Lightfield

The bet

Lightfield is betting that computing over a complete, lossless customer memory beats retrieving from a schema. Founders Keith Peiris & Henri Liriani shut down Tome (20–25M users) and rebuilt for sales teams. Notably, they're running the AI-CRM bet on repurposed Tome capital — $81M raised as Tome (through a $43M Series B at a $300M valuation), with no new Lightfield-branded round disclosed. That's a Series C signal worth watching: fresh money would mean investors buy the pivot.

Snapshot

Verification note: the "2,500+ customers" figure is a vendor press release syndicated via Yahoo Finance — self-reported, not independently audited. At that scale with zero G2/Capterra/TrustRadius reviews, the right claim is that Lightfield is unreviewed, not "poorly reviewed."

What's genuinely good

• Code execution over customer memory (shipped Feb 2026) — the agent plans, writes Python, runs it in a sandbox against the full customer graph, evaluates, and iterates. A real, technically specific, falsifiable capability — not marketing puffery.
• Schema-less "customer memory" — semantic key-value storage that auto-backfills up to 2 years from email/calendar/Slack/meetings/tickets on connection.
• One-Hour Migration Agent — ingests HubSpot/CSV exports at up to 90,000 records/hour, mapping contacts/companies/deals/stages without manual cleanup.
• Founder pedigree — ex-Meta (Instagram Direct, Messenger); the product polish shows up in early reviews.
• Genuine shipping velocity — roughly weekly changelog cadence sustained across the whole window.

What's weak

• The flagship "agentic" SKU is a service, not software — Agentic Pipeline Generation's own product copy says "Lightfield does this with a forward-deployed team… we scope hypotheses with you… we meet weekly," with results in "4–6 weeks of live sending." It's a managed-service engagement wrapped in CRM data access, gated behind the $3,000+/mo Growth tier.
• Zero third-party validation — no G2/Capterra/TrustRadius reviews at 2,500+ signups; AI claims are essentially unaudited.
• Pricing volatility — from ~$40/seller (Nov 2025) to a live $89–249/seat (today), a 2.2–6.2× jump in ~8 months, with intermediate figures that don't cleanly reconcile.
• Astroturf concern — an independent blog flags "no organic user community, just sponsored posts and paid reviews," and much of the "Lightfield review" long tail is templated SEO-farm content.
• No mobile app reported — a real gap for a product claiming to run your entire GTM motion U.

AI capability lens

The most technically interesting architecture in the cohort — undercut by a services-heavy "agentic" flagship and total absence of neutral validation.

Shipped & verifiable

• Code execution over memory — deterministic, at-scale computation over structured + unstructured customer data, rather than pure LLM retrieval; the vendor's argument is that retrieval-and-summarize plateaus because it can't compute why a deal moved
• Auto-capture + 2-year backfill eliminating manual upkeep (the "65% of rep time" pain)
• Meeting intelligence — recording, transcription, prep briefs, follow-up drafting
• NL BI with citations back to source conversations; MCP + Claude Opus 4.7 support

Customer voice

Praise

"The most impressive AI-native CRM we have used" — auto-updates after meetings "actually work and save real time."MakerStack review, 7.5/10 · platform: independent blog

Skepticism

"Outside of paid influencers, we have not heard many users genuinely raving about it… It is one to watch rather than one we would bet a sales team on today."Breakcold, 2026-06-30 · platform: independent blog

No dedicated Hacker News thread and no primary Reddit thread surfaced despite the newsworthy founder-pivot story — a conspicuous absence of organic community footprint. Sentiment leans on channels Lightfield controls (own blog, LinkedIn, SaaStr features).

Pricing teardown

Priciest of the modern cohort. Live pricing (fetched 2026-07-05):

The one verifiable trajectory: ~$40/seller (Nov 2025) → $89–249/seat (today), a 2.2–6.2× increase in ~8 months. Whether that signals confidence or churn-risk is the open question.

Head-to-head (10-seat deal)

Watch

Two signals: (1) a Series C — fresh money would validate the pivot beyond repurposed Tome capital; (2) whether Agentic Pipeline Generation productizes — if the forward-deployed service becomes real self-serve software, their "AI runs your GTM" claim finally matches the product. First real G2 reviews appearing would also change the "unreviewed" read.