top of page

AI Factory: changing perspectives

The very idea of building an organizational AI maturity model rose from the existence of AI factory kind of organizations. There were the Googles, the Ubers, the Amazons, the Airbnb who leveraged on the capabilities of AI and reinvented industries. We studied their business models, their motivation structure, their talent strategy, their computational architectures and wanted to learn from them. But in the breaks of the Executive MBA class where we taught these people came to us privately: "I understand these fantastic concepts, but we will not be Google, what is our journey? What can we do to improve?" These ideas were so far-fetched, that it seemed impossible to get there. That is why we started to build a roadmap, to build this bridge to the last stage, but at the same time to acknowledge, that AI Factory might not be for everyone. Yet it was a very exciting experience from time to time, that when managers understood the concepts behind AI Factories and inspired to find disruptive ideas they surprised themselves as well to have been able to find these opportunities. Maybe not for the whole organization, but a significant business opportunity.

AI Factory

What is an AI Factory?

We borrowed the term "AI Factory" from the fascinating book of Marco Iantisi and Karim R. Lakhani titled Competing in the age of AI. They use this label to highlight the pattern of companies whose competitive advantage rely on AI, therefore they must be at the forefront of innovation in AI. Companies who were born digital or who reinvented themselves or a spinoff of theirs around the novel capabilities of AI. They usually leverage on both of the network and the learning effects and with these can identify and manage a value creation potential where more clients are not harder but easier to serve and more clients bring more value to the existing clients.


The network effect is the phenomenon of the fax machine: one fax machine has no value no matter the quality of the tech. You simply can't send a fax to anyone else. If there are others who have faxes we can send faxes to each other. This is the basic value add of social media connecting people with each other and people with advertisers. The principle behind Amazon, Netflix and Airbnb connecting sellers and buyers at scale. But not all companies who leverage on the network effect are AI factories. Fax machines, traditional telecom networks, or railway networks are not AI factories, although they exploit the network effect. The competitive advantages of AI factories kick in when they start to leverage the learning effect as well.


The learning effect is when we can combine the data capital flowing through our systems with the information analysis potential of artificial intelligence to create additional insights for every customer we serve or improve every solution we produce. A typical example is Google Search or the recommendation systems behind Amazon marketplaces or Netflix: because they have so many customers they know so much that information surplus starts to be beneficial for everyone. A less trivial model is John Deer's predictive maintenance model: because they know so much about all the tractors and combines, they could leverage this to predict when a specific machine is going to break down. But this still feels like the core business, although this is a significant competitive advantage.


Self-disruption starts when we can produce additional business models, or revenue streams because we have access to these data and technologies. Like as a telecommunications company starts to sell marketing insights about the aggregated information of the users (where can you put your billboard or where should you locate your retail shop?) and lower their original prices because of this. Or a physical security company starts to sell their own camera based stuff allocation software to their potential competitors. Or a traditional glass manufacturing company starts to invest in a platform to provide 3D printer glass ink and with it connect sellers and buyers and learn more about the problems running through them.


Although there is a significant amount of AI Factories who are just leveraging the capabilities of AI, we saw that they are typically digital natives, not more traditional companies who transform themselves and exploit the capabilities of AI. Although still rare, the more common form of entering the stage of an AI Factory (rather than be born as one) is the combination of exploiting the data capital, the technological capabilities and adding an innovative business model thinking to it.



Becoming an AI Factory

From the Implementation stage companies will face opportunities that they let go. They realize that in a very segment where they implemented an AI solution they could even power it up and sell it to others as well. Traditional companies are traditional because they have a market, they can meet client needs, they know their job. They have an expertise in a niche, and they run an operation. When they start to automate it or upgrade it with the power of AI and create a temporary competitive advantage they ask themselves: "this is so great, why don't we make more money with this?" But they end up saying that they are not a digital business, they don't have a startup culture and they don't follow up on these. Those who do often find themselves stuck between two worlds: the corporate world and the startup world. As they improve from implementing AI to becoming data-driven and adaptive, the data-capital is more and more available, and the opportunities are more and more evident.


Becoming an AI Factory is usually not an integrative, but a separative journey. There is a new department who is provided with resources and an opportunity to build a separate culture to start exploiting scalable opportunities. They are allowed to build on top of the already existing data capital, market insights and technology stack, but they are not bound by the vision and the strategy of the company. If they are born and grown internally, they are usually aligning with the mission of the company, but totally rethinking what that might mean, because we have the technology to do so. In glass manufacturing the mission might be to provide glass based products with supreme chemical engineering skills. But with digitalisation and AI we can provide this without manufacturing to end customers and providing 3D printers with ink and a platform to know as much as possible about the problems.


There are two levels of aim for when wanting to play in the competition of AI Factories. We can build a leadway for emerging ideas or we can actively explore disruptive opportunities. Usually, a leadway is an organic consequence of a great idea and a great team trying to shake the thinking and a great management who realize that we have to make space for them: exceptions in policies, access to data, resources to follow-up on their idea and mission. The management contribution is to realize the pitch when they hear it and to open a path to let it grow. Management wisdom is when they can actively keep that path open for later ideas. There are also companies, who are actively and continuously exploiting the AI and data enabled disruptive opportunities. If this becomes part of the strategy, then new roles might appear (like AI intrapreneur), there has to be a segment of the architecture that is very much state-of-the-art, there are processes for choosing internal AI startup candidates and providing them with resources, or even an internal incubator where there are people with startup experiences, scalable business mindset, where business model innovation is just as important as technological innovation. Providing products and services that build on AI comes with a new array of challenges from availability of computing power through developing solid AI governance to comply with upcoming regulations and avoid AI scandals to a very nuanced value-chain thinking with ecosystem mindset to be able to find a segment of the AI based value chain that we can contribute to permanently. It becomes more often than not a B2B business rather than a B2C with new sales, new talent acquisition and new maintenance problems. But aiming for it is the only way to face the final threat of AI: being disrupted by a technology first newcomer.



Looking back on the maturity stages

There will always be companies who make a huge jump in technological breakthroughs, like OpenAI or Google. They will always steal the show from everyone else who is already on the market making efforts to meet specific needs. But they are not the real role models for companies, who are not digital natives, no matter how tempting they are. They have a single focus competitive advantage and a single focus strategy: AI. The journey of a normal company is to build a secondary focus besides their core business excellence: exploiting AI in their niche domain. Starting by exploring and implementing and building a solid foundation as data-driven then even be able to adapt to the speed of technology and the impact in brings to the domains beyond improving efficiency. Meeting new needs or finding totally new ways of meeting existing needs. AI is not going to do that. People and organizations using AI will.


コメント


bottom of page