How AI Companies Really Make Money: A Look Inside the Strategy Kitchen

You see it on the news all the time. Some AI-startup raises a billion ringgit. Another AI-startup releases a new chatbot. The world seems to be totally obsessed with “AGI” and “compute power.” And when you sit with the people who really run these AI-companies (not their founders that are always on stage), they are having an entirely different conversation. They have a vastly different outlook on what AI technology firms strategy. Sounds fancy huh? But fundamentally they are talking about how do we keep from running out of money while we build this thing? Allow me to break it down like you are in a kopitiam, no fancy talk, no “synergy,” just raw data.

Why Most People Get AI Business Models Wrong

Most of us use ChatGPT. It’s free (mostly) so we tend to associate AI as being inexpensive to create. However, it actually costs a lot of money to run a large language model. Every time someone asks a question, there’s an electric charge and computing power being used from a data center somewhere. So, how are these companies able to stay in business? Here’s the mistake everyone makes. They believe that AI companies are like how software companies operated back in 2010. They built apps and sold subscriptions; simple. But the reality is AI is different. The technology evolves every three months requiring you to revise your product roadmap every quarter, otherwise, if you have a highly structured AI product roadmap you will fail.

I was speaking to a buddy of mine that works in cloud infrastructure, and he said, “The companies that are winning in the AI space are not the ones with the best models, but the ones who can alter their pricing models at the drop of a hat.” If you look at the most recent movements within the industry, you will find the companies that are winning are moving from “one size fits all” to verticalised AI market strategies. These companies are targeting specific industries instead of trying to develop an answer for all humans. They are targeting specific industries such as lawyers, doctors, and insurance agents. Why? Because lawyers have deep pockets, and they typically have very specific problems. If you develop a model that can review contracts relevant to Malaysian corporate law (not simply “legal stuff”) you can charge 10x more. That is the transition; from generalised to specialised; from cool to useful.

AI technology firms’ strategy- “Platform vs Product” Debate Nobody Talks About

Every AI company needs to answer one crucial question; Are you building a platform or a product? This question may seem technical, but more importantly this question is about money. A product does one thing really well. You buy it, you use it. An example of a product would be an AI that can write marketing emails for shoe stores. A platform on the other hand, allows other people to build things on top of it. In this case, the underlying model that the shoe store AI uses would be a platform. Why does all of this matter from AI technology firms’ strategy? Because products have a ceiling or are limited by how many sales can be made. Platforms however, can continue to grow indefinitely.

If you create a product, you will have to sell your product to every customer individually. If you create a platform, then your platform will be sold by the many different developers creating products on top of your platform. You will take a percentage of the revenue from those sales. The downside, is that creating a platform is much more difficult than creating a product. You need to create security, reliability, and the necessary boring infrastructure that delights developers. We are currently witnessing this shift in real time, as many well-known AI brands are now heavily investing in platforms by developing APIs, developer tools, and building partnerships. Conversely, many are creating premium products designed for high-value customers. Which is preferable really comes down to your available resources, but as you will see with historical evidence, the major players in the AI ecosystem will always attempt to become a platform for long term financial gain.

How Data Changes AI technology firms strategy

As we often hear, a moat in business refers to an area that gives you an advantage over your competition and prevents them from copying you. In artificial intelligence, the moat is usually not the algorithm. That is often open-source in some form. Anyone can write (or copy) the code. Instead, the moat exists in your access to data about your business or your industry. Let me provide an example of this. There is a company developing AI for hospitals in Malaysia that has over 10 years’ worth of anonymised patient records, doctors’ notes, and outcomes from treatment as part of their data feed to help develop their system. A new startup can create a better algorithm but it won’t matter because it does not have the same level of access to patient data that the first company does (i.e. they won’t understand how Malaysian doctors write their notes or what slang they use to describe different types of symptoms). This illustrates how a data-driven business strategy is executed in the real world.

You will also see many companies entering into partnerships or alliances with other companies to use AI. Companies that develop large-scale implementations of artificial intelligence will usually only partner with other companies that have access to unique data sets such as banks, telcos, and government agencies. The common agreement is to provide these companies with AI solutions on either a free or low-cost basis in exchange for access to their data. This is a win-win for both the AI company and its partner company and ultimately a win for the end user because they have access to better services than before.

The “Go-to-Market” Reality Check

You have a model, you have data and you have a platform. Next you need customers and this is where your AI GTM strategy will separate the winners from the “we raised 20M and went out of business.” Most techies hate doing sales and they believe that if they build a great solution that people will come. In AI, this is a death wish! Here’s why – AI is still very much a mystery to everyday business folks. The owner of a restaurant in Penang doesn’t care about “neural networks.” He only wants to know, “Can your product help me figure out how much chicken to order so I don’t have to throw food out?” Therefore, your GTM strategy must begin with translating the features of your AI solution into business case outcomes.

The next step in your GTM strategy is your pricing model. Pricing for an AI solution is very different than pricing for a traditional software solution. With software, you will make money as more users use your application. But with an AI solution, every user query is costing you money. Therefore, if a business uses too much of your solution you will lose money. Therefore, you must have an AI monetization strategy that takes into consideration all three types of usage; limits per user, flat fee per user and overage per user. Some companies are going to a “consumption pricing model”. If you use it you pay for what you use. Others are remaining on a monthly subscription model. The smart ones are developing a hybrid pricing model. But here is the big secret, the real money in the AI space is not in subscriptions, it is in the competitive advantage of the AI firms. If you create a solution that solves a problem no one else is solving, you can charge what you want. It’s that straightforward!

Where Is All This Going?

We have talked about strategies: Tactical pricing. Go-to-market. Proprietary data. Platforms. However, how do these things play out in 5 years?When evaluating an organization’s long-term innovation strategy for Artificial Intelligence (AI), two key growth trends emerge: Agentic AI – AI that “does” not just responds. For example, an agentic AI that can negotiate with vendors, reserve flights, & process refunds will redefine how we utilize AI today. No longer is AI a “co-pilot,” but instead a full-time employee (i.e., agentic AI). Vertical AI Dominance – Vertical AI models will become commoditized, likely free-to-use offerings. The unique value of vertical AI will exist in specialized models designed for distinct market segments. Players who focus in this vertical AI sector will earn significant margins as their customers will find it very difficult to switch from their specialized models.

The establishment of a sound product vs platform strategy is the key to maintaining success in the future AI marketplace. If an organization invests heavily in its product offering too soon, it may run out of funds before welcoming their developer ecosystem to join them on the platform. Conversely, if excess funds are invested into the platform before product, there may be a growth ceiling related to those funds. The larger companies will continue pursuing an AI ecosystem strategy of acquiring start-up organizations with valuable data assets and unique distribution opportunities.