Services companies plus AI doesn't equal software margins
VCs are backing companies that use AI to achieve software margins in service businesses, but research shows 40% of workers spend extra time fixing low-quality AI output that eats the projected gains.
The VC thesis here is that you invest in services companies—or create new ones—that use AI to wring software-like margins out of traditionally labor-intensive work. General Catalyst put $1.5 billion behind this idea. They're incubating companies like Titan MSP and Eudia from scratch, providing them with AI tools to deliver services at software economics, and then using them as acquisition vehicles to roll up established firms. Titan MSP claims to automate 38% of typical IT management tasks. The math looks clean.
The problem is the Stanford research on what they call "workslop." It turns out that 40% of employees are spending extra time—an average of two hours per instance—dealing with AI output that appears finished but isn't. For a 10,000-person organization, that's $9 million annually in hidden costs.
So the margin improvement depends entirely on whether your AI actually reduces work or redistributes it to cleanup duty. For product teams, the quality bar isn't "does it produce output" but "does it produce output that doesn't create downstream problems."
What I find telling is that even Marc Bhargava from General Catalyst admits, "it's really hard to transform a company with AI." He sees the workslop problem as validation that you need specialized AI engineering talent, not just off-the-shelf tools. Maybe he's right. But if the work requires that level of sophistication, the roll-up strategy—where these portfolio companies acquire multiple businesses quickly to achieve returns—starts to look harder than the pitch deck suggests.
Connie Loizos
