Venture Capital’s AI Services Bet Faces Hidden Challenges as “Workslop” Costs Mount
Venture capital firms are racing to apply artificial intelligence to traditional services businesses, convinced they can replicate software’s sky-high margins. Leading the charge is General Catalyst (GC), which has earmarked $1.5 billion from its newest fund for a “creation strategy” aimed at building AI-native software companies in specific industries, then acquiring established firms and their customer bases. From legal departments to IT management, GC is already active in seven verticals and plans to expand to as many as 20, betting that automating up to 70% of routine tasks could dramatically boost profits.
Examples of this playbook are already live. Titan MSP, a GC portfolio company, raised $74 million across two tranches to develop AI tools for managed service providers before acquiring RFA, a respected IT services firm. In pilot tests, Titan automated 38% of typical MSP tasks, freeing up cash flow for further acquisitions. In the legal sector, GC incubated Eudia, which offers Fortune 100 clients such as Chevron, Southwest Airlines, and Stripe fixed-fee AI-powered legal services instead of traditional hourly billing. Eudia’s purchase of Johnson Hanna, an alternative legal services provider, reflects GC’s plan to double EBITDA margins at acquired firms.
Other investors are adopting similar tactics. Mayfield has set aside $100 million for “AI teammates” startups, including Gruve, an IT consultancy that bought a $5 million security consulting business and scaled it to $15 million in revenue with an 80% gross margin in just six months. Solo investor Elad Gil has pursued the same approach for three years, arguing that owning and transforming a business directly allows for faster AI integration than simply selling software to third parties.
But cracks are appearing in the model. A new study by Stanford’s Social Media Lab and BetterUp Labs of 1,150 full-time employees found 40% are dealing with “workslop” — AI-generated output that looks polished but creates extra work to decipher, correct, or redo. Employees spend an average of nearly two hours handling each instance, at an estimated hidden cost of $186 per month per person. For a 10,000-employee organization, that translates into more than $9 million a year in lost productivity, according to the researchers’ Harvard Business Review article.
GC’s Marc Bhargava disputes claims of AI overhype, saying the complexity of implementing the technology validates the firm’s strategy of pairing AI specialists with industry experts. Yet the hidden costs of “workslop” raise real questions about whether the dramatic margin gains investors are counting on can actually be achieved. If companies cut staff as AI takes over but lack people to catch and fix errors, quality may slip; if they maintain staff, the expected savings evaporate. For now, though, GC’s model remains profitable because it acquires cash-flow-positive firms — a notable shift from the traditional VC approach of funding high-burn startups — and the firm insists that as AI models improve, new industries will open up for disruption.
source: techcrunch