Tome founders pivot from presentation app to AI-native CRM Lightfield.

2 days ago7 min read5 comments

In a move that speaks volumes about the current trajectory of enterprise software, the founders behind the once-viral presentation tool Tome have executed a dramatic pivot, abandoning their 20 million users and $43 million in funding to launch Lightfield—an AI-native customer relationship management platform that fundamentally rethinks how businesses interact with customer data. This isn't merely another SaaS product; it's a philosophical and architectural bet that large language models have matured sufficiently to displace the structured relational databases that have underpinned business software since the mainframe era.Keith Peiris, Lightfield's CEO, frames the problem with characteristic clarity: traditional CRMs like Salesforce and HubSpot, despite generating billions in revenue, represent perhaps the most universally despised category of enterprise software, plagued by manual data entry, rigid schemas, and a complete failure to capture the nuanced, unstructured reality of human relationships. Lightfield's approach is architecturally distinct from the ground up.Instead of forcing sales teams to compress rich customer conversations into predefined dropdown menus and custom fields, the system acts as a passive intelligence layer, automatically ingesting and transcribing every customer interaction—emails, sales calls, product usage data—and storing them in their raw, lossless form. This creates a comprehensive 'relationship timeline' for every account.When structured information is needed, Lightfield's AI models perform on-demand extraction, allowing the data model itself to be fluid and adaptable, a stark contrast to the schema-locked legacy systems that force companies to live with decisions made during their earliest, least-informed days. The early results, while anecdotal, are striking.Tyler Postle of Voker. ai reported that Lightfield's AI agent helped him revive over 40 stalled opportunities in a single two-hour session—leads that had languished for six months in HubSpot, a tool he described as turning him into a 'data hygienist' rather than a closer.Another user, Radu Spineanu of Humble Ops, highlighted the system's ability to proactively identify neglected follow-ups, a feature he credits with preventing at least three deals from going cold this quarter alone. This points to a deeper shift: AI is moving from being a feature within a software category to being the foundational architecture of the category itself.The strategic bet here is that incumbents like Salesforce, despite their vast resources and frantic announcements of AI features, are architecturally incapable of retrofitting their decades-old, database-centric platforms to operate in this new paradigm. Their core product is a structured database; any AI they add is necessarily a layer on top of that constrained data model.Lightfield, by contrast, is built from the first line of code on the premise that the primary record *is* the unstructured conversation, with structure being a derived, ephemeral view. This creates a genuine technical moat, but it also introduces significant risks that Daniel Reed, an AI researcher, would be quick to identify.Storing complete conversation histories raises profound privacy and data sovereignty questions, though the company claims SOC 2 compliance and a strict no-training policy on customer data. More critically, the reliance on LLMs for interpretation and action introduces the ever-present specter of hallucination—the generation of plausible but incorrect information.Peiris acknowledges this inherent unreliability, and Lightfield's design reflects a cautious, augmentation-focused philosophy, requiring human approval for critical actions like sending emails or updating deal stages. This is a wise concession to the current state of the technology, positioning the system as a copilot that amplifies human judgment rather than an autonomous agent that replaces it.The company's trajectory will serve as a crucial case study for the entire enterprise AI sector. Its initial beachhead—over 100 Y Combinator startups—is a classic disruption strategy: capture the next generation of companies before they develop entrenched habits and vendor relationships.The ultimate test, however, won't be technical feasibility but human trust. Will sales teams, whose livelihoods depend on commission checks, ever trust an AI's interpretation of a customer's tone or its drafted follow-up email enough to base multimillion-dollar decisions on it? The success of Lightfield hinges on the answer to this question, determining whether AI-native architecture is a compelling differentiator or merely a technological solution in search of a problem that salespeople are willing to let it solve.

#Lightfield

#AI-native CRM

#Salesforce competitor

#enterprise software

#generative ai

#featured

Stay Informed. Act Smarter.

Get weekly highlights, major headlines, and expert insights — then put your knowledge to work in our live prediction markets.