The AI services transformation may be harder than VCs think (techcrunch.com)

Venture capital firms, led by General Catalyst, are betting that AI can turn labor-heavy professional services into high-margin, software-like businesses by incubating AI-native companies and using them to acquire established service firms. General Catalyst has earmarked $1.5B for this “creation” strategy across multiple verticals and cites pilot wins — Titan MSP automated ~38% of managed‑service tasks and Eudia (legal ops) has landed Fortune 100 clients — with ambitions to automate 30–50% of target companies (up to 70% in call centers) and materially boost EBITDA. Other investors (Mayfield, Elad Gil) report similarly bold margin claims (Gruve citing an 80% gross margin post‑AI), reinforcing a roll‑up thesis: improved cash flow funds more acquisitions and scales incumbents into platform businesses. But the approach faces material technical and operational headwinds that could erode the promised economics. A Stanford/BetterUp study highlights “workslop” — AI output that looks polished but requires an average of nearly two hours to fix — imposing an estimated hidden cost of $186 per employee per month (≈$9M/yr for a 10k workforce). Beyond garbage‑in/garbage‑out, successful transformation demands rare applied‑AI engineering expertise to pick models, integrate them into reliable workflows, and guard against error cascades. The core tension: cut headcount to realize margins and risk fewer people to catch AI mistakes, or keep staff and blunt margin gains. The thesis is plausible but hinges on sustained model improvements, high‑quality implementation, and effective human‑AI orchestration.