Investors Bet on AI to Transform Low Margin Services into High Margin Software

Investors in the technology sector are increasingly betting on the idea that generative AI can transform low-margin service businesses into high-margin software companies. This strategy involves acquiring traditional business process outsourcing (BPO) companies, such as call centers and accounting firms, at modest valuations. These companies typically operate with 10-15% EBITDA margins due to their reliance on human labor for repetitive tasks. The plan is to deploy generative AI to automate these workflows, reduce headcount, and expand EBITDA margins to 40% or more. The ultimate goal is to sell these newly AI-enabled services companies at software multiples, as buyers and public markets recognize the transformation from a human-heavy service business to a scalable AI business.

However, this thesis is flawed. It rests on a fundamental category error: confusing operational improvement with business model transformation. While AI can make workflows more efficient, it does not turn a services company into a software company. This misconception is evident in the valuations of AI-transformed BPO firms, which trade at 5-23x EV/EBITDA, compared to their pure software counterparts, which command valuations of 22-92x EV/EBITDA. This gap cannot be bridged with press releases about AI partnerships; it reflects a fundamental difference in how markets value human-dependent businesses versus true software platforms.

For instance, Concentrix, often cited as a BPO transformation success story, has invested heavily in AI but still trades at a low EV/EBITDA multiple with an EBITDA margin around 10%. This indicates that automating workflows does not change the fundamental business model. PolyAI, a leading conversational AI company, explored acquiring human-driven contact centers in 2019 but ultimately decided against it. The company identified several structural barriers, including the illusion of control, the pricing trap, and zero switching costs, which remain unchanged today. PolyAI chose to remain a software company, partnering with BPOs rather than acquiring them, and is now valued at over $500 million.

Services businesses are inefficient by design, as clients pay for flexibility, customization, and someone to blame when things go wrong. Automating away the human element changes what is being sold, and clients who wanted software would have already bought it. The most successful services firms use AI to augment their humans, not replace them, and maintain margins through pricing power and relationships. Ultimately, they still trade at services multiples because that is what they are.

The AI rollup thesis is a familiar pattern in technology investing: the conflation of technological capability with business model transformation. In the early 2000s, e-commerce was thought to transform retail margins, but Amazon succeeded by building a native digital retailer. In the 2010s, investors believed software would eat traditional industries, but the winners built new software-native businesses. The same lesson applies today: AI may transform some corners of professional services, but this is different from the AI rollup thesis that assumes low-margin, labor-heavy service businesses can be turned into software-like platforms simply by embedding AI. Transformation will come from new, AI-native companies with fundamentally different economics.

The AI rollup thesis is venture capital’s attempt to arbitrage the multiple gap between services and software. However, services businesses face different constraints, different economics, and different customer relationships than software companies. PolyAI saw this in 2019, and public markets see it now. The AI revolution is real, and the opportunity to improve services businesses with AI is real, but the idea that this improvement transforms them into software companies is unlikely to be real today, just as it wasn’t in 2019. AI rollups may still deliver returns, but not the kind VCs are underwriting. At best, they are tech-enabled private equity: operationally heavy, valuation-capped, and unlikely to scale like software.