I drew the onion on a napkin at an advisory board meeting at a university about a year ago, and it has shown up in every conversation I have had since.
The provost had just finished a presentation about the school's AI adoption progress. The slides showed numbers. Percentage of faculty using AI tools. Number of departments with AI policies. Hours of professional development delivered. All of the numbers were going up. The room was nodding.
I raised my hand and asked a question that stopped the nod.
"How many of those uses changed the actual work, and how many just made the same work faster?"
Nobody had the number. Because nobody was tracking that. And because the answer, once you look at it honestly, is almost always the same: most of it is surface. Almost none of it is core. And the distance between those two things is where the entire competitive story of the next decade lives.
I grabbed a napkin and drew four circles, one inside the other. That napkin has been doing more work than any slide I have ever built.
The Four Layers
The Skin (Surface). AI helps an individual do an existing task faster. Email drafting. Slide formatting. Transcription summaries. Meeting notes. This is where roughly 70% of "AI adoption" lives in 2026. No process changes. No org changes. Just the same work, slightly faster, done by the same people in the same roles.
This layer is not nothing. I want to be clear about that. A team that learns to use AI for email drafting in 2024 builds the literacy that lets them attempt something harder in 2026. The surface is the on-ramp. The mistake is parking there and calling it a strategy.
The Outer Layers (Augmentation). Teams build AI into recurring workflows. Sales call summaries auto-routed to the CRM. Client onboarding partially automated. Marketing assets generated against a brand kit instead of briefed from scratch. Some process change. No structural change. The org chart looks the same.
The Inner Layers (Redesign). The job itself changes shape. The customer service team is half its former size and twice as fast, because the work was rebuilt around an agent that handles the first 60% of inquiries. The marketing team ships in days what used to take weeks, because the deliverable was redefined. New roles appear. Old roles merge. The org chart starts to look different.
The Core (Infrastructure). AI is part of the operating system of the organization. It is not a tool someone uses. It is the substrate the business runs on. New roles exist that did not exist two years ago. Old roles have been retired. Customers experience the organization differently. Revenue per employee looks different. The financial profile of the company has changed.
This is where the returns live. Not at the skin. At the core.
The Data on the Gap
Deloitte's 2026 State of AI in the Enterprise report puts the gap in plain numbers. 37% of organizations report using AI only at the surface, with little or no change to underlying business processes. Only 30% are redesigning key processes around AI.
The 30% that got to redesign are pulling away fast. Deloitte found that AI leaders achieved 1.7x revenue growth, 3.6x greater total shareholder return, and 1.6x EBIT margin compared to laggards. The differentiator is not which model they use. It is whether they redesigned the work around it.
Glean's enterprise data adds another angle. "Frontier" workers, people redesigning their job around AI, save six times more time than median AI users. That is not a productivity story. That is a bifurcation story. The people at the skin are getting a little faster. The people at the core are operating in a different category.
The gap between surface adoption and core integration is the new competitive moat, and it is widening, not closing.
An OpenAI enterprise report from 2025 found the same pattern from the customer-facing side: the gap between organizations treating AI as a peripheral tool and those treating it as infrastructure was visible in customer experience metrics, not just internal productivity numbers. The customers can feel the difference. They cannot name it, but they can feel it.
Why Most Organizations Stay at the Surface
If the core is where the returns live, why does almost nobody get there?
Three reasons, and I need to be honest about all of them.
First: getting to the core is hard, expensive, and politically dangerous. The middle managers whose jobs are most threatened by redesign are the ones sitting on the AI committees. Surface adoption is what those committees produce by design, not by accident. Recommending "use ChatGPT for email" is safe. Recommending "rebuild the customer service function around an agent" threatens someone's headcount, budget, and identity. The surface is the path of least organizational resistance, and most organizations follow the path of least resistance until the market forces them off it.
Second: surface adoption is genuinely useful, and it builds the literacy that makes deeper adoption possible. Skipping the surface and demanding the core is like asking someone to run a marathon on their first day at the gym. The muscle has to be built. The mistake is not starting at the surface. The mistake is measuring the surface and believing the measurement means you are done.
Third: the Deloitte 1.7x number is correlation, not causation, and I am going to say that out loud because most people quoting it will not. Companies that successfully redesign around AI are also companies with strong leadership, capital, and talent. They might have outperformed without AI. The honest claim is that AI amplified their existing advantage. It did not create it from nothing. The onion model is a diagnostic, not a guarantee.
Same Story In Home Services
Here is what the four layers look like in a single industry, so you can see the whole arc at once.
Skin. Every contractor with ChatGPT is generating polished proposals, faster review responses, AI-drafted job descriptions. Useful. Ubiquitous. No moat. Your competitor has the same tool and figured it out the same week you did.
Outer layers. The shops that have integrated AI into their field service software for dispatch optimization, predictive scheduling, and AI-assisted customer service call handling. Real productivity gains. The business is still recognizable as the same business, with the same roles, just moving faster.
Inner layers. A small but growing number of home services operators running agentic customer service fully. No human picks up the first call. The agent qualifies, books, and dispatches. Their cost-per-booked-job is 30-50% lower than the shop next door. Their org chart already looks different. Roles that existed two years ago do not exist anymore. New roles (agent supervisor, exception handler) have appeared.
Core. The regional player who has rebuilt the entire customer journey around AI, from first ad click to post-service follow-up to upsell to review request, as one continuous, agent-orchestrated flow. These operators are starting to acquire the surface-only operators below them. The consolidation is already happening, and it is happening because the core operators have a fundamentally different cost structure and customer experience than the ones still parked at the skin.
The contractor whose AI use is "ChatGPT for proposals" is competing against contractors whose AI use is "the entire business." Same trade. Different layer. Different outcome.
The lesson translates one-for-one to higher education, to marketing agencies, to consulting firms, to any organization that has been measuring AI adoption by counting how many people use the tool instead of asking what the tool changed.
The Diagnostic Question
Here is the question I leave with every audience, every advisory board, every client.
Which layer of the onion is your organization actually operating at, and what would have to be true to move to the next one?
Not which layer your strategy deck says. Not which layer your AI committee reported. The actual one. The one you would admit if nobody was keeping score.
Most organizations are at the skin and measuring the skin and reporting the skin and believing the measurement means they are making progress. They are making progress the way a person running on a treadmill is making progress. The effort is real. The distance traveled is zero.
The organizations that win the next five years are the ones that looked at the onion, were honest about which layer they were at, and started doing the harder, slower, more politically dangerous work of moving inward. Not because a consultant told them to. Because the market is going to stop waiting for the ones who did not.
The napkin I drew that day at the advisory board is pinned to my office wall. Every time I look at it, I ask myself the same question I asked the provost. I do not always like the answer.
If you want to run the diagnostic on your own organization and figure out what "one layer deeper" looks like for your specific situation, that is the conversation I have with people at bensaibrain.com. Come say hi.
Sources
Created with ❤️ by humans + AI assistance 🤖