Why the investment strategy “AI Rollup” is defective

First, obtaining companies using external sources of commercial operations (BPO) such as communication centers and accounting companies with modest reviews of 1x revenue. These companies usually operate 10-15 % of EBITDA (pre-interest, taxes, depreciation and consumption), which are weighed by the armies of human workers who perform frequent tasks, and automation faces the greatest structural resistance.

Second, the obstetric artificial intelligence spread to automate the basic workflow, cut the number of employees, and expand profit margins before benefits, taxes, depreciation and consumption to 40 % or more. What is required by hundreds of accountants or communication center agents can now be done by a handful of people who run artificial intelligence systems.

Third, get out of the newly enabled services company in Software Multiples because buyers and public markets realize that you have transformed human heavy services to AI to be developed. Where the traditional BPOS is trading in 6X Ebitda, software companies are worth 20x or more.

On paper, it is a great arbitration. In practice, it is a mirage. It depends on a basic category error: confusion of operational improvement with the transformation of the business model. Yes, Amnesty International can make the workflow more efficient. No, this does not turn a service company into a software company.

Indeed, five years ago, Amnesty International has now conducted this accurate experience, and has moved away. Its results should be a warning to believers today. Let’s dig.

Evaluation Valley 69X

The most suitable evidence against the Rollup Ai is hiding in the sight of public markets. BPO companies “converted” today, which have invested extensively in automation-among them the focusand GenpctAnd Infosys-Trade at 5-23x EV/ebitda (the value of the institution to Ebitda). Successive software, such as Salesforceand servicenowWorkday, 22-92x EV/epitda leadership. Here is a scheme for the story’s novel:

This is not a gap that can be blocked with press data on the partnerships of Openai, Anthropic or Gemini. It is a key difference in how the markets estimate the human -based companies in exchange for real software platforms.

Consider the focus, often cited as a BPO transformation story. Despite the main batch of the launch of the Gen-E products in 2024 and now Publishing operations In more than 1000 customers, the company is still multiple EV/ebitda the company stuck in the low individual numbers, and the EBITDA margin is still hovering about 10 %. The market message is clear: automating the workflow does not change your basic work model.

Polaya prophecy

In 2019, Polyai, the leading company in the artificial intelligence conversation, spent six months to explore whether to get the current communication centers that a person moves to accelerate its growth. After analyzing the opportunity by visiting more than 10 communication centers, building relationships with three main BPOS, and employing industry advisers, the answer was clear.

“It is not trusted by companies using external sources of commercial operations, not their reward for innovation, and not allowing them to innovate,” she reads the roof of the Board of Directors.

The structural barriers that you set are still unchanged today:

• And they are controllingBuy BPO does not mean having the business you support. You simply rent the right to provide employment on customer terms. Chimneys, operations and technical approvals remain firmly in the hands of the customer. Artificial intelligence deployment continues to require its permission, integration and control. You are not in control. You are an replaceable seller.
• Pricing: Most of the service work is an hourly bill. Efficiency improvements that reduce filling watches directly from revenue action. He also discovered Polyai, BPOS is innovated in winning contracts, then returns to maximizing the hours of the margins. It is a business model at odds with automation.
• Zero replacement costsAs service contracts were for 10 years, it was now common to see three years or less. This reduces the ability to recover artificial intelligence investments, especially when there is a little closure of the customer, no effects on the network, and no trench.

Polyai chose to remain a software, partnership with BPOS instead of obtaining it. Today, it is value More than $ 500 million with clients like PG & E and Marriott and FEDEX. Meanwhile, the BPOS that they are thinking about buying is still a trade in one complications.

Why this time is not different

Here is what investors miss: companies are not effective by chance. It is ineffective by design. Implementation is the product. Pay customers in exchange for flexibility, customization, and a person who is blamed when things get worse.

Automation of man not only reduces costs, but mainly changes what you sell. BPO technology is not registration. And customers who wanted software have already bought programs.

The most successful service companies understand this. They use artificial intelligence to increase humans, not to replace them. It maintains margins through the strength of pricing and relationships, and not by reducing the number of employees. In the end, they are still trading in the complications of services because this is what they are.

History lessons

The AI ​​Rollup thesis represents a familiar pattern in technological investment: mixing technological capacity with the transformation of the business model. We have seen this movie before.

In the early first decade of the twentieth century, believers believe that e -commerce would turn the margins of retail. Amazon It proved them right by building an original digital retail seller, not by obtaining and converting serums or Barz and Nobel. In 2010, investors believe that programs will eat traditional industries. The winners built new software companies instead of the update modification of old companies.

The same lesson applies today, but with a narrower range. Artificial intelligence may convert some vocational services angles, especially when the current companies are pushed to adopt new tools by private stock owners who have clear control and incentives. We have seen it in sectors such as health care and financial services, where PE companies have prompted the adoption of AI-driving tools that are driven by artificial intelligence. But this differs from the Rollup Ai, which is chased by VCS-those that assume that companies with low margin and heavy service can be converted into software-like platforms by simply by including artificial intelligence. For those companies, the shift will not come from owning the service layer. It will come from new companies of artificial intelligence with various economies mainly.

The bottom line: Owning the program, not the service

AI Rollup thesis is an attempt to invest in investment capital to arbitrate the multiple gap between services and programs. But this gap is present for some reason. Services companies, and even very automated companies face different restrictions, different economies, and different customer relationships from software companies.

Polyai saw in 2019. Public markets see it now. AI is real. The opportunity to improve service services with artificial intelligence is real. The idea that this improvement turns them into software companies? It is unlikely to be real today, just as it was not in 2019.

Rollups may still offer returns, but not of the type VCS subscription type. At best, it is private property rights that support technology: heavy operational, evaluating them, and is unlikely to expand like programs.

The opinions expressed in cutting comments Fortune.com are only the opinions of their authors and do not necessarily reflect opinions and beliefs luck.

Read more: