Merge Built a $100M ARR Enterprise Motion on Forward-Deployed Engineers, Then Bolted PLG on Top

Merge went from $1M to $100M ARR selling enterprise-first, then added product-led growth on top — the reverse of how most B2B software companies claim they need to grow. In a July 22 GTMnow conversation, CEO Shensi Ding credits “forward-deployed engineers” — specialists who embed post-contract to solve the “1% edge case” blocking enterprise adoption — with unlocking 5-10x usage growth once deployed, calling the role an “enterprise cheat code.” Merge, backed by $75M from Accel and NEA, now runs both motions at once: 400-plus enterprise customers including OpenAI, Netflix, Uber, Mastercard, JPMorgan, and Amex, alongside more than 20,000 self-serve organizations.

The more concrete signal for sellers evaluating their own AI stack is buried in the same conversation: Ding describes 2026 as the year Merge moved away from 2025’s “token-maxing” behavior, routing simple tasks to cheaper models like Haiku and reserving frontier models for complex work through Merge’s own Gateway product — plus rebuilding her marketing function to run like an engineering team, automating changelogs and event coordination and taking field events from 1-2 per quarter to 3-5 per month.

That FDE motion isn’t unique to Merge. A separate GTMnow piece from Sophie Buonassisi traces the same playbook across Legora ($1M to $100M ARR in 18 months selling to law firms), Sierra ($100M ARR run rate in 21 months, working with 40% of the Fortune 50), and Decagon (tripling revenue to roughly $35M across 100+ enterprise logos, deployments live in about six weeks). The mechanism cited there is identical: MIT research found 95% of enterprise AI pilots stall, and an embedded senior engineer who ships the actual integration is what gets an agent past that wall — Anthropic’s $1.5B Deloitte deal and OpenAI’s $10B Bain alliance are, per that piece, FDE staffing pacts at the same scale, just priced in the billions instead of the low hundreds of thousands per hire.

The Ceiling Method read: for AI-native vendors selling into enterprise accounts, the FDE isn’t a support cost, it’s the close.