When the Rules are Changed Forever

As a futurist, I spend my life looking at trends, technologies and historical patterns of human behavior to assess how these elements will shape our future. When Shailesh asked me to consider writing a foreword for his book, I wanted to parse his book through that lens. How will AI and collaborative intelligence reshape customer expectations and experiences moving forward?

10-15 years ago, conventional wisdom was that the best form of customer service was interacting with a human. I've worked extensively reviewing customer surveys and market research for banking and financial services over the last two decades, and a decade ago you could see this assumption reflected in the data. Internet banking had already led to a significant reduction in branch traffic by the mid 2010s, but customers still cited the location of a branch as a major element behind their choice of bank – just in case something went wrong there was someone you could go yell at to get access to your money. Today, that data is increasingly absent from the research.

In November of 2025, the American Bankers Association released their annual National Survey on American's banking habits. More than half (54%) of surveyed customers said mobile was their top option for banking, and then 22% for online via PC. Only 9% of customers indicated the branch.

In every market around the world, we see pure play digital banks setting a clear standard in customer experience that incumbent banks are being forced to measure themselves against. Let me illustrate. NuBank is the largest company by market capitalization in Brazil, it is the largest bank in Latin America by customers (135 million as of 2026), and has the highest positive customer referral rates of any bank in the region. This reduces cost of acquisition and has helped NuBank scale. Itaú, their largest traditional competitor, has just 58 million customers.

In Europe, Revolut is on track to become the largest retail bank there by 2028, perhaps sooner. Revolut reported 2025 ratings of 4.9 on Apple's App Store, 4.8 on Google Play and 4.6 on Trustpilot. Its annual report also describes customer NPS as more than twice the industry average.

Today the top 20 Fintechs are outpacing the growth of the world's top 20 traditional retail institutions by 10x, showing nearly 150% customer growth in just the last 5 years compared with an anemic 2.8% for the traditionals. Wall street should be taking notice of the revenue performance of these players also with fintechs growing at 7 times that of the incumbents, at 32% Compound Annual Growth Rates (CAGR) versus less than 10%.

These fintechs now serve nearly 1 billion more customers than their 20th century counterparts, with an aggregate of 130% of the customer base of the blue chip FIs. NuBank serves more customers than all the top retail banks in Latin America combined. WeBank today serves more customers than all but 3 of the top Chinese banks, and they serve more customers than all the top 10 traditional banks in both the US and EU markets combined. This phenomenal customer growth simply could never have been accomplished by traditional banks operating branches or human-led advisor networks in the past.

Digital pure plays generally outperform traditional banks on customer advocacy and day-to-day satisfaction, but not yet on institutional trust. Their advantage comes primarily from low friction, transparency, speed, pricing and useful digital functionality—not from completely eliminating human advisors and support staff.

The strongest model is therefore not “digital versus human.” It is digital by default, human at the moment of consequence. But, AI changes all of that.

Nubank's recent customer-service AI work suggests that digital interaction can approach human performance. Across several deployed use cases, its AI support reached satisfaction levels within a few percentage points of expert human agents. In one card-delivery use case, improved AI produced a 37-percentage-point increase in transactional NPS and a 29-point increase in self-service.

For China's WeBank, the largest digital bank in the world, AI has already revolutionized the boundary where humans are needed to mitigate or address moments of genuine consequence. WeBank reports that they have been able to remove 98% of all traditional requirements for human interaction or support using Artificial Intelligence. Henry Ma, the CIO of WeBank has said “In most cases, if you need to interact with a human, that is a failure of design.”

As AI gets more capable, the gap between human performance in providing a differentiated experience and that of interactions with AI-level support diminishes down to zero. In fact, when we heavily personalize the experience with generative AI and better modeling, it is more than reasonable to expect that humans won't be able to compete with AI simply because they lack the context and data that AI will use to frame the experience with.

Let me illustrate using healthcare and medical advice. Already today we are seeing AI being used in a wide range of diagnostic applications in medicine. We are seeing companies like Insilico Medicine generate better drug candidates using AI. Google DeepMind's founders won the 2024 Nobel prize in Chemistry for their work in predicting protein structures, subsequently known as AlphaFold. Now his team has turned its attention to decoding the human genome in the same way. But let's look at the experience in a clinical setting to understand the dynamic emerging.

In any single year we add millions of new scientific research journals, the largest percentage of which is typically in healthcare and medicine. The PubMed Journal database adds over 1 million new citations annually, and major entities like the National Institutes of Health (NIH) traditionally fund research that adds more than 250,000 published articles each year. Now think about what this means when you step into a doctor's clinic with an ailment that requires diagnosis and treatment.

How many physicians will have read even 0.1% of those newly published insights and research to ensure that they have adequate data to provide you with an optimal diagnosis? AI can read every single research paper and article to ensure the best result.

In anticipating how customer experience is going to change over the next few years with AI we need to consider this information asymmetry that exists.

The reason we go to see a doctor for medical advice is he knows more about medicine than we do. The reason we go to see a lawyer is he undoubtedly has more legal experience than we do. The same is true of seeing a banker or financial advisor about our money problems. This is information asymmetry.

But when an AI can have massive information advantages, can understand your context, your likes, dislikes, behaviors and expectations of outcomes far better than a human expert sitting in a bank branch, a doctor's clinic, or a lawyer's office. Then the information asymmetry between human and AI is no longer a gap humans can bridge.

On top of that, we have advancing Agentic AI, which will within 24-48 months start to radically reframe customer expectations. Instead of exhaustively researching different brands, products or reviews, your agent will simply go out based on your expectations of an outcome and find the very optimal solution that meets your needs. It will do so by deep research that can be conducted across every vector in milliseconds, even negotiating with multiple other agents to package or repackage the optimal solution.

This is why you can't make this shit up anymore. The problem of how to design the best experience for a customer, is no longer one that is determined by the skill of the human on the frontline speaking with another human. Increasingly it is going to be abstracted by AI in a way that means the best experience is one limited by data injected into model training, and experiences generated in real-time by advanced, adaptive liquid interfaces and generative capabilities.

In this world if your business thinks that the best advice and experience is funneling a customer to a channel where they need to speak to a human, then you will be losing out to AI faster than you could ever imagine. As Shailesh articulates extremely well in the pages following, the best experience is no longer one that is simply a default to speak to a human, it is one that is designed for optimal and efficient engagement to simply solve the problem or reach the outcome your customer needs.

This is a design problem, and ultimately AI is going to design humans out of most day-to-day experiences because we just slow things down, or we don't have the data to optimally solve the problem.

The rules have forever changed when it comes to customer experience.