The Pivot from Presentation to Context

Keith Peiris, CEO of Lightfield, describes his transition from building Tome—a generative presentation tool that reached 25 million users—to founding Lightfield. Despite Tome's explosive growth, Peiris realized the product was limited by its inability to capture the deep context required for professional storytelling. He argues that founders must be willing to abandon products they don't personally love or find indispensable, even when those products show high vanity metrics.

Intelligence Over Schema

Lightfield’s core thesis is that "intelligence is greater than schema." Traditional CRMs like Salesforce rely on rigid, pre-defined data structures (schemas) that require manual entry and constant maintenance. Peiris argues that this approach is fundamentally broken because it forces users to conform to a database's limitations rather than capturing the reality of a business relationship.

Lightfield replaces this with a "business world model." By treating every email, call, meeting, and product interaction as an entry in a chronological activity log, the system builds a canonical record of the relationship. This allows AI agents to reason across disparate data points, enabling complex workflows—such as identifying expansion opportunities or matching clinical trial participants—that would be impossible in a traditional, siloed CRM.

Architectural Primitives

To solve the "needle in the haystack" problem inherent in unstructured data, Lightfield employs a semi-structured approach. It stores vast amounts of unstructured data in an activity log but uses an underlying architecture that allows the system to infer causality. This setup enables a "schemaless" user experience: customers connect their data sources (email, data warehouses, call recorders), and the system assembles the business reality in real-time, allowing users to define fields and stages dynamically rather than upfront.

Peiris discusses the difficulty of entering the "brownfield" CRM market, where incumbents hold "hostages" rather than customers. He emphasizes that early-stage success in such markets requires a willingness to ignore traditional product categories and follow the "heat" of user demand. By offering a product that provided immediate value—even when it was unfinished—Lightfield was able to secure early adopters who were willing to provide hourly feedback, effectively using "negative pricing" (offering office space in exchange for product usage) to build a feedback loop that incumbents lack.

Key Takeaways

  • Follow the Heat: When pivoting, look at your existing user base to see where they are trying to use your product for unintended, high-value tasks.
  • Intelligence > Schema: Don't force users to fit their reality into rigid database fields. Build systems that can reason over unstructured activity logs.
  • Build for the End State: Focus on the "business world model" rather than just creating a repository for data entry.
  • Embrace Naivety: Being an outsider to a legacy industry can be an advantage, as it allows you to question fundamental assumptions about how software should work.
  • Prioritize Velocity: In the AI era, shipping speed and the ability to iterate based on real-world usage are more important than perfecting a feature set before launch.