The AI-Native CRM
Joe Schmidt, Alex Rampell, Keith Peiris
AI-assisted summary of The AI-Native CRM — original source. Check the original for full context.
Source published:
Key ideas
- After scaling presentation tool Tome to 25 million registered users, co-founder Keith Peiris chose to pivot away due to lack of founder conviction and the fundamental limitation that isolated prompts lacked relational business context. (0:54)
- Lightfield raised a $47 million Series A led by a16z to build an AI-native CRM centered on a 'business world model' that replaces rigid schemas with semi-structured, chronological activity logs. (0:00)
- To seed its initial alpha, Lightfield offered free San Francisco office space to 10 startups on the condition that they run their daily sales operations entirely on the unproven platform. (6:40)
- Rejecting purely seat-based or 100% consumption billing, Lightfield adopted a hybrid pricing model that bundles core CRM upkeep into a predictable platform fee while charging usage for pipeline generation, automations, and strategic forecasting. (33:15)
Detailed notes
## Background Published on September 16, 2026, across RSS and web podcast platforms, this episode features a16z partners Alex Rampell and Joe Schmidt hosting Lightfield co-founder and CEO Keith Peiris. The discussion focuses on Peiris's decision to walk away from Tome after reaching 25 million users to build Lightfield, an AI-native CRM engineered around a unified 'business world model' that replaces rigid relational database schemas with contextual machine intelligence.
## TL;DR Keith Peiris explains how the limitations of Tome's presentation tooling led his team to uncover a deeper enterprise challenge: reconciling fragmented, conflicting customer data across emails, call recorders, and stale CRM records. Lightfield addresses this by modeling company relationships chronologically through semi-structured activity timelines, allowing AI agents to dynamically deduce context without manual data entry or upfront schema design. The conversation delves into Lightfield's journey from offering free office space to alpha users ('negative pricing') to penetrating brownfield markets through cross-functional non-sales adoption. Peiris also details the operational shift to a 40-person team of un-siloed generalists, the rationale behind hybrid usage-and-platform pricing, and the long-term vision of turning the CRM into an autonomous strategic scenario engine.
## Atmosphere & style The conversation carries an energetic, intellectually rigorous insider vibe, balancing technical architecture debates with candid operator post-mortems. Peiris speaks with refreshing vulnerability about the emotional toll of abandoning a viral hyper-growth product and ignoring organizational noise, while Rampell peppers the dialogue with wry historical analogies ranging from vintage BMW switchboards to early SQL character limits. The dynamic feels like a high-bandwidth, collaborative whiteboard session between trusted investment partners, defined by pragmatic realism rather than empty AI hype.
## Bulletin - `00:54` After scaling presentation tool Tome to 25 million registered users, co-founder Keith Peiris chose to pivot away due to lack of founder conviction and the fundamental limitation that isolated prompts lacked relational business context. - `00:00` Lightfield raised a $47 million Series A led by a16z to build an AI-native CRM centered on a 'business world model' that replaces rigid schemas with semi-structured, chronological activity logs. - `06:40` To seed its initial alpha, Lightfield offered free San Francisco office space to 10 startups on the condition that they run their daily sales operations entirely on the unproven platform. - `33:15` Rejecting purely seat-based or 100% consumption billing, Lightfield adopted a hybrid pricing model that bundles core CRM upkeep into a predictable platform fee while charging usage for pipeline generation, automations, and strategic forecasting.
## Key takeaways - `00:54` Founders must personally love their product; viral metrics like 25 million users cannot substitute for lack of conviction or enterprise utility. - `00:54` Waiting on foundation model capability improvements ('hopium') fails when the bottleneck is missing enterprise context rather than raw reasoning power. - `05:00` Enterprise data systems present conflicting realities, where call recordings, emails, and CRM records often actively contradict one another. - `06:40` Intense user anger and hourly critical feedback from daily alpha users is a stronger signal of real utility than passive consumer growth. - `10:00` Treating human and organizational relationship timelines as the primary primitive allows semi-structured logs to replace rigid relational database schemas. - `15:00` 'Intelligence is greater than schema': systems that capture raw activity allow future data modeling changes to be backfilled retroactively via LLMs without expensive upfront data modeling. - `17:40` Winning greenfield startups provides a faster initial path to scale than competing directly for entrenched legacy 'hostage' customers. - `20:00` Commoditized outbound email automation is an unsustainable wedge against incumbents; the enduring advantage lies in modeling holistic company reality. - `22:15` Granting free CRM seats to cross-functional departments like engineering and finance builds internal network effects that insulate the tool against incoming sales leaders demanding legacy software. - `25:10` Enterprise software should reject aesthetic dogmatism, offering familiar tables and dashboards alongside frontier natural-language and CLI interfaces. - `33:15` Purely outcome-based pricing is impractical in sales tooling because software cannot control the underlying product-market fit of its customers. - `35:46` Operating without rigid departmental swim lanes and running a unified 40-person standup with a stack-ranked backlog enables extreme shipping velocity. - `40:00` Insurgent platforms must engineer maximalist features tailored to their fastest-growing breakout customers to stop them from churning back to legacy enterprise vendors. - `48:15` Attempting to build homebrew CRMs or internal 'company brains' consistently fails due to the immense complexity of data synchronization, entity resolution, and context engineering. - `50:00` Executing a successful pivot requires ruthlessly tuning out internal organizational noise and peripheral grievances to focus exclusively on solving verified customer pain.
## Detailed walkthrough ### `00:00` Building Lightfield and the Pivot from Tome
Three out of the five founding members of Lightfield originated from Facebook, bringing a background in consumer product design to enterprise software. Keith Peiris and his team gravitated toward building software for revenue teams because revenue generation represents a universally shared, high-stakes problem that every business cares about and constantly strives to improve. Today, Lightfield operates as an AI-native CRM built around the concept of a "business world model," which digests unstructured communications—such as customer emails, sales calls, and meeting notes—and transforms them into a grounded, unified record that autonomous AI agents can reason across to perform real-world work.
```mermaid flowchart LR accTitle: Unstructured Data to Actionable Business Model accDescr: Pipeline showing raw inputs transformed by Lightfield into an actionable record for AI agents. subgraph Inputs["Disparate Business Inputs"] E["Emails"] C["Sales Calls"] M["Meeting Transcripts"] end subgraph Engine["Lightfield Business World Model"] R["Resolution & Context Reconciliation"] end subgraph Execution["Downstream Work"] A["AI Agents & Workflows"] end Inputs --> Engine Engine --> Execution ```
Illustrating the impact of this approach, Keith highlights Power, a marketplace that connects patients suffering from complex, chronic illnesses with frontier clinical treatments. Power modeled their entire complex patient matching workflow inside Lightfield, which successfully helped an Alzheimer’s patient locate and access an experimental treatment within days.
Alex Rampell frames the broader thesis behind a16z leading Lightfield’s $47 million Series A round as a fundamental shift where "intelligence is greater than schema," moving enterprise software away from brittle, human-maintained relational tables toward systems powered by probabilistic reasoning.
- Keith emphasizes the competitive reality of enterprise software: “As a CRM company in a red ocean space, we have to be an expansion company.” - The core operating philosophy: “If we can help you completely model your business and your customer reality, then the rest will be easy.”
---
`00:54` Why 25 Million Users Were Not Enough
Before founding Lightfield, Keith built Tome into a viral phenomenon, achieving explosive consumer adoption right as GPT-3 and ChatGPT launched. Coming from consumer backgrounds where they worked on selfie and camera design at Instagram and Messenger, the founders originally set out to modernize how people share ideas. Tome allowed users to generate complete presentations and visual documents using generative AI, scaling at a clip of 2 million new users per month to reach 25 million registered users, with server waitlists forming due to severe inference compute shortages.
```mermaid flowchart TD accTitle: Tome Dilemma accDescr: The trade-off between massive viral growth and product conviction that led to Tome's pivot. V["Viral Growth: 25M Users"] --> P["Product Reality Check"] P --> C1["Individual / Student Utility"] P --> C2["No Context on Relationship or Audience"] C2 --> D["Decision: Hard Pivot Away"] ```
Despite the headline metrics, Keith and his co-founders made the radical decision to walk away entirely from the hyper-growth product. The core issue was an absence of founder conviction: the founders realized that they did not actually like using the product themselves, and they could not make presentations that met the standard of discerning enterprise professionals, investment bankers, or management consultants writing serious memos.
When Alex Rampell asks why the team did not simply wait for models to improve along the curve from GPT-3 to GPT-4 and beyond—warning against the trap of "hopium"—Keith explains that raw foundation-model reasoning was never the real bottleneck. The fundamental ceiling was that an isolated presentation prompt possessed zero context about the human dynamics in play.
- Keith on founder alignment: “Deep down, I think at an instinctual level, none of us liked the product, which is kind of a funny thing to say. I think at the end of the day, if you're a founder, you have to love the product that you're building and you have to be excited for your customers to use it.” - Keith explaining why waiting on better LLMs would not fix the core problem: “The model just didn't have enough context to really understand the audience, the relationship between the presenter and the audience, and no amount of general reasoning gets you past that.”
---
`05:00` Discovering Disparate Realities Across Enterprise Stacks
To pivot, the team went back to their founding mission of helping professionals tell difficult, high-stakes stories. Sifting through their 25 million users, they isolated the segment generating genuine commercial value: B2B sales and marketing teams. Keith secured free pilots across 12 mid-market companies ranging from 500 to 1,000 employees, originally offering to draft new business pitches and proposals for free.
Once inside these revenue teams, customer demands immediately drifted beyond slide decks into operational territory. Sales teams asked them to conduct account research, qualify inbound leads, and identify accounts primed for expansion. Keith asked for access to what he assumed was their operational single source of truth—the CRM:
```mermaid flowchart TD accTitle: Data Reconciliation Problem accDescr: Fragmented tools feeding contradictory realities into a central system of record. subgraph Siloed["Fragmented Business Systems"] CRM["CRM / Salesforce: Stale & Manual"] CR["Call Recorders: Unstructured Ground Truth"] DW["Data Warehouses: Event Metrics"] EM["Email Threads: Multi-threaded Conversations"] end Siloed -->|"Conflicting Data"| LF["Lightfield Reconciliation"] LF -->|"Reorganized Reality"| SOT["Ground-Truth Business World Model"] ```
The team quickly uncovered that enterprise business data was broken, contradictory, and deeply fragmented. A call recorder like Gong often reflected an entirely different operational reality than Salesforce; notes in the CRM omitted key deal risks, while email threads revealed procurement blockers that never appeared in pipeline reports. Reconciling this fragmented operational reality emerged as a vastly larger and more foundational problem than generating slides.
- Keith on the realization that sparked Lightfield: “The hardest part of doing this work was actually making sense of all of the data across all of these disparate systems. And it was sort of incomplete. It was conflicting. You know, the call recorder often had a different view of reality than the CRM.” - Keith on the thesis of the new company: “If you can reorganize reality for a company in a way that machines can understand and also for humans to understand, that feels like a way more interesting and enduring company than the one that we're on right now.”
---
`06:40` "Negative Pricing" and Rebuilding the CRM from First Principles
The path from the initial discovery to a full CRM was not an immediate straight line. The team initially built a lightweight go-to-market assistant, but they found themselves commoditized alongside ten other startups competing for the same workflows without owning the underlying data layer. Keith made the call to shrink the team down, discard the assistant product, and spend four months in complete isolation rebuilding the CRM from first principles.
When the alpha version was ready, they confronted the adoption wall: no operating business wants to trust their revenue pipeline to a four-month-old, unproven CRM. Bound to an expensive, oversized office lease leftover from Tome, Keith deployed an unconventional go-to-market hack: he posted on X and LinkedIn offering free San Francisco office space to startups on the condition that they run their sales operations entirely on Lightfield.
```mermaid flowchart LR accTitle: The Negative Pricing Loop accDescr: How trading office space for alpha CRM usage unlocked intense, high-frequency user feedback. A["Unused Office Space"] -->|"Free Desks Offered on X / LinkedIn"| B["10 Startups Move In"] B -->|"Mandatory Daily Usage"| C["Clunky, Slow Alpha CRM"] C -->|"Hourly Slack Feedback"| D["Rapid Iteration & Product Conviction"] ```
Ten startups accepted the offer—what Alex Rampell jokingly terms "negative pricing." The setup generated immediate product-market feedback. While the alpha was slow, buggy, and missing essential enterprise features, the founders sat in the product every single day, pinging the engineering team with impassioned critiques in Slack every two hours. For Keith, this visceral customer frustration was the ultimate validation, serving as a total contrast to Tome’s passive consumer churn.
Alex Rampell points out the classic dichotomy of traditional CRM systems: on paper, they are supposed to be the canonical repository of customer relationships, yet in reality, sales reps resent data entry and leave them hopelessly stale.
- Keith on the visceral feedback from alpha users: “They were mad about everything missing. They were mad about how slow it was. But they were in it every day. And they were giving us Slack feedback about it every two hours. And it's like, oh, this is so different than Tome.”
`10:00` Building the Facebook Timeline for Business
Reflecting on why most CRMs fail to capture reality, Keith Peiris notes that Lightfield benefited immensely from a beginner's mindset. In a typical growth-stage startup, sales representatives treat a CRM as a digital notepad to remember tasks, run low-level automations, and generate sales forecasts. Most founders entering the market rushed to build "CRM for work," automating repetitive sales activities like cold outbound emails. Lightfield went in the exact opposite direction:
> “And I think our naive view is actually the most important thing here is the latter. You know, if we can help you completely model your business and your customer reality, then the rest will be easy.”
Three out of Lightfield’s five founding team members came from Facebook, which shaped their architectural philosophy. Drawing inspiration directly from the early Facebook timeline, they treated the chronological, human relationship between two organizations as the foundational atomic unit of software. Rather than forcing users into standard sales pipeline stages right away, they designed an immutable activity log first: tracking the very first outreach, what was said back and forth, meeting records, shared documents, and eventually product usage telemetry and billing events. This log serves as a canonical truth that subsequently triggers standard CRM functions like updating deal stages and populating metadata fields.
```mermaid flowchart TD accTitle: Lightfield Semi-Structured Architecture accDescr: Raw activity streams feed an unstructured log that intelligence layers convert into structured schema updates.
subgraph Ingestion["Data Ingestion"] A["Emails & Call Recordings"] B["Product Usage & Telemetry"] C["Billing & Warehouse (Snowflake)"] end
subgraph Core["Core Primitive"] D["Canonical Activity Log\n(Semi-Structured Timeline)"] end
subgraph Intelligence["Intelligence & Reasoning Engine"] E["Causal Inference & Traversal"] end
subgraph Output["CRM Interface"] F["Deal Stages & Field Updates"] G["Custom Object Reasoning"] end
A --> D B --> D C --> D D --> E E --> F E --> G ```
- Standard CRM design is crippled by manual rep data entry, brittle schema rigidity, and poor API integrations. - Keith describes Lightfield's core technical goal as building a comprehensive "business world model" rather than simply dispatching emails. - Rather than maintaining a rigid relational schema or an unusable raw data lake, Lightfield settled on a semi-structured activity log that makes causal inference possible.
`12:35` Semi-Structured Logs and the Expansion Use Case
When Lightfield initially experimented with purely unstructured storage, they immediately hit performance barriers. Query latencies exploded because retrieving answers created a massive needle-in-a-haystack problem. To fix this, they engineered a hybrid semi-structured model capable of housing massive amounts of unstructured communication while indexing causality across the activity timeline.
This design fundamentally alters how downstream workflows operate. Keith offers the practical scenario of a Customer Success Manager (CSM) evaluating accounts for expansion. In a legacy CRM, a rep looks at disconnected static fields or runs brittle SQL queries. In Lightfield, a CSM can ask an open-ended question: "What should we sell to them, and when should we sell to them?" Lightfield traverses the customer's chronological communications—such as open support tickets, sentiment, and past meeting notes—and reconciles them with product usage streams pulled directly from warehouses like Snowflake. If an account has not had a single user log in for an entire month, the system knows not to recommend an upsell, or it can prioritize accounts showing intense product engagement across entire departments.
`15:00` “Intelligence Is Greater Than Schema”
Alex Rampell and Keith Peiris dissect the traditional enterprise onboarding nightmare: hiring costly CRM consultants. When surveying these consultants, Lightfield discovered that their most consequential job was helping customers design an initial relational data model. If an implementation team picks the wrong custom fields or sales pipeline stages at the outset, the entire implementation fails—sales reps will never retroactively update historical accounts to match a new schema.
Lightfield opted for an effectively "schemaless" paradigm. Customers do not spend months mapping database entities; they simply plug in their corporate email, connect a meeting call recorder, and sync their data warehouse. Lightfield’s built-in enrichment sources assemble relationships dynamically in five minutes, allowing teams to declare or alter fields at any point in the future. Because the entire history is retained in the activity log, changes can be backfilled retroactively using LLMs:
> “Basically, intelligence is greater than schema. That's the crux.”
Alex Rampell draws a humorous historical contrast to the early days of SQL databases, where systems engineers obsessively managed every byte of storage: - Relational tables required developers to hardcode character limits using data types like `VARCHAR(25)` for a user's name. - If a customer had a name that exceeded 25 characters, the system broke. - Computing has transitioned from manual byte accounting to an era where raw AI intelligence bypasses rigid schemas entirely, offering an enterprise platform that sets up like a consumer product.
`17:40` Clinical Trials and Brownfield vs. Greenfield Markets
Because Lightfield treats objects and relationships as arbitrary, customers use it to run atypical business models that traditional CRMs cannot accommodate. Keith highlights a health-tech customer that operates a two-sided marketplace matching pharmaceutical companies with patients seeking frontier treatments for severe diseases. The company modeled both the B2B pharmaceutical side and the B2C patient intake side in Lightfield. They then built automated agents that scrape clinical trial registries directly from the FDA and clinicaltrials.gov to create a real-time world model of ongoing studies. Lightfield autonomously correlates patient medical conditions with matching open trials, and recently helped a patient suffering from Alzheimer's disease secure access to an experimental treatment in just a few days.
Alex Rampell turns the conversation toward market entry strategy, explaining the classic startup dilemma of greenfield versus brownfield expansion:
> “The best companies have hostages, not customers. Like SAP has hostages. That's a brownfield.”
- **The Brownfield Challenge:** Brownfield markets represent environments heavily entrenched by legacy enterprise incumbents (e.g., SAP in ERP). Selling directly into these organizations is notoriously difficult because existing systems hold workflows hostage. - **The Cloud Paradigm Shift:** The first generation of cloud software conquered brownfields by waiting for legacy infrastructure to rot—such as on-premise Siebel systems running on aging IBM AS/400 mainframes whose sole 92-year-old system administrator had retired—and offering modern web delivery. - **The Greenfield Alternative:** In contrast, a greenfield strategy ignores entrenched legacy customer hostages altogether. Instead, a startup focuses on building a fundamentally superior tool for brand-new companies that are completely untethered from legacy software lock-in.
`20:00` Finding the Brownfield Wedge and Moving Beyond Outbound Spam
When Lightfield first launched, Keith Peiris and his team lacked a razor-sharp thesis on how to crack "brownfield" accounts—established companies entrenched in existing enterprise software. Because customer relationship management is universally cared about and perpetually ripe for improvement, they knew an opportunity existed, but they chose to start by competing for new, greenfield companies. To win those early startups, Keith and his chief of staff personally spearheaded outbound efforts, prospecting on LinkedIn and cold-emailing Y Combinator founders with the pitch: can we beat your existing startup CRM? They believed that by listening intently to these early users, the wedge required to break into brownfield accounts would eventually reveal itself.
That wedge crystallized over several months of observing their customers scale at unprecedented velocity. In the current startup climate, companies progress from pre-seed to seed to Series A and B faster than ever; several Lightfield customers that began with zero sales representatives quickly scaled to over a hundred reps. Watching them operate made it clear that the core value was far more elevated than the prevailing Silicon Valley narrative. For years, billboards across the Bay Area promoted AI CRMs as automated labor bots designed to pump out cold emails. Keith’s team recognized that incumbent software like Salesforce would inevitably automate outbound emails anyway, meaning task-execution features could never serve as a sustainable wedge against legacy systems. Instead, the real challenge was providing a holistic understanding of the business so leadership could steer their organizations through chaotic growth.
Keith Peiris summarized their realization: “And we were like, I think the wedge in Brownfield has to do with better understanding your company.”
- Keith and his chief of staff personally prospected early startup customers via LinkedIn and Y Combinator founder outreach. - Customers scaled rapidly on the platform, growing from zero sales reps to over 100 reps across subsequent funding rounds. - Lightfield rejected the AI billboard trend of focusing strictly on automated outbound tasks, recognizing incumbents would commoditize basic email automation.
---
`22:15` The Seasoned VP of Sales and Internal Network Effects
As a startup matures, a classic tension emerges between greenfield agility and brownfield inertia: founders and technical teams embrace frontier technology, but the seasoned VP of Sales brought in to scale the team has spent years conditioned on legacy software. Alex Rampell recalled encountering this exact obstacle when building one of his earliest companies. Adamant about avoiding Salesforce’s $85-per-month seat cost, Alex initially set up SugarCRM because it was free. Despite the tool working well, he was forced to cave and pay for Salesforce the moment he hired an experienced VP of Sales who flatly refused to work in anything else.
To counter this inevitable friction, Lightfield introduced a specific organizational distribution model: on all sales-led tiers, they provide free access to anyone outside the core sales team. Rather than walling off the CRM inside sales, engineering leaders use it to understand customer feature requests, customer success uses it for account scoring, and finance uses it for revenue recognition. When a newly hired VP of Sales enters the company demanding a return to legacy Salesforce, they are met with pushback from the rest of the organization, which already relies on Lightfield as an operational nervous system. This cross-departmental dependency creates powerful internal network effects that insulate the software from being abruptly ripped and replaced.
Alex Rampell highlighted the stubborn pull of legacy tools from personal experience: “I actually remember when I started one of my first companies, I was so adamant about not paying $85 a month for Salesforce that I use this thing called SugarCRM, which is free.”
```mermaid flowchart TD accTitle: Lightfield Network Effects Strategy accDescr: Diagram illustrating how distributing Lightfield freely across non-sales departments protects the platform against rip-and-replace attempts by incoming sales leadership. subgraph FreeOrgSeats ["Company-Wide Distribution (Free Non-Sales Seats)"] direction TB E["Engineering: Customer Context"] CS["Customer Success: Account Scoring"] F["Finance: Revenue Recognition"] end subgraph SalesLeadership ["Incoming Legacy Pressure"] VP["New VP of Sales (Salesforce-Conditioned)"] end subgraph Outcome ["Retention & Defense"] Net["Cross-Functional Network Effects"] Defense["Rip-and-Replace Defeated"] end FreeOrgSeats --> Net VP -.->|"Demands Legacy CRM"| Net Net --> Defense ```
- Alex Rampell was forced to abandon SugarCRM for Salesforce purely because his incoming sales leader refused to use anything else. - Lightfield’s sales plans grant free seats across non-sales departments to build broad organizational buy-in. - Cross-functional adoption makes it significantly harder for an incoming sales executive to mandate a rip-and-replace of the system.
---
`25:10` Skeuomorphism, Pragmatism, and the Tesla Gear Stalk
When rethinking enterprise software, founders face a fundamental interface design question: should the platform retain skeuomorphic elements—familiar blue dashboards, dense tables, and spreadsheets—or replace them entirely with natural-language prompts and proactive notifications? Keith explained that Lightfield purposefully rejected dogmatism in favor of pragmatism. Even as CEO and the company’s de facto sales manager, Keith still runs internal pipeline meetings directly from a table view. Lightfield provides polished spreadsheet tables and traditional dashboards alongside frontier interaction methods like natural language and command-line interfaces (CLI), allowing users to work in whatever format suits them best.
Alex Rampell connected this design tension to automotive hardware, comparing older BMW cockpits—packed with what felt like 9,000 physical switches and dials reminiscent of an airplane—to modern Teslas that have removed even the physical gear-shift stalk. While stripping away physical controls seems intimidating at first, the resulting interface is so intuitive that even a child could understand it, reflecting the popular Reddit concept "ELI5" (Explain Like I'm 5). In software, moving from strict schemas and endless configuration knobs toward plain-language intelligence mirrors that transition: advanced capability no longer requires an overwhelming wall of buttons.
Keith Peiris articulated their product philosophy: “I think we had this view of if we really want to be like a real enterprise CRM, we can't be religious about the way people work.”
- Keith Peiris serves as Lightfield’s de facto sales manager and continues to manage weekly internal reviews from spreadsheet views. - Alex Rampell compared traditional software interfaces with hundreds of dials to older BMW cockpits and airplane instrument panels. - Modern Teslas eliminated physical gear selector stalks entirely, exemplifying how radical simplification can make complex systems intuitive without sacrificing functionality.
---
`27:30` World-Model-Driven Sequences and the "Software is Dead" Build-vs-Buy Debate
While conversational AI handles complex synthesis, Keith conceded that the software industry has not eliminated deterministic dashboards. Business leaders still want a dependable, identical chart to look at every morning over coffee to confirm that revenue is growing. However, Lightfield has fundamentally re-architected classic sales workflows, such as email sequences. Traditional sales tools force operators to build rigid flowcharts filled with arrows, conditional branch statements, and brittle template variables. In Lightfield, outbound recipes query the company’s underlying business world model and dynamically adapt to context. Although veteran sales leaders initially balk at losing their manual toggles, they quickly realize that autonomous, context-aware recipes save immense time and eliminate steep learning curves.
This shift touches on the broader industry debate around whether "software is dead" and whether companies can simply build their own internal tooling using modern foundation models. Keith noted that as an enterprise system of record, Lightfield must respect that customer data belongs entirely to the customer. However, assembling a durable system of record involves complex foundational plumbing that internal teams underestimate. Lightfield invested heavily in engineering bespoke, high-performance synchronization engines for email, Slack, and cloud data warehouses to construct an accurate, real-time representation of customer business logic.
Keith Peiris outlined their posture regarding customer ownership and underlying data architecture: “First of all, I think just being a system of record, we have to be open-minded.”
- Traditional sales sequences rely on hard-coded conditions, branching logic arrows, and static variables, which Lightfield replaces with world-model recipes. - Lightfield engineers built proprietary sync engines for Slack, email providers, and cloud data warehouses rather than relying on shallow wrappers. - Despite the narrative that internal teams can build custom CRMs with LLMs, the underlying data integration and business modeling represent immense engineering hurdles.
`30:00` Harnesses, the SaaS Apocalypse, and Finding the Right Pricing Model
Lightfield gives customers complete control over their underlying data, allowing them to integrate via MCP (Model Context Protocol) or the command-line interface (CLI). Keith notes that when a couple of customers tried using Lightfield merely as a core system of record while building their own internal harness on top, they soon realized the depth of Lightfield's purpose-built layer: “Wait a minute, your harness was actually doing quite a bit. It had better entity recognition. It had better precision and recall. It was faster.” Keith welcomes these experiments, emphasizing: “I think our take is, it's your day to do what you want, and we're going to work hard to earn the right to your productivity every day.”
```mermaid flowchart TD accTitle: Lightfield Harness Architecture accDescr: Diagram showing how customer interfaces connect through Lightfield's harness to the core system of record.
subgraph InterfaceLayer["Access Channels"] CLI["CLI Access"] MCP["Model Context Protocol (MCP)"] Custom["In-house Custom Harness"] end
subgraph LightfieldCore["Lightfield Platform"] Harness["Lightfield Harness Layer\n- High precision & recall\n- Advanced entity recognition"] SOR["System of Record / Business World Model"] end
CLI --> Harness MCP --> Harness Custom -.->|"frequently re-converges on"| Harness Harness --> SOR ```
The conversation shifts to the broader state of SaaS pricing in the AI era. Alex Rampell observes that legacy seat-based models no longer make sense for products like Zendesk, where AI could plausibly drive the human seat count down to zero. Alex pushes back against the doom-laden narrative of the "SaaS apocalypse"—the notion that software is entirely dead, which he compares to absurd extremes like claiming supermarkets are dead because everyone will grow their own food, or car manufacturers are dead because everyone will weld their own aluminum. He asks Keith how Lightfield found its pricing strategy, particularly between the polar choices of seat-based, consumption-based, and outcome-based models.
Keith explains that to discover the "efficient frontier," Lightfield tested both extremes: - **Pure Seat-Based Pricing:** Although it felt familiar to buyers raised on Salesforce and HubSpot, the top end of the user base was consuming 10,000 times more computing resources than the long tail, making it economically unsustainable long-term. - **Pure Consumption Pricing:** Lightfield then flipped to an all-credit consumption model where every single action burned credits. Keith confesses that this triggered "the worst three weeks of the company's life," because signups poured in but users became paralyzed, terrified of taking routine actions that would deplete credits.
`33:15` The Four Work Buckets and the Limits of Outcome-Based Pricing
To resolve the pricing dilemma, Keith and Joe broken down Lightfield’s operational value into four discrete buckets to align buyer expectations with fair charging mechanics:
1. **Everyday CRM Work:** Core baseline tasks like capturing meeting transcripts, filling out database records, and updating tasks. Customers do not want volatile bills for baseline upkeep; as Keith puts it, “They're like, I don't want to think about the error bounds of my core CRM when I look at my budget over the next year.” Lightfield therefore packages this under a predictable, fixed platform fee. 2. **Pipeline Generation:** Inbound and outbound enrichment and prospect discovery. Because users see clear alpha, more meetings booked, and higher conversion to revenue, they gladly accept consumption-based pricing for this tier. 3. **Workflow Automations:** Triggered processes that perform tangible labor—such as researching a demo inbound signup and routing deep-tech prospects to Henry and health-tech accounts to Matt. The distinct return on investment makes usage-based fees feel justified. 4. **Frontier Intelligence and Scenario Forecasting:** The most forward-looking capability of Lightfield, acting as a dynamic "crystal ball or snow globe" of the business. Keith explains: “I changed my sales process after letting GPT-6 rip on the light field for a couple of hours on the weekend.” Users readily pay consumption rates for this high-leverage strategic insight.
Addressing why Lightfield does not charge on a purely outcome-based model, Keith points out that sales outcomes are inextricably tied to the customer's underlying product-market fit. Providing outbound automation for a viral product like OpenAI would yield runaway positive outcomes, whereas prospecting for an unpopular product would struggle through no fault of the CRM. Consequently, Keith highlights their current operating philosophy: “I think where we've landed at the moment is we have to charge for the work. We can't quite charge for the outcome in the space for now.”
`35:46` Ditching Swim Lanes for Extreme Velocity and Continuous Planning
Joe notes that Lightfield’s shipping velocity has skyrocketed compared to Keith’s previous venture, Tome, and asks how internal culture, tooling, and decision-making evolved to support that speed. Keith candidly reflects that Tome suffered from organizational dysfunction, citing his co-founder Henry's phrase: “We had a lot of people playing house, you know, just to say that you have the product leader and they all have their swim lanes and they get really mad if someone gives them feedback about something in their swim lane.” These rigid silos slowed execution to a crawl and made it impossible to pivot because the operational appendages were disconnected from the brain.
```mermaid flowchart TD accTitle: Lightfield Operating Rhythm accDescr: Daily and continuous planning flow showing how 40 generalists pick up stacked priorities across engineering, design, and customer success.
Standup["Daily All-Hands Standup\n(40-person team)"] StackRank["Stack-Ranked Priority List\n(Delivery, Eng, CS)"] Generalists["Unassigned Generalists\n(Engineers, Designers, CSMs)"] Execution["Direct Execution via Modern Tooling\n- Context via Lightfield\n- Design via LLM-in-Figma\n- Tasks auto-created in Linear"] Review["High Bar to Ship\n(Weekly editorial pruning & company bug bashes)"]
Standup --> StackRank StackRank --> Generalists Generalists --> Execution Execution --> Review ```
At Lightfield, Keith eliminated rigid swim lanes and extended multi-quarter roadmaps. He remarks on the shifting tech landscape: “You know the expression, man plans and God laughs. Now it's man plans and OpenAI laughs.” Today, the company operates with 40 people under a highly fluid structure: - **Universal Ownership:** “Everyone owns product and everyone owns customer success, which is kind of an interesting setup,” Keith explains. - **Single Daily Standup:** All 40 employees attend the exact same morning standup to review a single, stack-ranked backlog of engineering, delivery, and customer success tasks. Whoever is unallocated simply claims the highest-priority ticket. - **AI-Powered Fast Onboarding:** Modern tooling enables generalists to jump across disciplines effortlessly. Any teammate can understand an account through Lightfield, interact with design systems using LLMs inside Figma, and auto-generate task tickets directly in Linear. - **Low Bar to Start, High Bar to Ship:** While anyone can initiate a feature, Keith maintains a strict editorial gate to keep the product aligned with the broader company mission. Lightfield relies on weekly roadmap pruning and full-company "bug bashes"—a quality-control habit Keith retained from his tenure at Instagram: “The company needs to like this before it goes to customers.” - When asked what keeps him up at night despite this momentum, Keith states plainly: “I would say speed.”
`40:00` Building for Fast-Growing Accounts and Maximalist Prioritization
Before launching Lightfield, Keith Peiris spent time studying Tegus expert interview transcripts to analyze why startups switch away from modern, insurgent software back to legacy giants like Salesforce. One particular case stood out: an a16z portfolio company, ElevenLabs, had adopted a trendy startup CRM, but the provider couldn't ship analytics dashboards fast enough. After four months of unfulfilled requests, ElevenLabs simply churned and migrated to Salesforce. Keith admits this reality fuels his deepest paranoia: Lightfield must capture new companies as their core wedge, but they also have to build fast enough to prevent those customers from ever needing the legacy world.
Alex Rampell questions how one makes product trade-offs in an era where AI enables generalists to prompt their way into software features almost instantly. In the past, a team composed of a product manager, a designer, and nine engineers would map roadmaps years out. Keith explains that in an ultra-competitive "red ocean" market like CRM, Lightfield cannot rely solely on the initial land size; they must evaluate accounts based on their multi-year expansion potential. Because winning means scaling with breakout accounts, Lightfield skews its development toward the needs of its fastest-growing customers rather than the median user.
```mermaid flowchart TD accTitle: Lightfield Customer Prioritization Model accDescr: Flow showing how Lightfield weighs fast-growing accounts to drive long-term expansion over median customer requests.
subgraph Evaluation ["Account Evaluation"] A["New Inbound Account"] --> B{"Assess 3-5 Year Expansion Value"} end
subgraph Roadmapping ["Product Prioritization"] B -->|"High Growth / Velocity"| C["Build Fast & Unblock (Prevent Churn to Salesforce)"] B -->|"Average / Slow Growth"| D["Deprioritize Niche Custom Requests"] end
subgraph Strategy ["Product Posture"] C --> E["Maximalist Engineering (Ship Everything Needed)"] end ```
- Keith Peiris emphasizes the imperative to over-index on high-velocity clients: “I think because of that, we've leaned more towards building for our fastest growing customers than our average customer.” - Regarding the sheer scope of engineering required to prevent customer churn to legacy systems, Keith Peiris notes: “It's sort of a time of being a little more maximalist than usual.”
---
`43:10` Silicon Valley Hype vs. Rest-of-World Referenceability
Alex Rampell points out a distinct contrast: inside Silicon Valley, AI is almost overhyped, whereas outside the Bay Area, it remains heavily underhyped. Many enterprise buyers tested ChatGPT once in late 2022, saw a hallucination, wrote off generative models, and now view them with skepticism. Alex asks how Lightfield plans to cross the chasm from local startups to mainstream, non-tech businesses throughout the rest of the country.
Keith responds that system-of-record businesses inevitably generate the vast bulk of their ultimate revenue far outside Silicon Valley. However, he treats local tech clients not merely as early revenue, but as high-profile reference logos. When a well-capitalized tech firm that raised $200 million scales its go-to-market team on Lightfield, that operational track record becomes an asset when selling to traditional industries, from manufacturing to legacy logistics.
- Alex Rampell characterizes the broader market's sluggish reaction: “I kind of like to joke that AI is almost overhyped in Silicon Valley. I don't think it is, but it's like massively underhyped outside of this little region of the world.” - Keith Peiris frames the Bay Area focus as an acquisition mechanism for proof points: “I just see the focus on Silicon Valley not as efficient revenue capture, but actually just like efficient marketing capture to go out there and be like, alright, we're excited to get into manufacturing.”
---
`45:00` The Psychology of Trust and Moving Off the CRM
Alex and Keith unpack the psychology of enterprise software purchasing. Even though companies across disparate sectors require the exact same underlying software primitives—tracking customer contacts, pipeline stages, and communication histories—buyers fundamentally want to buy what their peers use. A toothbrush manufacturer hesitates to trust a platform adopted by a beverage giant simply because they operate in different sectors. In the enterprise, the old adage "nobody gets fired for buying IBM" remains potent, especially for a central system of record.
Keith highlights that choosing a CRM creates as much operational friction as switching corporate banks. To unlock risk-averse sectors like health tech and digital health, Lightfield invested heavily upfront in enterprise-grade security controls, including executing Business Associate Agreements (BAAs) and enduring years of deep penetration testing. That diligence, coupled with handling complex deals involving fifty distinct stakeholders, makes it trivial to close new health tech accounts today because prospects see peers already deployed safely on the platform.
- Keith Peiris explains the extreme friction of replacing a database: “In many ways, your CRM is maybe harder to move off of than your bank.” - Discussing why reference architectures carry so much weight with buyers, Keith Peiris reflects on his own purchasing instincts: “If I imagine even myself, like, I don't want to choose the wrong ERP. So I probably want to choose the ERP that companies who look like me chose. So that I never have to think about this after I buy it.”
---
`48:15` The Fallacy of DIY "Company Brains" and the Future Scenario Engine
Alex asks whether Keith encounters companies attempting to vibe code their own internal CRMs from scratch rather than buying an off-the-shelf platform. Keith notes that while seed-stage founders occasionally threatened to hack something together over four weekends, they inevitably returned a few weeks later when their homebrew tools hallucinated customer emails or broke down. At the enterprise scale, companies do not claim they are building a standard CRM; instead, they claim they are engineering their own internal "company brain." After stumbling on customer modeling, context engineering, and real-time world models, those enterprises usually concede defeat and buy Lightfield.
Looking ahead, Keith shares that his primary ambition is transforming Lightfield into an intelligent scenario-planning engine for executive teams. Rather than a static system of record, it functions as a mirror that guides existential decisions, such as identifying hiring velocity, pinpointing missing products, or recognizing that an enterprise business needs a dedicated mid-market line. Joe Schmidt underscores that having an AI system infused with comprehensive business context transforms frontier models into actionable decision engines.
- Keith Peiris shares what drives his product vision: “I'm stoked about Lightfield becoming your sort of crystal ball for scenario planning. And the thing that gets me up in the morning is actually having folks use Lightfield to think about how many reps should I hire? What product should I build next? Where do I go?” - Keith Peiris captures the core mission of an AI-native CRM: “I'm just excited about us being that sort of mirror for a company to make their hardest, most consequential decision.” - Joe Schmidt observes how holistic context unlocks automated reasoning: “If you just have all of this data and all this context, you can actually be data-driven, and this is the exact opportunity to throw frontier intelligence at a really complicated decision.”
`50:00` Ignoring the Noise During a Startup Pivot
Reflecting on how Lightfield operates, the hosts note that setting up business operations and strategic planning historically required hiring exceptionally smart people, maintaining dedicated operations staff, writing endless SQL queries, and stringing together complicated tools. With Lightfield, an operator can simply talk to the system across an afternoon or a weekend and construct an intelligent, actionable plan for moving the company forward. Looking back at Keith Peiris's journey from scaling Tome to pivoting and launching Lightfield, Joe Schmidt asks what single piece of advice Keith would give his past self if he had to embark on that pivot journey all over again with the benefit of hindsight, especially for founders currently facing their own transitional moments.
Keith emphasizes that during a pivot, nearly all the ambient chatter and internal friction must be actively disregarded. He recalls the petty distractions that surfaced during their transition—employees reminiscing about how the physical office reminded them of "the good old days," complaining that the catered food was uninspiring, or anxiously asking how their stock options would be repriced. For Keith, surviving and succeeding through a radical pivot requires completely shutting out those peripheral anxieties and staying locked onto genuine customer utility.
As Keith put it: “I think the most important thing to remember is that, almost none of the noise around you matters when you're in a pivot.” He underscored that founders must ruthlessly strip away distractions to focus purely on value: “You just need to find pain. You need to be inspired to build a product or service that solves that pain. And you need to be like maniacally focused on your customers.”
- **The old ops stack vs. AI:** Tasks that previously demanded specialized operations headcount, deep data analysis, and SQL queries can now be resolved through natural conversations with Lightfield over a few hours. - **Common pivot distractions:** Internal complaints regarding office nostalgia, team perks, catering quality, and equity repricing are non-essential noise that founders must block out with blinders.
`51:32` Wrapping Up and Podcast Disclosures
Closing the conversation, the hosts express immense admiration for what Keith and his team have accomplished in shifting directions and building Lightfield into such a compelling product. They encourage listeners who haven't yet experienced the tool to try it out directly at lightfield.app, emphasizing how proud and excited they are to partner with Keith as the company shapes the future of AI-native enterprise workflows.
Keith thanks the hosts for having him on, after which the show wraps with standard a16z podcast outro credits and mandatory investment disclosure notes.
- **Product URL:** Lightfield can be accessed and evaluated at lightfield.app. - **Show channels:** Listeners are directed to subscribe and follow across YouTube, Apple Podcasts, Spotify, X (@a16z), and Substack (a16z.substack.com). - **Compliance note:** Andreessen Horowitz (a16z) notes that it and its affiliates maintain investments in companies discussed on the podcast, with full public disclosures accessible at a16z.com/disclosures.
## Notable quotes > `17:40` The best companies have hostages, not customers. Like SAP has hostages. — Alex Rampell
> `15:00` Basically, intelligence is greater than schema. That's the crux. — Keith Peiris
> `45:00` In many ways, your CRM is maybe harder to move off of than your bank. — Keith Peiris
> `00:00` As a CRM company in a red ocean space, we have to be an expansion company. — Keith Peiris
> `50:00` I think the most important thing to remember is that, almost none of the noise around you matters when you're in a pivot. — Keith Peiris
> `00:00` If we can help you completely model your business and your customer reality, then the rest will be easy. — Keith Peiris
## Mentioned Lightfield, Keith Peiris, Alex Rampell, Joe Schmidt, Tome, Facebook, Salesforce, Power, a16z, Instagram, Messenger, GPT-3, ChatGPT, GPT-4, GPT-6, OpenAI, Gong, Snowflake, SAP, SugarCRM, HubSpot, Zendesk, Linear, Figma, ElevenLabs, Tegus, FDA
## Video description a16z’s Alex Rampell and Joe Schmidt sit down with Lightfield co-founder and CEO Keith Peiris to discuss what it takes to rethink the CRM for an AI-native world, and the unusual pivot that got him there. Keith previously built Tome to 25 million users, but eventually walked away from the product after concluding that the underlying technology couldn’t capture enough context about a presenter, their audience, and the relationship between them. Starting again, his team followed customers from AI presentations into sales workflows and eventually found a harder problem: making sense of the fragmented and often conflicting data spread across a company’s emails, calls, CRM, and other systems. They unpack Lightfield’s idea of a “business world model,” why Keith believes intelligence can replace much of the rigid schema behind traditional software, and what changes when AI has enough context to reason about a company and its customers. They also get into building for greenfield versus brownfield markets, AI-era pricing, running a company where everyone is a generalist, and Keith’s lessons from making a hard pivot.
[Source](https://mgln.ai/e/1344/afp-848985-injected.calisto.simplecastaudio.com/3f86df7b-51c6-4101-88a2-550dba782de8/episodes/6229af03-9c3d-49f0-b84c-fbbd1b7a789b/audio/128/default.mp3?aid=rss_feed&awCollectionId=3f86df7b-51c6-4101-88a2-550dba782de8&awEpisodeId=6229af03-9c3d-49f0-b84c-fbbd1b7a789b&feed=JGE3yC0V)