Nathan Benaich is the founder of Air Street Capital and author of the AI report of the state of AI. Nikola Mrkšić is the CEO of Polyai.
In the entire world of technology investments, investors scale their bets on a seductive thesis: Generative AI will convert companies with low margins in companies with high margin software companies. Several well -known platform companies have committed billions of things this strategy and started making their bets. This is how the thesis goes:
First acquire traditional business process -Outsourcing companies (BPO) such as Call Centers and accounting companies on modest ratings of 1x income. As a rule, these companies work with 10-15% EBITDA (profits before interest, taxes, depreciation and amortization margins), which have been weighed up by armies by human workers who perform repeating tasks, and automation faces the greatest structural resistance.
Second, you provide generative AI to automate core workflows, reduce removal and to expand the EBITDA margins to 40% or more. What once hundreds of bookholders or call center agents needed can now be carried out by a handful of people who manage AI systems.
Third, end the newly shaped company for AI-capable services for software multiples, since buyers and public markets recognize that you have transformed a human service business into a scalable AI company. When traditional BPOs act at 6x EBITDA, software companies 20x or more commands.
It is a brilliant arbitrage on paper. In practice, it is a fraud picture. It is based on a fundamental category error: confusion of operational improvement with the transformation of the business model. Yes, AI can make workflows more efficient. No, this does not transform a service company into a software company.
In fact, five years ago, a nunner carried out this exact experiment and went away. His results should serve as a warning for today’s believers. Let’s dig in.
The 69x assessment canyon
The most muddy evidence against the Ki -Rollup thesis combines visibility in public markets in the areas of visibility. Today’s “AI-transformed” BPO companies that have invested heavily in automation concentrationPresent Pactand infosys-over 5-23x EV/EBITDA (company value for EBITDA). Your pure software counterparts such as SalesforcePresent Serviceand working day, command reviews of 22-92x EV/EBITDA. Here is a table to tell the story:
This is not a gap that can be bridged with press releases via Openai, Anthropic or Gemini partnerships. It is a fundamental difference in how markets depend on human -dependent companies compared to real software platforms.
Consider Concorlix, which is often referred to as BPO transformation success history. Despite a great advance in the introduction of their gene AI products in 2024 and now have Deployment In the case of over 1,000 customers, the company’s EV/EBITDA multiplier remains stuck in the low individual digits, and the EBITDA margin is still 10%. The message of the market is clear: the automation of workflows does not change your basic business model.
The Polyai prophecy
In 2019, Polyai, the leading Ki company of the conversation, examined for six months whether reigning contact centers should be powered in order to accelerate its growth. After analyzing the opportunity to visit 10 contact centers, to build relationships with three large BPOs and to stop industry consultants, the answer was a clear no.
“Business process -Outsourcing companies do not trust to innovate innovations, not reward innovations and not to be approved for innovations,” was the board deck.
The structural barriers she identified remain unchanged today:
- The illusion of control: Buying a BPO does not mean owning what you support business. You simply rent the right to deliver work on the conditions of the customer. Tech stacks, processes and permits remain firmly in the customer’s hands. KI deployments still require their permission, integration and supervision. You have no control; They are an interchangeable provider.
- The price trap: Most service companies expect hours. Efficiency improvements that reduce the billable hours can directly exploit the income. As Polyai stated, BPOS innovation promise to win contracts, and then return to maximize the billable hours to protect the margins. It is a business model that basically contradicts automation.
- Zero turnout costs: When 10-year service contracts were once the norm, it is now becoming more and more common to see three years or less. This reduces the ability to regain AI investments in advance, especially if little client-lock-in, there are no network effects and no moat.
Polyai decided to stay a software company that worked with BPOS instead of acquiring them. Today it is estimated With over 500 million US dollars for customers such as PG & E, Marriott and Fedex. In the meantime, the BPOs, which they consider for the purchase, still have to act with single -digit multipliers.
Why this time is not different
Here is what investors are missing: Service companies are not accidental inefficient. They are inefficient through design. Inefficiency is the product. Customers pay flexibility, adaptation and someone who blame when something goes wrong.
The automation of humans not only reduces the costs, but also in fundamentally changing what they sell. BPO technology was never the restriction. And customers who wanted software have already bought software.
The most successful service companies understand this. They use AI to expand their people and not to replace them. They keep the edges through pricing and relationships, not by reducing the main focus. Ultimately, they still act with services multiples because they are.
The lessons of history
The KI -Rollup thesis represents a well -known pattern in the investment in technology: the fusion of technological skills with the transformation of the business model. We have already seen this film.
In the early 2000s, the faithful thought that e-commerce would change retail margins. Amazon They correctly demonstrated by building a local digital retailer, not by acquiring and transforming Sears or Barnes & Noble. In the 2010s, investors believed that software would eat traditional industries. The winners have built up new software native companies instead of exterminating old ones.
The same lesson applies today, but with a closer area. AI can certainly change some corners of professional services, especially if existing companies are urged to pursue new tools from private equity owners with clear control and incentives. We have seen this in sectors such as health and financial services in which PE companies have promoted the introduction of AI-controlled instruments. However, this differs from the AI Rollup thesis that VCS pursue-one that takes on a lower, labor-subject service company, can be converted into software-like platforms by embetting AI. For these companies, the transformation will not arise from the possession of the service layer. It will come from new AI-native companies with a fundamentally different economy.
Conclusion: Do you have the software, not the service
The Ki -Rollup thesis is the attempt by Venture Capital to arbitrate the diverse gap between services and software. But this gap exists for a certain reason. Service companies, even highly automated, are exposed to different restrictions, different economy and different customer relationships as software companies.
Polyai saw it in 2019. Public markets see it now. The AI revolution is real. The possibility of improving service transactions with AI is real. The idea that this improvement is transforming you into software companies? It is unlikely that it is real today, just as it was not in 2019.
Ki rollups may continue to deliver returns, but not the type of VCS are insured. At best, they are technically capable private equity: operational heavy, evaluation capital and unlikely how software scales.
The opinions that were expressed in Fortune.com comments are exclusively the views of their authors and do not necessarily reflect the opinions and beliefs of against Assets.
Read more:
