Platform companies are challenging every industry and already dominate many. Why is that?
Platforms thrive on the network effect, i.e, every new user creates value for other users of the platform. With data and AI, the network effect becomes even stronger. AI enables platform companies to learn faster than the competition, creating a virtuous circle.
Data is the raw material for AI
Platform winners collect data on their users and extract learnings all the time. Every improvement attracts more users from whom they can learn, even more, accelerating the value resulting from network effects.
Data is the raw material for AI. Therefore, companies need to understand what data they need to train an AI system, what data they already have, and what data needs to be acquired. Sometimes data comes from unexpected sources. For example, Uber uses data from your phone’s battery status to determine your willingness to pay. Namely, an empty battery means more urgency to get a ride.
Data is everywhere if you just know where to look. For example, existing floor plans are data that can be used to teach AI how to design new ones. So-called GAN networks (General Adversarial Networks) are advanced deep-learning-based AI models that learn underlying patterns from a set of training examples, by themselves discovering the rules of how they are composed. Silo AI has worked together with an architect company Belatchew in Sweden to pilot the use of GANs in creating novel floorplans. In this case, the images of floorplans teach the generative AI how floor plans are generated. As a result, the AI solution is able to generate novel floor plans to help architects in their design work.
Network effects are amplified with data
Network effects are the most important success factors behind platforms. The value of the network increases for every new user. Think Facebook or telephone network. But how are network effects created? Data plays a critical role. Every new user creates more data that can be used to improve the system. For example, Upwork is a freelancing platform where users rank each task. This is data that helps the platform to create better matches between individuals and enterprises. Every new user adds value to other users.
Orica is a more than 100-year-old Australian company and the world’s largest provider of commercial explosives and blasting systems. The company realized its engineers were advising customers on implementing blasts. They had valuable knowledge that they could teach to AI. Orica developed an AI system for this. Every time a new blast is designed, new data is fed into the system. The more its customers use its service, the more data it collects, and the better its AI becomes. Every new user creates value for other users through data and thus, amplifies network effects.
Learning loop creates a competitive advantage
When we use Google, it learns from every search and adjusts its AI algorithms to be able to provide better search results. Every user contributes to this learning loop and Google’s competitive advantage in search extends further.
Learning loop is like a hard-working student. It constantly draws in new learnings and gets better. However, compared to humans, there is one crucial difference: the speed and amount of experience cumulation. That’s why AI-based learning loops are so powerful in creating competitive advantage and why incumbents face challenges from newcomers. Even if a new entrant is not better when they start, they can become better much faster with an AI-powered learning loop. Of course, incumbents can do the same. But, their success makes them slower. Therefore, it’s critical to create a new mindset, the platform mindset.
A good example of a newcomer is Tesla. When you start driving your Tesla, it collects data through cameras, radar, and other sensors. Tesla uses this data to train and improve its AI algorithms. Step-by-step, Tesla’s self-driving capability becomes better. And the more cars it sells the more data it collects.
Putting AI into practice – machine learning operations (MLOPs)
But how to make this all happen. Many companies try individual AI projects, proofs-of-concept but don’t make an impact in business. Why is that? To succeed, companies need to make AI and machine learning development activities systematic. Furthermore, they need to connect the machine learning projects to its business and IT infrastructure while bringing automation to the relevant parts of the machine learning workflow. Machine learning operations (MLOps) improves all of this.
MLOps is a way of productizing machine learning projects. It’s about considering the whole lifecycle from the more experimental R&D phases to the deployment of production-ready models. The most significant potential is in scaling the machine learning activities from individual projects to the entire organization across the whole company through shared practices and enabling developers to a more considerable extent than before.
Silo AI customer Beamex, a company focused on calibration, wanted to introduce computer vision-based ways to read digital screens during instrument calibration. In the past, users were required to manually enter values, and as there can be hundreds of measurements to be done during calibration, the risk for human error was high. Silo AI’s modern data-driven machine learning approach helps to automatically read the value from the calibrated device by using the mobile device’s camera and AI models to recognize the measured value from the calibrated device display.
Connect the dots: network effect, learning loop, and new business model
In addition to network effects and the AI learning loop, successful platform companies have developed a new business model. However, changing your business model is hard for successful incumbents. Especially, if the current business is fighting change. That is why Timo Vuori and I wrote our new book Platform Strategy, to spell out the business mechanics of creating a platform, to tell people how to take the first step and get the management team and their stakeholders, board, and employees on board.