Your AI Is Aimed at the Wrong Customer Problem

\ When I review how a company has deployed AI in its customer-facing functions, I have learned to ask one question before any other: which layer of the customer relationship is it actually aimed at? Most of the time the answer is the same, and it is the layer that matters least. Over years of building and scaling digital products, I have come to see customer-facing AI as three distinct layers, stacked by how much strategic value they create. I call them doing, knowing, and anticipating. The pattern I see again and again is that companies pour their effort into the bottom layer, where the work is easiest to measure, and leave the top two layers, where the real advantage lives, almost untouched. \ \ Layer 1: Doing, where everyone starts The first layer is automation of customer-facing tasks. Drafting support replies, generating marketing content, summarizing call transcripts. This is where most companies begin, and for an understandable reason: the savings are easy to count and easy to justify to a budget owner. The trouble is that time saved on a task is a thin kind of value. It improves the cost of serving a customer you already understand. It does not deepen that understanding, and it does not tell you what to build next. The data points in the same direction. In its November 2025 study The State of AI in 2025: Agents, Innovation, and Transformation , McKinsey found that nearly nine in ten organizations now use AI in at least one function, yet only 39 percent report any EBIT impact at the enterprise level. A great deal of doing, in other words, has produced very little measurable value. Activity at Layer 1 is not translating into advantage. Layer 2: Knowing, where value begins The second layer uses AI to turn unstructured customer signal into structured insight. Most organizations sit on enormous volumes of messy customer data: support conversations, reviews, session recordings, sales notes. The capability that matters is pulling meaningful patterns out of that material and organizing it the way a strategist would, by customer segment and by the job the customer is trying to get done. This is where AI stops being a cost tool and starts being an insight engine. On one engagement, I designed the commercial core of a national last-mile delivery company built from a blank page for a global restaurant brand. To ground it, I ran more than thirty in-depth interviews across the delivery chain and a conjoint study of more than two hundred respondents that measured how much customers actually valued each feature of a delivery. The team had assumed price would be the deciding factor; the research showed the opposite, that customers valued the reliability and quality of the courier more highly than the price of the delivery. That single finding, what customers did value rather than what we assumed they wanted, changed the product, the pricing, and the order in which the company entered its markets. This is also where the money is. In The Economic Potential of Generative AI: The Next Productivity Frontier (June 2023), McKinsey estimated that generative AI could add 2.6 trillion to 4.4 trillion dollars in annual value across the use cases studied, and that about 75 percent of it falls into four functions: customer operations, marketing and sales, software engineering, and research and development. Two of those four are customer-facing. The value is not sitting in the back office. It is sitting in how well a company knows and serves its customers. \ \ Layer 3: Anticipating, where advantage compounds The top layer predicts the next job the customer will need done and acts before the customer asks. This is the layer that changes the economics of a customer relationship, because it moves the company from reacting to demand to shaping it. The shift is already visible in how leading companies listen. In its article Enhancing Customer Experience in the Digital Age , McKinsey notes that the survey, long the backbone of customer understanding, is losing power as response rates fall, and that the advantage is moving to companies that use operational and behavioral data to predict what will resonate rather than asking after the fact. In the same work, McKinsey reports that 71 percent of consumers expect personalized interactions, and 76 percent will switch if they do not like the experience. While working for a large food-delivery platform, we stopped treating loyalty as one program for everyone and instead matched the mechanic to the segment and its economics, using order frequency, discount sensitivity, and contribution economics to choose different interventions for heavy users, occasional users, and the discount-driven crowd rather than applying a single offer across the base. The point was to anticipate which customers would respond to which move before spending on it, which is a different discipline from reacting to whoever happened to churn last month. Why companies stall at the bottom If the value is higher up, why do so many companies stay at Layer 1? The answer is rarely the technology, and the data supports that. In its September 2025 study The Widening AI Value Gap: Build for the Future 2025 , based on more than 1,250 firms worldwide, Boston Consulting Group found that about 5 percent of companies are pulling substantial value from AI, roughly 35 percent are starting to, and the remaining 60 percent show minimal gains. The leaders are not winning on tooling, which is largely shared. They are winning on intent and operating discipline. BCG reports that the leading group shows about 1.7 times the revenue growth of the laggards. McKinsey adds a telling detail: 80 percent of companies set efficiency as an AI objective, but the ones capturing the most value add growth and innovation as objectives, and most of them redesign their workflows to do it. In my experience, that is the core difference. Layer 1 asks the technology to make an existing process cheaper. Layers 2 and 3 ask the organization to change how it works, which is harder, slower, and far more valuable. Companies stall at the bottom not because the higher layers are out of reach, but because climbing requires reshaping workflows and accepting that the goal is growth, not just savings. The payoff is starting to show up in the data The pattern is not limited to consulting research. In the November 2025 analysis The State of Generative AI Adoption in 2025 , economists Alexander Bick, Adam Blandin, and David Deming at the Federal Reserve Bank of St. Louis found a positive correlation between how much time an industry saves with generative AI and how fast its labor productivity grows. Industries with one percentage point more time savings showed about 2.7 percentage points higher productivity growth relative to their pre-pandemic trend, with an overall correlation of 0.32. The authors are careful to call this suggestive rather than causal, but it is an early, independent signal that the knowledge and customer work where AI concentrates is beginning to move real productivity. \ \ How to move up a layer If you lead customer strategy, the practical step is to audit where your AI effort actually sits. In most organizations I have seen, the honest answer is that nearly all of it is Layer 1, dressed up in the language of transformation. From there, three moves help. Pick one high-value customer decision and ask what insight you would need to make it better, then build the Layer 2 capability to produce that insight from data you already hold. Resist the urge to spread AI thinly across many small tasks, because a portfolio of Layer 1 wins does not add up to advantage. And measure the work in customer and revenue outcomes, not hours saved, because the metric you choose quietly decides which layer you optimize for. None of this replaces the discipline of knowing your customer. It raises the return on it. The tools are now available to everyone, which means they are no longer the differentiator. The companies that win will be the ones that aim those tools at the layers that matter, and rebuild the way they work around that answer. \ Sources McKinsey, The Economic Potential of Generative AI: The Next Productivity Frontier (June 2023) McKinsey, The State of AI in 2025: Agents, Innovation, and Transformation (November 2025) McKinsey, Enhancing Customer Experience in the Digital Age BCG, The Widening AI Value Gap: Build for the Future 2025 (September 2025) Federal Reserve Bank of St. Louis, The State of Generative AI Adoption in 2025 (Bick, Blandin, Deming, November 2025) \