It is easy to take a reductionist view when thinking about digital transformation. Fix enough of the granular systems that run your finance, logistics, marketing, and HR, and you will eventually reinvent yourself — or so the wishful thinking goes. In truth, when an organization is reborn with machine intelligence at its core, it is not just faster or better than its peers; it becomes different. And different is what you need if you plan to reshape industries and redefine competition in your market.

A successful digital transformation can be hard to predict or plan; it is often the result of new customer interactions, new combinations of talent and teams, unexpected alliances with new partners, and entirely new business models. These components are constantly evolving, shaped, and influenced by algorithmic systems, aggregated in such a way that their collective behavior is more than the sum of their parts. More is different. Just as water becomes ice when cold enough, or graphite turns into diamond under enough pressure, at a critical point, more data and algorithms can transform an organization or an industry into something else entirely.

That raises a question for leaders: how do you navigate a transformation from what you know to what you have yet to define? What you need is an emergent approach to digital transformation, focused on three principles:

Act ahead of the phase transition.

In July 2021, something extraordinary happened in the car industry. Bugatti, founded in 1909 and maker of some of the world’s most expensive hypercars, announced their intention to merge with Rimac Automobili, a Croatian automotive startup begun in 2009 and run by Mate Rimac, its 33-year-old founder. The distance between the two organizations was more than a mere hundred years; it was the technological and philosophical gap between two very different forms of mobility.

Volkswagen, Bugatti’s parent company since it took control in 1998, had spent over $2.4 billion to develop a combustion engine vehicle that could exceed 300 mph and achieve 0-60 mph in under 2.5 seconds. However, in an increasingly electric-powered future, Bugatti’s fate was far from certain. Creating the next generation of high-performance electric cars would not only require substantial investment, but also an entirely new set of capabilities to develop an AI-powered electric power train. Despite Rimac’s newcomer status, Volkswagen’s leaders realized a phase transition was at hand and that they had to take early action. Rather than simply outsourcing to the startup, they proposed a partnership.

It was a wise move. AI, automation, and algorithms now infuse every stage of modern hypercar development — from aerodynamic design to battery optimization. However, the value of the Bugatti merger for VW was more than just AI expertise; Rimac had evolved a unique set of capabilities due to its challenger status in the automotive industry. Without the resources to take a more traditional approach, they were forced to innovate.

“When we started our company, we had no choice but to develop our own technology,” Mate Rimac explained to me. “We couldn’t afford to pay a supplier royalties for existing technology or for them to develop technologies for us. Also, the performance and features that we wanted were not possible with existing technology. So, we built most key systems ourselves, doing things quickly and inexpensively, with extensive 3D modeling and digital testing.”

Phase transitions, unfortunately, are rarely one-offs. Technology disruption is an overture, setting the stage for a cascade of changes in business models, customer behaviors and industry dynamics. While VW was swift to partner with a emerging technology player, transportation is still in the early stages of its 21st century reinvention. “It will be the biggest disruption we’ve ever had in the automotive industry,” Rimac says. “When people don’t drive and own cars any longer, and mobility becomes a service, owning a car is not as necessary as it once was. OEMs will have to become mobility providers, offering their cars on subscription, available at the press of a button for autonomous journeys. It’s a huge opportunity for new players.”

Amplify learning and adaption.

In an emergent digital strategy, learning is what allows you to leverage your digitalization efforts to evolve faster than your competition. For Kevin Johnson, President and Chief Executive Officer at Starbucks, a key driver powering the company’s digital transformation has been their ability to learn at scale. In his view, increasing the innovation velocity at Starbucks demanded a learn and adapt approach. Their current mantra is to “go from idea to action in 100 days.”

The Starbucks approach to AI-powered amplification has supported the firm’s transformation from a coffee retailer to a data-driven technology platform. Serving more than 100 million customers weekly at 31,000 stores, with 24.8 million registered and active mobile app users in the U.S. alone, Starbucks has built a learning machine for aggregating vast amounts of valuable data about customer behavior and preferences. At the core of this platform is Starbucks’ digital flywheel strategy that links a powerful program of rewards, simplified payment methods, personalization in the form of special offers, and quick, convenient order processes.

Starbucks was quick to embrace mobile ordering and payment, well ahead of more technologically sophisticated peers. It started accepting mobile payments nationally back in 2011. By the time Apple got around to rolling out mobile payments in 2014, Starbucks was already processing 7 million weekly mobile payment transactions in the U.S., while growing its database of mobile app users. In the fourth quarter of 2021, 51% of its U.S. company-operated sales were driven by customers who were Starbucks Rewards members.

But the company went further, using its growth in mobile data to amplify its ability to lock in users with rewards, digital payments, and mobile orders. The Starbucks AI engine processes everything from data about what times of day people usually order to which drinks they typically like, which can then be combined with other data like geolocation, weather, and seasonality to offer up personalized recommendations, offers, or even quests and challenges to earn extra rewards points. The level of digital engagement generated by this platform became particularly important during the Covid-19 crisis, when many physical stores had to close. Today, drive-thru and Mobile Order & Pay (MOP) together account for 70% of transactions — a 15% increase from pre-pandemic levels.

The digital flywheel is just part of Starbuck’s efforts to leverage learning. They are also embracing AI and automation as part of their broader operating model. According to Johnson, “Our Deep Brew, artificial intelligence platform that has automated daily inventory management and store staffing and training improvements was designed to reduce complexity in our stores.”

Invest in capabilities, not competencies.

How do you plan for an unpredictable future? Counterintuitively, the best response to uncertainty is not a retreat to the familiar, but a bet on your capacity to explore the unknown. When news of the pandemic first emerged, Moderna Therapeutics was working on several mRNA-based medicines, including those focused on heart disease, Zika virus, and cancer. However, Moderna’s founders believed that if mRNA technology worked for one application, it could work for countless more, simply by changing the information and coding for a new application. It was thanks to this capability-centric approach that not long after Chinese scientists first put the genetic sequence of the novel coronavirus online, Moderna was able to develop and release an entirely new Covid-19 vaccine in a matter of months — an extraordinary achievement.

Companies often invest in competencies (things they do well), rather than capabilities (things that they might do well). In a way, it’s a trade-off that’s similar to the classic explore-exploit dilemma. How much time and resources do you spend investigating your options before you pick one? While upgrading your legacy systems may be initially appealing, this type of reductionist approach risks cutting off your exploration phase too early. Instead, what if you could design your future organization as an open-ended, digital platform with the potential to unlock new opportunities?

When I spoke with Dave Johnson, Chief Data and AI Officer at Moderna, he explained that it was the software-like, digital nature of mRNA technology that inspired the company to fashion themselves in the form of a new type of digital biotech company with AI, algorithms, and automation at its core.

“We had constructed this large-scale pre-clinical manufacturing suite that allowed our scientists to order mRNA through online digital tools, to use AI algorithms to help optimize them, and then feed them into a high throughput, massively parallel, highly automated small-scale manufacturing facility to produce them as quickly as possible,” says Johnson.

Arguably, what makes Moderna so effective is its capacity to align its digital infrastructure with its business strategy of pursuing multiple mRNA-based therapeutics in parallel. According to Johnson, they think holistically about their platform: investing in entirely digital systems and algorithmic design tools, capturing data in a very structured and rich way, and finally integrating those systems and AI models together in an efficient and reliable production environment.

“Our digital platform enabled us to build a research engine that let us go from drug concept to clinical-grade material in only 42 days,” says Johnson, speaking of the company’s experience in rapidly developing their Covid-19 vaccine. “Most companies have to go through a much different process where you have to not only invent the idea, but invent how to manufacture it at the same time. Because we had built this platform, we were just able to leverage it.”

The problem with any digital transformation plan is just that; it is a plan, rather than a path. Organizations and markets are complex adaptive systems; they have emergent properties that are not present in their smaller pieces and cannot be replicated simply by digitizing processes or integrating new software. Nevertheless, if you can overcome the need for reductionist certainty, there is an elegant symmetry to taking a bottom-up approach to digital transformation. After all, machine learning systems themselves are self-organizing networks from which emerge insights, predictions, and recommendations. Whether you are a startup trying to disrupt an industry or a traditional incumbent reimagining itself, an emergent digital strategy allows you to maintain your optionality while also acknowledging that when things do change, they are likely to do so overnight.