The AI services transformation may be harder than VCs think

Venture capitalists are increasingly optimistic about using artificial intelligence (AI) to enhance profitability in traditionally labor-intensive service sectors. The strategy involves acquiring established professional services firms, integrating AI to streamline operations, and then leveraging the resulting cash flow to acquire more businesses. General Catalyst, a prominent player in this space, has committed $1.5 billion from its latest fund to a "creation" strategy aimed at nurturing AI-driven software companies across various sectors, ultimately using these firms to acquire legacy businesses in similar fields. General Catalyst has diversified its investments across seven industries, including legal services and IT management, with plans to expand to up to 20 sectors. As Marc Bhargava, who is spearheading these initiatives, pointed out in a recent TechCrunch interview, the global services industry is worth a staggering $16 trillion, while software stands at only $1 trillion. The appeal of investing in software lies in its higher margins, which can significantly increase profitability as companies scale. If AI can successfully automate a substantial portion of service tasks—potentially up to 70% in specific areas like call centers—the financial benefits could be considerable. The enhanced cash flow could then be reinvested to acquire additional firms at prices that traditional buyers may find unaffordable, creating a cycle of growth that investors find attractive. For instance, Titan MSP, a company in General Catalyst's portfolio, received $74 million to develop AI solutions for managed service providers and subsequently acquired RFA, a recognized IT service firm. Titan has shown that it can automate 38% of standard MSP tasks, and it now intends to use its improved profit margins to acquire more MSP businesses. Similarly, Eudia, another incubated company, focuses on providing legal services to in-house departments rather than conventional law firms. Eudia has secured contracts with major corporations like Chevron and Southwest Airlines, offering fixed-price legal services powered by AI, and has recently expanded through the acquisition of an alternative legal service provider. General Catalyst aims to at least double the EBITDA margins of its acquired companies, a strategy echoed by other firms as well. Mayfield, for example, has earmarked $100 million for investments in AI-driven business models. They recently supported Gruve, an IT consulting startup that successfully acquired a smaller security consultancy and saw its revenue triple within six months, achieving an impressive 80% gross margin. However, emerging studies suggest that the anticipated transformation of the services industry may be more complex than VCs expect. A survey from Stanford Social Media Lab and BetterUp Labs revealed that approximately 40% of full-time employees are dealing with "workslop"—AI-generated outputs that seem polished but often require extensive revision. This inefficiency translates to an average loss of nearly two hours per incident for employees, imposing a hidden cost of $186 per month per person. For organizations with 10,000 employees, this could result in over $9 million in lost productivity annually. The implication is clear: merely implementing AI does not guarantee improved outcomes. Bhargava argues that the challenges faced by companies in adopting AI further validate General Catalyst's approach. He contends that if integrating AI were straightforward, many major corporations would already have achieved significant transformations. The critical factor is the technical sophistication required for successful AI implementation. Bhargava emphasizes the need for applied AI engineers who understand the nuances of various technologies to effectively integrate them into business processes. This complexity is precisely why General Catalyst's model of collaborating AI specialists with industry veterans is strategic. Yet, the issue of workslop raises concerns about the economic viability of these investments. A crucial question remains: how significant is this problem, and will it evolve over time? If businesses reduce their workforce in line with AI efficiency expectations, they may lack the personnel needed to address AI-generated errors. Conversely, maintaining staffing levels to manage the increased workload from problematic AI outputs might hinder the margin gains that investors are counting on. These challenges could slow down the aggressive scaling strategies that are central to VCs’ roll-up plans, potentially jeopardizing the financial metrics that make these acquisitions appealing. However, it often takes more than a few studies to deter Silicon Valley investors, who typically target businesses with established cash flows. As Bhargava concludes, as long as AI technology continues to advance and investment in these models persists, there will be ample opportunities for industry disruption.