The AI-Native CRM

Episode Summary

Executive Summary: Keith Perez described how Tome’s explosive AI presentation growth gave way to a pivot into Lightfield, a business world model for CRM that unifies emails, meetings, docs, product usage, and warehouse data. The discussion centered on why schema-less, intelligence-first systems can outperform traditional CRMs, how Lightfield pricing and product strategy evolved, and why the company is optimizing for expansion, speed, and real customer pain rather than legacy workflows.

Main Topics: From Tome’s viral growth to a hard pivot (Priority: 5/5): Keith explained that Tome reached massive usage quickly, but the team didn’t love the product and couldn’t see a durable path for high-end users. The pivot was driven by product conviction, not just metrics, and by limits in model context for presentations. Why Lightfield is a “business world model” (Priority: 5/5): Lightfield was framed as a CRM reimagined around a canonical activity log that models the relationship between companies and customers across every touchpoint. The core thesis: intelligence and context matter more than rigid schema. Schema-less architecture and semi-structured data (Priority: 5/5): The team moved from trying to be fully unstructured to a semi-structured activity-log approach that preserves context while staying queryable. This lets Lightfield infer causality and update traditional CRM fields automatically. Greenfield vs. brownfield CRM adoption (Priority: 4/5): The conversation contrasted new-company adoption with entrenched incumbent environments. Lightfield initially focused on startups to prove value, then found expansion use cases and network effects that help it move into more established accounts. Pricing and monetization in AI software (Priority: 4/5): Lightfield tested seat pricing and pure consumption pricing before landing on a hybrid model: platform fee plus seats for core CRM, consumption for higher-value AI work like enrichment, automation, and forecasting. Culture, speed, and operating model (Priority: 4/5): Keith described a flat, highly cross-functional team where everyone owns product and customer success, planning is continuous, and ship speed is balanced with rigorous internal bug bashes before release. AI as a company brain and decision engine (Priority: 4/5): The team’s longer-term vision is to help companies make harder strategic decisions—hiring, product direction, expansion timing—by turning Lightfield into a real-time mirror and scenario-planning tool.

Key Arguments: A founder must love the product they are building; no amount of growth can compensate for deep product mismatch. Traditional CRMs fail because they store fragmented, conflicting data instead of a coherent model of customer reality. A canonical activity log is a better primitive than rigid schema because it can reconstruct relationships and update fields from evidence. AI in CRM should not only do tasks like outbound and lead scoring; it should help model the business itself. A semi-structured design is the practical middle ground between fully structured databases and fully unstructured retrieval. Startup CRM adoption is easier in greenfield companies, but real scale comes from expansion into larger accounts and proving referenceability. Pricing should reflect distinct work types: core record-keeping, pipeline generation, workflow automation, and strategic intelligence/forecasting. In fast-moving AI markets, continuous planning and a low bar to start projects are necessary, but the bar to ship must remain high. Cross-functional generalists and shared ownership are better suited to AI-era product velocity than rigid departmental swim lanes. The biggest threat to Lightfield is speed—specifically, preventing customers from feeling the need to move back to legacy systems.

Data Points: Tome users: 25 million users - Referenced as the scale Tome reached before the pivot Tome monthly usage: 2 million users a month - Described as the explosive launch scale for the presentation product Lightfield Series A: $47 million - The company recently raised a Series A led by a16z Founding team background: 3 of 5 founding members from Facebook or Meta data teams - Used to explain the team’s instinct for modeling relationships and activity logs Early pilots: 12 pilots - They initially tested the product with sales teams at 500–1,000 person companies Initial startup test group: 10 startups - Used the office-space incentive to get early CRM users when rebuilding the product Company size: 40 people - Current team size mentioned when describing the operating model Legacy pivot window: about 4 months - Time spent building the first version of the reimagined CRM in the dark Planning horizon: 3-year time revisit - Lightfield prioritizes accounts based on expansion potential over a multi-year horizon Security/healthcare diligence: years of penetration testing - Used to explain how Lightfield won healthcare and health-tech deals Example customer outcome: within days - Lightfield helped someone with Alzheimer’s find frontier treatment within days through Power

Pivotal Quotes: "If we can help you completely model your business and your customer reality, then the rest will be easy." — Keith Perez: Explaining Lightfield’s core thesis for replacing rigid CRM workflows with a business world model "Intelligence is greater than schema." — Alex Rampell: Summarizing the company’s design philosophy after discussing Lightfield’s schema-less architecture "Almost none of the noise around you matters when you're in a pivot. You just need to find pain." — Keith Perez: Advice to founders considering or undergoing a major pivot

Implications: The episode suggests AI-native systems of record will win by modeling reality, not just storing it. For founders, the lesson is to follow real pain, build for expansion, and use AI to create decision leverage, not just automation.