Lightfield Raises $47M to Build the Agent-Era CRM
Andreessen Horowitz just led a $47 million Series A into Lightfield, a company making a direct argument that Salesforce and HubSpot are structurally incompatible with AI agents. The funding, announced September 9, 2026, signals that at least one top-tier investor believes the next system of record for enterprise sales will not be a legacy CRM with AI features tacked on. It will be a platform built from the start so agents can understand and act on customer data reliably.
This matters for any organization that has started deploying AI agents across GTM: if your agents are reading stale, manually entered Salesforce fields, they cannot produce work you can trust.
Why Legacy CRMs Break for Agents
The core problem Lightfield is solving is context starvation. Every popular CRM in use today was designed for human users: a person reads the data, forms a judgment, takes action, and then logs what happened. The log is always incomplete, often days late, and reflects what the person remembered, not what was actually said.
When you route an AI agent through that same record, the agent inherits every gap. A prospecting agent told to personalize outreach based on “last conversation notes” is reading a field that may be empty, wrong, or three weeks old. A deal coaching agent told to flag risk factors cannot flag what was never logged.
According to Lightfield’s announcement, the problem is structural: “When layered onto legacy CRMs like Salesforce or HubSpot, agents inherit incomplete fields, stale close dates, and notes written from memory after the call, and produce work companies cannot verify or trust. Solving this requires a new system of record designed for deep understanding, not AI features added to the old one.”
This is an honest description of where most enterprise AI deployments currently sit. The model is capable. The context it receives is broken.
What Lightfield Actually Builds
Lightfield’s approach rests on four design choices that separate it from adding AI features to existing CRMs.
Automatic capture. Lightfield ingests every customer interaction across email, calendar, calls, Slack, and LinkedIn and structures it into the record without requiring human input. No one logs a call after the fact. The transcript, sentiment, and commitments made are captured and structured as they happen.
A business world model. Rather than a flat table of fields, Lightfield builds a connected representation of every customer: what was said, who said it, how it changed the state of the deal, and how it connects to other contacts, accounts, and opportunities. The company calls this a “business world model,” and it is the substrate agents reason from when asked to predict what happens next or to diagnose why a deal stalled.
An agent harness for consistent output. Agents that access Lightfield operate through a standardized SDK, run analysis in a code sandbox, and are monitored by automated evaluations. The company claims the same request returns the same quality of output each time, not a variable result depending on what context happened to be available. For organizations trying to build reliable agentic workflows, consistency is the actual problem, not raw capability.
An open platform. Every record in Lightfield is readable and writable through API, MCP, and CLI. The company has also shipped an MCP server as an official connector in Claude’s Connectors Directory. This means any tool in an organization’s stack can access Lightfield’s data under proper permissions, and any agent can write back to the record, not just read from it.
The Investor Signal
Alex Rampell, General Partner at Andreessen Horowitz, framed the thesis plainly in the announcement: “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. Companies building with agents need more than a CRM with AI features. They need a system designed from the ground up for agents to work from.”
The comparison to Salesforce is deliberate and historically grounded. Salesforce’s dominance came not from being the best database, but from being the first CRM built for the web, the cloud, and the browser-based workflow that came with it. The bet here is that the shift to agentic workflows is the same kind of platform transition, and that legacy CRMs cannot adapt fast enough to own it.
Participating investors alongside a16z include Maverick Capital, Coatue, Audacious, Alumni Ventures, Greylock, and Lightspeed Venture Partners. The roster covers both consumer and enterprise-focused growth investors, suggesting confidence that Lightfield’s current early-stage customer base can expand into mid-market and enterprise accounts.
Traction and Trajectory
Lightfield launched publicly in November 2025 and has accumulated more than 5,000 company sign-ups in under a year. The founders, Keith Peiris and Henri Liriani, previously built Tome, an AI presentation tool, before shutting it down in mid-2025 and pivoting to this problem. Their product cadence since launch has been aggressive: native outbound sequencing, Slack and Teams integration, dashboards, formula fields, and an agent builder have all shipped in the months since launch.
The company’s changelog shows Claude Opus 5 and Sonnet 5 added as selectable models for automations, which tracks with the broader pattern of enterprise GTM tools plugging into frontier model APIs rather than training proprietary models.
What This Means for Enterprise GTM Teams
For organizations still running agents on top of Salesforce or HubSpot data, this funding round is a useful signal about where the market is heading. It does not mean you need to migrate your CRM today. It does mean that the problem Lightfield is solving, unreliable agent output caused by incomplete context, is real and growing more expensive as more agentic work gets deployed.
The table below summarizes the architectural differences between legacy CRMs and Lightfield’s agent-native design:
| Dimension | Legacy CRM (Salesforce, HubSpot) | Agent-Native CRM (Lightfield) |
|---|---|---|
| Data capture | Manual, human-logged | Automatic, always-on |
| Record completeness | Depends on rep discipline | Structural, all interactions captured |
| Agent access | Read-only via API, stale fields | Read/write via API, MCP, and SDK |
| Agent output consistency | Variable (depends on context quality) | Monitored by evals, consistent |
| Update latency | Hours to days | Near real-time |
| Designed for | Human-driven workflows | Human and agent workflows |
| Open platform | Partial (API with limits) | Full API, MCP, CLI |
The pattern here aligns with what Enera has documented in its work on autonomous GTM systems: the bottleneck for agentic GTM is almost never the model. It is the quality and structure of the data the agent operates on. A prospecting agent built on a world model of every customer interaction outperforms the same agent built on a legacy CRM by a margin that compounds over time.
The Broader Context for September 2026
Lightfield’s raise lands in the same week as several other AI infrastructure rounds and model launches, but it is notable for its specificity. Most AI funding in 2026 has targeted model training, inference infrastructure, or security layers. Lightfield is betting on the data layer: the structured, agent-readable representation of what is actually true about a business’s customer relationships.
This is consistent with a broader shift visible in the market. As enterprise AI leaders described in Salesforce’s agentic enterprise adoption data from earlier this year, the organizations seeing the highest ROI from agents are the ones that invested in structured, clean context before deploying agents, not after.
The $47 million is small by 2026 AI funding standards. The bet it represents is not small: that the system of record for the next decade of enterprise software will be agent-native from the foundation, not retrofitted.
For GTM leaders evaluating whether to build agent workflows on existing CRM infrastructure or to consider purpose-built alternatives, this round is a data point worth tracking. The thesis is coherent, the early traction is real, and the investor base is credible.
If you are building agentic GTM systems and want to evaluate what a context-first approach looks like for your organization, Enera works with enterprise teams on exactly this problem.
Sources: Lightfield Series A announcement, PRNewswire, September 9, 2026 | Lightfield product page | Adam’s GTM Report: Lightfield brand coverage