Your Business Model Is Obsolete: How AI Is Disrupting Value Creation

Developing an AI strategy or building AI products requires a vision for what’s next. Otherwise, you end up building an AI version of a digital product or business. Variations of digital or cloud themes don’t live up to expectations.

The AWS console’s AI assistant is a variation on a digital search theme. Search strings are more flexible with it, and there’s incremental value in that feature. However, generative search hasn’t made configuring instances, discovering features, and managing cloud usage easier. Generative search was stapled over the digital workflow and fits the Human-Machine Tooling paradigm (an early maturity phase of product design).

For AI to live up to the hype, we must design products and features that support Human-Machine Teaming and Collaboration. The first mindset shift required for AI strategy and products is transitioning from thinking about technology supporting tasks and delivering work products to technology enabling outcomes.

There’s value in some Human-Machine Tooling-style AI products and features, but the highest value comes when opportunities are guided by a vision that leverages AI’s disruptions. Disruption is a fuzzy term that requires a clear definition to be actionable. Most AI strategies are aspirational, and defining terms like disruption makes them actionable.

Fuzziness leads to interpretation. Multiple interpretations lead to different business units taking AI in very different directions. Enterprise-wide alignment is critical, so actionable AI strategies are the only ones that succeed.

Defining AI’s Disruptive Nature

Disruptions happen when a long-standing assumption or rule is no longer valid. Business and operating models, as well as products built on those assumptions, become vulnerable. Update, replace, or remove the old assumption/rule, and the new opportunities become easier to see. That’s how startup founders and enterprise visionaries discover the opportunities that the rest of the market can’t.

NVIDIA is a great example. Data centers were built on the assumption that CPUs are the ideal architecture for enterprise workloads. AI disrupts that assumption since GPUs are a better architecture for model training and inference workloads. Intel and AMD continued delivering products built on an assumption that no longer held. NVIDIA pulled ahead by positioning its products based on a more accurate view of the marketplace.

Knowing the assumption that NVIDIA’s dominance is built on is the recipe for disrupting it and taking NVIDIA’s market share. Dominant strategy requires finding a new architecture that’s significantly better suited for AI workloads or changing the nature of AI workloads to fit a different architecture.

I teach this framework in my courses (Disruption is part of Decision Dominance and Innovation Management) because the most significant AI opportunities are the least obvious. Frameworks provide first-principles definitions of fuzzy concepts and make managing a business’s fuzziness more efficient.

Frameworks have multiple purposes in modern business strategy. They are heuristics for decision making, and in the AI paradigm, heuristics must be built for people and technology-driven agents.

In a modern business, strategy must be actionable because value is delivered by agency, not just ideas. Think of it in terms of potential energy (strategy) and kinetic energy (agency). Without potential energy, there’s nothing driving agency. Without action, a strategy’s potential energy is never realized.

NVIDIA’s GPUs had a smaller amount of kinetic energy in gaming. Pivoting strategically to support AI workloads increased its kinetic energy by orders of magnitude.

The Physics Of AI Strategy

It’s the same for AI agents. ChatGPT with no agency is just potential energy. Cursor’s parent company, Anysphere, is one of the fastest-growing startups in history because Cursor has agency in the coding task domain. As its ability to act grows, so will its kinetic energy and value.

In the past, people were the primary source of a business’s kinetic energy and agency, so we built businesses around that tenet. Technology is a new source, which is why Satya Nadella called AI a new input for economic growth. AI enables the business’s technology platforms to take on more agency. We won’t see AI achieve its full potential as a disruptor and growth driver until we build frameworks that enable its agency.

That’s the separation between a legacy and modern business. Legacy firms have business and operating models. Modern firms also have a technology model.

People who have read my book or taken my classes are helping their businesses develop a technology model by implementing frameworks that optimize the transformation from legacy to modern business. The frameworks are heuristics for decision-making. In the legacy business paradigm, they would only target people, giving them greater agency. In the modern business paradigm, frameworks must support people and technical agents.

Firms have complex geometries and run on mathematical models. When I designed my frameworks, I built them to increase agency across the business with two audiences in mind: people and technology.

If you ask many LLMs with reasoning capabilities to turn my frameworks into a function, they will do a pretty good job. Here’s ChatGPT o3’s interpretation of Continuous Transformation:

Is it a perfect interpretation? No, and it doesn’t have to be. It’s a hypothesis, and that makes it a testable, improvable starting point. All you need are frameworks for managing those processes, which is why I created The Flywheel. Experimentation and continuous improvement cycles in modern businesses are managed by technology and people.

This is how we implement Human-Machine Teaming and Collaboration paradigms. We must transform the business to enable people to use technology like a teammate. As the technology agent becomes more reliable, it transitions into the collaborator paradigm.

Strategy is the business’s core operating system. Agency is how strategy is implemented and executed. People and technology agents work together to make the business more efficient. However, most businesses aren’t built this way, so they can’t leverage technology efficiently. The AI Last Mile Problem must be addressed at the first mile.

The AI Last Mile Problem – Reducing The Cost Of Agency

Cursor’s value jumped because it increased the technical platform’s agency to deliver code and reduced the cost of agency in that domain. Before Cursor, people were limited to the Human-Machine Tooling paradigm to reduce the cost of coding. IDEs and the developer’s ecosystem of tools helped optimize some parts of the workflow, but none of those tools were good enough to have agency.

Cursor has limited agency in the Human-Machine Teaming paradigm, where an engineer describes what code must be written and validates the model’s output. Vibe-coding takes that a step further into the Human-Machine Collaboration paradigm. However, the model isn’t reliable enough to take on that level of agency yet.

Still, you can see the slow march towards greater agency delivering greater value by reducing the cost of actions and work products. Once reliability crosses the threshold into Collaboration, the cost of agency plummets because the model is capable of reducing the cost of outcomes. The workflow vs. outcome mindset shift is where the greatest value comes from.

We’re limited by our view of AI, just reducing the cost of agency in terms of our current tasks and work products. The highest value use cases are built on delivering entirely new types of agency, not just agency that wasn’t possible to automate in the past.

The AI Last Mile Problem – Enabling New Types Of Agency

The assumption that Cursor is built on is that code is the most efficient way to deliver technical artifacts. Programming languages are translation layers between people and machines. All code is based on logic, so well-written code can be expressed mathematically. It’s more efficient to build the mathematical model first, optimize it, and then implement it in your favorite programming language.

Disrupting Cursor requires finding a more efficient translation layer between people and machines. The obvious answer is math, but functions behind logic require a strong foundational mathematical understanding. Most people, even programmers, don’t have all the knowledge required. Code creates a simpler way to represent logic that abstracts away the math.

Does AI provide a more efficient method that removes coding from the workflow altogether? Above, an agent defined the heuristic or framework for Continuous Transformation in terms that an agent can understand without code. An agent can ground its actions on that mathematical model. It basically follows the function’s instructions, the same way machines follow the instructions embedded in code. The root was natural language, and the intermediate step of coding has been removed altogether.

As the logical or algorithmic complexity of the agency we’re automating rises, the code and mathematical complexity do too. We truly are at war with complexity, and technology is our primary weapon against it. If we can use the technology platform to reduce the business’s complexity, we can make it more efficient. That’s the function behind my Core-Rim framework.

Without strategy, businesses can build exceptional agency, but it won’t be deployed efficiently or in support of high-value opportunities. Without a technology model, agency will remain expensive and inefficient. For businesses to profit from data and AI, they must transform. Technical strategy manages that transformation efficiently.

I am pleased to announce that the next instructor-led cohort of my Data & AI Strategist Certification is open for enrollment. See the course overview and enroll here.