AI and the Digital Operating Model

The robots aren’t taking over -- they already have, and that’s not a bad thing.

The New Rules

The robots aren’t taking over — they already have, and that’s not a bad thing.

Artificial Intelligence as it currently exists isn’t all that sophisticated. It’s not “strong” AI which is the stuff of science fiction where computers can simulate human behaviors or enact human reasoning. In fact, today’s AI is “weak” AI consisting of programs that simply perform routine tasks traditionally assigned to humans. AI recommends songs on Spotify like a DJ, sets prices on Amazon like a clerk, and dispatches the Uber closest to you like a dispatcher.

By eliminating traditional restraints such as human response time and individual bias, “weak” AI streamlines decision making into an instant, accurate process that is devoid of human error, transforming the rules of competition and changing how successful companies do business.

The AI Factory

At the center of this change is the AI Factory, the software and algorithms that make these billions of simple decisions every day. The AI Factory runs the millions of daily ads on platforms like Google and Facebook, sets the prices in stores across the country, and updates traffic and weather conditions instantaneously.

There are four aspects of the AI process that are key in allowing the factory to work successfully. First, there must be a data pipeline that gathers, combines, and safeguards data. The second is algorithm development, which generates predictions for the best decisions to make. The third is experimentation, where the algorithms created are tested to ensure they achieve the intended results. And the fourth is infrastructure, the systems that enable the successful decisions to connect with users.

Gather data, make predictions, test predictions, implement successful decisions, lather, rinse, repeat. The AI Factory does this millions of times a second across all industries, improving their decisions all the time. Of course, with AI making all the decisions, the humans who once did the job are being pushed to the side, which has worrying consequences but has allowed companies to scale like never before.

Scale has been a central concept in business since at least the First Industrial Revolution when it became clear that large modern industrial businesses could reach previously unattainable levels of production at much lower unit cost, giving them a significant edge over smaller rivals. In addition, these larger companies could create a larger variety of products, expanding their scope and market share even further. In order to maintain their powerful position, these firms continued to push for improvement and innovation, introducing a third requirement for successful modern businesses: learning. Since then, scale, scope, and learning have come to be thought of as the three key components that drive a company’s success.

No Limit

In traditional operating models, scale inevitably reaches a point where it begins to deliver diminishing returns. That is, when a company is founded, every new customer adds great value to the company. As the company scales up, the value of each new customer declines, until only incremental growth is possible, regardless of how many new customers they attract. With AI, the limits of scale, along with scope and learning, are practically removed, allowing for businesses to grow like never before.

Unlike traditional processes, AI-driven processes can be scaled up rapidly, and allow for much greater scope by easily connecting to other businesses. With AI at the helm, every second is an opportunity for learning and improvement, giving companies the power to create more accurate and in-depth customer behavior models than ever before, and the ability to quickly adapt new strategies as needed.

In the digital operating model, as new customers are acquired, data builds on itself and creates new channels of value and connectivity. Using this data, AI reinforces itself over time and accumulates even more data that directly leads to more efficient and accurate decision making. Of course, this model takes time to build. In the traditional model, early users are the most valuable. In the digital model, the full value of AI only comes into play when a sufficient data threshold is crossed, at which point scale, scope and learning are exponentially accelerated and traditional economies of scale are eliminated.

Closing the Silos

When we look at the current landscape and see businesses with traditional models collide with AI-driven businesses, it’s clear that traditional businesses with highly specialized internal divisions are the ones losing out the fastest to their AI-enabled competition. The reason is simple, these businesses’ internal operations are siloed, and silos are the opposite of AI powered growth. To be properly informed and make the best decisions, AI needs an integrated core of data, not organizational divisions that don’t collaborate. When the various aspects of a company work independently of each other in their own little silos, the company is disconnected and unable to fully tap into the biggest asset they have: their customers.

For traditional companies, the shift to a digital operating model is difficult but possible, and the first step is eliminating silos. Ultra-specialization is a dated model that even in the best case scenario will eventually hit the ceiling of its own success. Once the silos are gone, it becomes possible to accumulate data across all aspects of a business and allow a well-informed AI to begin making decisions, learn from those decisions, and improve upon those decisions to achieve maximum growth. AI-based models do take time to develop up front, and switching over from a traditional model can be tricky. However, the benefits far outweigh the growing pains, which is why traditional companies like Walmart, Nordstrom, Visa, and many more have started building greater focus on data and analytics into their businesses.

AI Leadership

Though the benefits are plain to see, as AI aggregates data and value for companies, it can also aggregate harm if deployed improperly, or without caution. Remember, all we have is “weak” AI; it can only perform the routine tasks we assign. However, where once one bad inspector may have let a few faulty products out of the factory, a faulty AI can make that mistake a million times a second, compounding damage at a previously unthinkable speed. In addition, it’s always worth remembering that while AI “learns,” it is programmed to do so by humans, and because we are human, bias, personal vulnerability, security, and connectivity can be compromised on a massive scale when we inadvertently insert these flaws into the algorithms we entrust with decision making.

The benefits of AI are far too great to ignore. From great music to low-priced travel, AI has improved our lives tremendously and is changing the way successful companies do business. Embracing AI and the digital operating model is the only way forward, but we must be careful and considerate in how we enact and implement this powerful technology. AI may be replacing people in many tasks, but it is only as good at the people who design and administer it.