For forty years, CRM has asked people to turn business relationships into fields: name, company, opportunity value, deal stage, notes. Artificial intelligence is challenging this architecture, because an agent that must operate autonomously cannot rely on what a salesperson remembered to type in last night. Lightfield was born precisely out of this fracture and has just raised $47 million in a Series A led by Andreessen Horowitz.
The company is the second act for the founders of Tome, the AI presentation app that reached 25 million users. Keith Peiris and Henri Liriani abandoned that product, drastically downsized the team, and rebuilt the company around a CRM designed for agents. Since its launch last November, Lightfield claims that more than 5,000 companies have signed up and that some customers are migrating entire Salesforce implementations.
Traditional CRM records what happened
Salesforce and HubSpot have become essential infrastructure because they organize customer relationships. But their fundamental unit is the record updated by a person or an integration. Emails, phone calls, and meetings, by contrast, contain far more context than what ends up in structured fields.
Lightfield attempts to automatically build a representation of the company based on interactions. Its temporal context graph preserves not only the current state of a deal, but how and why that state changed. This is a critical distinction when the reader is no longer just a manager, but software tasked with deciding which action to take.
Agents need company memory
An agent can write a flawless email and still cause damage if it does not know that the client just asked to pause talks, that there is an open technical issue, or that the procurement lead has changed. The bottleneck therefore becomes context, not text generation.
Lightfield's thesis is that the system of record must update continuously from calls and emails, giving agents a shared memory. APIs, MCP, and command-line tools then allow other automated systems to query that memory and take action.
Tome's pivot tells the story of AI after the wow factor
The company's story is almost as interesting as the product. Tome went viral at a time when creating slides with AI seemed like a whole new category. However, the rapid commoditization of content generation made that advantage difficult to defend. The founders chose to walk away from a large user base to pursue a deeper problem.
It is a lesson for the entire market: in AI, a spectacular feature can become a commodity in a matter of months. Systems that own data, workflows, and operational relationships are much harder to displace.
Salesforce will not disappear tomorrow
The challenge remains massive. Salesforce possesses distribution, an ecosystem, and integrations built over decades; HubSpot and new contenders are also adding agents. Migrating a CRM is one of the most delicate operations for an enterprise, because it means moving commercial memory.
Lightfield must therefore prove that an agent-first architecture delivers a significant enough advantage to justify the risk of migration. The $47 million buys time to prove it, not victory.
The enterprise's next operating system could be context
If agents truly take on an increasing share of sales, support, and operations, the most important software will not necessarily be the one that generates the best response. It will be the one that knows what is happening across the enterprise and can grant each agent the proper context.
This is where the CRM battle turns into an infrastructure battle. Lightfield does not simply want to add AI to the commercial database: it argues that the commercial database, as we know it, was designed for the wrong worker.
