Platform building in the AI era and the FDE gold rush

Summary

A market report on enterprise AI infrastructure winners and the strategic mistakes incumbent platforms are making. The author argues that most vendors have misunderstood platform building by launching extensibility and headlessness before solving a killer application; the real platform winners will own centralised data and operate like hyperscalers rather than traditional SaaS. Covers moves by Gainsight (moving toward AI-augmented services delivery), Snowflake (now winning on coding agents), Anthropic's FDE JV-acquired startup Fractional, OpenAI's talent raids on Salesforce, Meta's new FDE unit, and explosive ARR growth at inference vendors. Read the original thread.

Key Claims

• Most incumbent platforms commit a cardinal sin by building extensibility and headlessness (APIs, MCPs, CLIs) and declaring them platforms when they lack the killer application that earns platform status.
• Platforms that dominate the next era will resemble hyperscalers structurally, not legacy SaaS, because they'll need to operate at different margins and serve as infrastructure for agents, not just data sources.
• Gainsight is pivoting toward AI-augmented outcomes-based services delivery — Gainsight-employed FTEs working inside client environments — signalling that software companies are moving toward operating models that look like consulting.
• Snowflake's Q1 market surge was driven by growth in Cortex Code, its Snowflake-native coding agent, and AWS marketplace commitments; coding agents and committed spend now drive market perception more than traditional traction metrics.
• Anthropic's FDE JV (with Blackstone and Hellman & Friedman) acquired Fractional, an SF-based AI services firm founded by former LiveRamp executives, and is positioning it as a lab for AI transformation playbooks deployable across PE-owned portfolio companies.
• OpenAI is aggressively recruiting enterprise talent, including Salesforce's global CMO and confirmed or rumoured senior Salesforce executives, signalling serious enterprise market ambitions.
• Meta launched a Forward Deployed Engineer unit embedding PMs, data engineers, and software engineers inside large customers to integrate Meta's tools directly into client workflows, leveraging its existing relationship depth across enterprises.
• Kirkland & Ellis, the world's highest-grossing law firm, chose to build its own AI legal system in-house rather than buy from purpose-built legal AI vendors, likely prioritising IP and data ownership.
• Inference vendors are experiencing explosive growth in Q1 2026 — Baseten tripled to $600M ARR, Fireflies quadrupled to $800M+ ARR — outpacing traditional SaaS metrics.

Quotes

• "A platform provides more benefits to the ecosystem above than it captures for itself. A platform is the place that apps, or agents in this case, run."
• "It's remarkable to watch the last gen of winners commit the cardinal sin of platform building: starting with the platform before you've built the killer app."
• "The platforms that win the next era will look very different from the last. They'll look far more like hyperscalers than SaaS platforms of yesteryear, and that's a whole new company DNA."
• "As everything becomes agent infrastructure, expect all of these trends and numbers to continue to increase dramatically."