Lightfield says it is building a different kind of CRM for AI agents, and a $47 million Series A led by Andreessen Horowitz puts investor money behind that pitch. For RevOps leaders, CRM administrators, founders, and CTOs, the question is not whether the product has AI features, but whether replacing a legacy system of record is becoming a real architectural option.
What changed
SiliconANGLE reported that Lightfield raised $47 million in a Series A announced September 9, 2026, led by Andreessen Horowitz with Lightspeed Venture Partners, Coatue, Greylock, Maverick Capital, Audacious, and Alumni Ventures participating.
SiliconANGLE also reported that Lightfield is officially known as Magical Tome Inc., and positioned the company as an AI-native CRM startup with ambitions to take on Salesforce and HubSpot. The legal name points to the founding story: Andreessen Horowitz says co-founders Keith Peiris and Henri Liriani previously built Tome, an AI presentation tool that reached 25 million users, then pivoted, cut the team to a core group of engineers, and rebuilt around the missing AI-native CRM they had hit while scaling Tome. Buyers are therefore evaluating a roughly year-old product from a team on its second act, not an established platform.
Why B2B teams should care
The core product thesis in the reporting is architectural. SiliconANGLE said Lightfield believes legacy CRM architecture is ill-suited to AI agents, which need a fundamentally different system of record — so the company rebuilt the CRM from scratch rather than layering a copilot onto an existing one. Peiris told SiliconANGLE that agents fail today not because the models do not work, but because the data they are given is incomplete, inaccurate, and lacks the structure agents need.
SiliconANGLE described four technical differentiators, and they are the most checkable part of the pitch: a self-updating record that continuously ingests email, calendar, Slack, and LinkedIn interactions; a “world model” mapping those interactions to accounts and deals and recording how they change; an agent harness that forces agents through a standardized SDK and a monitored code sandbox; and an open design, readable and writable through APIs, command line interfaces, and the Model Context Protocol.
Andreessen Horowitz, writing as Lightfield’s lead investor rather than an independent validator, added that the product uses a flexible data model and a temporal context graph, and that agents and humans work from the same platform through the same API. General partner Alex Rampell stated the bet directly: “Every platform shift produces a new system of record. Salesforce defined it for the cloud era, and Lightfield is defining it for the agent era.” That is the investor’s framing, not a settled outcome — the same system-of-record question is contested a layer below in the enterprise agent architecture hyperscalers are assembling.
Who is affected
The primary audience is Salesforce-centric sales and RevOps teams already weighing whether AI belongs inside the CRM record or in an overlay. SiliconANGLE reported that Peiris said dozens of companies ditched Salesforce and replaced it with Lightfield.
Andreessen Horowitz says teams are switching from Salesforce, HubSpot, and Attio, which widens relevance beyond the large-enterprise Salesforce base.
Traction is early by enterprise standards. SiliconANGLE reported Lightfield has signed up more than 5,000 companies since launching in November 2025, from early-stage startups to scaling enterprises. But every traction figure here originates with the company or its lead investor, and the wording differs between them: SiliconANGLE reports more than 5,000 companies “signed up,” while Andreessen Horowitz says “thousands of companies have adopted” the platform. Signups and production adoption are not the same measure, and no independent audit of either exists in the current coverage.
What teams should check now
Teams evaluating Lightfield AI CRM should treat this as a system-of-record question first and an AI-feature question second.
- Check whether the current CRM data model is structured enough for agents at all, since that premise carries the whole pitch. If agent output is unreliable because records are stale rather than because the model is weak, confirm that diagnosis before it becomes a replacement project.
- Test the openness claims in a trial rather than on the deck: read and write real records through the API, the CLI, and MCP, and ask what the agent harness and code sandbox actually restrict. Those are demonstrable in an afternoon, unlike the architecture argument.
- Map migration requirements before treating this as a near-term replacement. Lightfield’s product page, as summarized by Unite.AI, describes a white-glove migration in which its agent ingests legacy CRM data and recreates the data model, plus two-way sync for running alongside Salesforce or HubSpot. Those are vendor claims to put under contract, not verified outcomes.
- Verify whether the use case needs a full CRM replacement rather than an overlay. Replacing a system of record moves reporting, integrations, and compensation data with it, and the incumbent cost it displaces is rarely just the seat price — the pattern in our Salesforce pricing guide.
What remains unclear
- Not yet confirmed: valuation, revenue, or total funding raised across all rounds.
- Not yet confirmed: detailed enterprise customer names or named reference deployments.
- Not yet confirmed: how many of the 5,000-plus signups are paying, in production, or running Lightfield as the primary system of record.
- Not yet confirmed: independent verification of any traction or migration claim; all trace back to the founder or lead investor.
- Not yet confirmed: pricing, beyond Andreessen Horowitz saying it follows work done and value delivered rather than static seats.
What to watch next
The next meaningful proof point is not another architecture claim. It is traction evidence beyond signups: named deployments and migration case studies showing where Lightfield replaced Salesforce, HubSpot, or Attio inside live revenue operations.
Buyers should also watch for production evidence that agents are handling updates, follow-ups, handoffs, and reporting inside the same environment human reps use — the areas Andreessen Horowitz named.
Sources
- SiliconANGLE, AI-native CRM startup Lightfield raises $47M to build an AI agent-ready replacement for Salesforce
- Unite.AI, Lightfield Raises $47M Series A Led by a16z to Accelerate Growth
- Digital Today, Lightfield raises $47 million in Series A for AI agent CRM
- Andreessen Horowitz, Investing in Lightfield
