprehension."

The pitch against Salesforce

The product centers on what Lightfield calls a temporal context graph, a data model that builds itself from emails, calls, calendar entries, Slack messages and LinkedIn activity rather than requiring manual setup, according to Andreessen Horowitz's investment announcement. Alex Rampell, a general partner at the firm who leads its roughly $1 billion Apps investing practice, said "Salesforce defined it for the cloud era, and Lightfield is defining it for the agent era."

Rampell, Joe Schmidt and Fabrizio Serafini, the three a16z partners who backed the deal, wrote that "Siebel was born on client-server, Salesforce on cloud, HubSpot on inbound SMB," arguing each prior generation of customer software won over a set of companies the incumbent could not see, rather than converting users away from the last one. Salesforce, founded in 1999, remains the dominant CRM by revenue; Lightfield's bet is that a system built around autonomous software, not human sales reps clicking through forms, requires different underlying data structures rather than a new interface bolted onto the old ones.

The bet lands the same week Meta shipped its own consumer AI agent, Muse, into an uneven public launch, part of a broader rush by well-funded companies to put agents in front of paying users before the underlying reliability problems are fully solved.

From Tome to Magical Tome

Lightfield's corporate name is Magical Tome Inc., a holdover from a previous company Peiris built with co-founder Henri Liriani: the AI presentation app Tome, which reached 25 million users before the pair moved on to build a CRM instead, according to the a16z announcement.

The agent harness underneath the CRM

Beyond the data layer, Lightfield ships a standardized software development kit that runs agent code inside a sandbox with built-in quality monitoring, according to the PR Newswire release announcing the round. The company also exposes the platform through a conventional API, a command-line interface and the Model Context Protocol, the specification that lets outside AI agents call a piece of software directly rather than through a human clicking buttons in a browser. That combination is the company's answer to a problem Peiris described in the same statement: agents that can reason well but act on records other software has left incomplete or contradictory.

What the round doesn't say

Neither Lightfield nor its investors disclosed a valuation for the round, a detail Harvey included when it raised $550 million this week. Lightfield also did not say how many of its 5,000 sign-ups are paying customers, or what share of those have finished migrating off Salesforce rather than simply trialing the product alongside it.