Firm Valuation When AI Shapes the Business Model: A Milestone-Based Real-Options Framework for the AI Valuation Uncertainty Problem

Valuing a company is traditionally a matter of looking at its past to predict its future. If a business has a long history of selling products, paying bills, and making a profit, experts can use those numbers to estimate what the company will be worth tomorrow. They calculate the money the business will likely make in the coming years and adjust that figure to account for the risk that things might go wrong. This method works well for stable businesses, like a bakery that has been open for twenty years or a factory that has been running smoothly for decades. However, the rules change completely when a company is built around artificial intelligence. These firms often have no history of profits, their business models are being rewritten in real time, and their value depends entirely on whether they can successfully reach a series of difficult future goals. When a company is trying to build something that has never existed before, the old tools for measuring value often fail because they cannot see the specific risks and opportunities hidden inside the plan.

A new study from the Swissi Institute for AI and the University of Applied Sciences Nürtingen-Geislingen proposes a different way to look at these companies. The researchers argue that standard methods are too vague when applied to artificial intelligence because they squeeze all the complex uncertainties into a single, opaque number. Instead of guessing a general growth rate, the team developed a structured protocol that breaks a company's value down into a series of specific, checkable steps. They treat the journey of integrating artificial intelligence not as a smooth line, but as a path with gates. At each gate, the company must prove it can do something specific—like getting a new software system to work, passing a regulatory check, or finding its first paying customer. If the company passes the gate, it moves forward and its value increases. If it fails, the path ends, and the potential value disappears. By mapping out these gates and assigning a realistic chance of success to each one, the researchers created a way to see exactly where the value comes from and where the risk lies.

The core of this new approach is a simple but powerful idea: not all companies using artificial intelligence are the same. The researchers created a classification system that separates firms into two main groups. The first group consists of companies that use artificial intelligence as a tool to improve their existing work, such as a bank using a computer program to check credit scores faster. The second group consists of companies whose entire product is the artificial intelligence itself, like a firm that builds a new type of smart software to sell to others. Within the first group, the researchers further divided companies based on how deeply they have integrated the technology. Some use it just a little, like adding a chatbot to a website, while others have rebuilt their entire operations around it. The study found that the deeper the integration, the more the company's value depends on future possibilities rather than current profits. A company that has only dipped its toes into artificial intelligence has a predictable value based on its current sales. But a company that has fully rebuilt its business around the technology is essentially a collection of future options, where the real value lies in what it might be able to do next.

To figure out how likely a company is to succeed at each step, the researchers introduced a method that replaces guesswork with a structured conversation among experts. They asked a panel of people, including the company's own leaders and outside specialists, to compare different factors that determine success. These factors included whether the technology actually works, if the company has the right people to run it, and if the government regulations will allow it to proceed. By having these experts compare the factors against each other in pairs, the system generates a score that reflects how ready the company is to move to the next stage. This score is then calibrated against real-world data from similar technology projects to turn it into a specific probability. This process ensures that the final valuation is not just a single opinion, but a transparent calculation where every number can be traced back to a specific assumption. If two experts disagree on the value of a company, they can now pinpoint exactly where their disagreement lies: is it about the chance of passing a specific test, or the amount of money that test will generate?

The researchers tested their new framework on a real Swiss startup that builds software to manage energy portfolios. This company, which sells its artificial intelligence product directly to customers, was in the early stages of development with no revenue history at the time of the study. Using the old methods, valuing such a company would have required a large, unexplained guess about its future growth. Using the new framework, the researchers mapped out four major milestones the company needed to reach over the next two years. These milestones included stabilizing the system for initial tests, securing paid pilot customers, generating its first recurring revenue, and finally proving it could sell the product to many customers at once. By calculating the value of each step and the probability of reaching it, they arrived at a total company value of 100 million Swiss francs. This number matched the company's own internal estimate, but the new method revealed something crucial: more than half of that value depended entirely on the final milestone, which was the hardest to achieve.

This finding highlights a major shift in how risk is understood in the artificial intelligence sector. The study shows that for companies selling their own artificial intelligence, the risk is not spread evenly across the journey. Instead, the risk concentrates heavily in the later stages. The early steps, like building a working prototype, are relatively safe and likely to succeed. But the later steps, which involve proving that the market wants the product and that the company can scale up, carry the most uncertainty. The framework makes this clear by showing that if the final milestone fails, the company loses roughly half of its total value. This is different from traditional businesses, where risk is often more evenly distributed. The new method also accounts for the costs that other valuation tools often hide. It separates the money spent on building the technology from the money spent on complying with laws and regulations, ensuring that these costs are counted clearly rather than being buried in a general estimate.

The study also addresses a common problem in how companies report their finances. In many European countries, the rules for accounting require companies to treat money spent on developing new technology as an expense, which makes the company look like it is losing money on paper. In other places, companies can record this spending as an asset, making them look more valuable. This difference in accounting rules can make two identical companies look completely different to investors. The new framework bypasses this confusion by looking at the actual economic reality of the business rather than the accounting labels. It treats the money spent on development as a real cost regardless of how it is recorded in the books, allowing for a fair comparison between companies in different countries or under different rules. This approach ensures that the valuation reflects the true potential of the business, not just the way it is written down in a ledger.

While the study provides a clear and logical way to value these companies, the authors are careful to note that this is a demonstration of a new method, not a final proof that it is perfect. They tested the framework on a single company with very clear information, which makes it easier to apply the method. They acknowledge that real-world situations can be messier, with companies that have mixed levels of technology integration or less clear data. The researchers suggest that the next step is to test this method on many different companies to see if it works as well in practice as it does in theory. They also point out that the rules for artificial intelligence are changing rapidly, and the framework will need to adapt as new laws are passed. However, the core idea remains solid: by breaking down the complex journey of artificial intelligence into smaller, measurable steps, we can stop guessing and start understanding exactly what drives the value of these modern companies.

The implications of this work extend beyond just calculating a number. It changes the conversation between investors and founders. Instead of arguing over a single, vague figure, they can now discuss the specific steps a company needs to take to succeed. If an investor thinks a company is too risky, they can point to a specific milestone that they believe is unlikely to be reached. If a founder thinks their company is undervalued, they can show how their plan to reach the next gate is more solid than the investor realizes. This transparency helps both sides understand the true nature of the risk and the opportunity. It also helps regulators and policymakers see where the real bottlenecks are in the artificial intelligence industry. By identifying which steps are the hardest and where the most value is created, the framework can guide decisions about where to invest resources and how to support the growth of these new technologies.

Ultimately, this research offers a way to bring clarity to a field that has often been shrouded in mystery. For decades, valuing a company has been a mix of art and science, relying on experience and intuition. With the rise of artificial intelligence, the old intuition is no longer enough because the rules of the game have changed. The new framework provides a map for navigating this new terrain. It does not promise to predict the future with certainty, but it does promise to make the uncertainties visible. By showing exactly where the risks lie and what it takes to overcome them, it allows everyone involved to make better decisions. Whether you are an investor looking for the next big thing, a founder trying to build a new company, or a policymaker trying to understand the economy, this approach offers a way to see the real structure of value in the age of artificial intelligence. It turns a black box into a series of clear, understandable steps, making the future of business a little less opaque and a little more knowable.