ChatGPT Sparked a Business Model Revolution

The Day the Business Model Broke

On November 30, 2022, OpenAI released a chatbot that could write poetry, debug code, and explain quantum mechanics to a fifth grader. Within five days, a million people had tried it. Within two months, the conversation about business had shifted permanently. Not because ChatGPT was a better search engine or a smarter assistant. Because it made something invisible suddenly visible: the business model itself.

For decades, companies treated their business model like a fish treats water. It was the environment, not the thing you examined. You sold products. You charged customers. You optimized operations. The structure was assumed, not questioned. Then generative AI arrived, and the assumptions cracked open.

Dominik Kanbach, Louisa Heiduk, Georg Blueher, and Maximilian Schreiter, researchers at the Leipzig University of Applied Sciences and the University of St. Gallen, watched this happen in real time. In their 2023 paper "The GenAI is out of the bottle," they analyzed 513 data points from academic publications, company reports, press releases, news articles, interviews, and podcasts to understand how generative AI was reshaping business models (Kanbach et al., 2023). What they found was not a list of new features or efficiency gains. It was a structural shift in how companies create, deliver, and capture value.

The revolution was not in the technology. It was in the permission the technology gave companies to rethink everything.

Why ChatGPT Broke the Mold That Other AI Could Not

Artificial intelligence has been inside businesses for years. Recommendation algorithms. Fraud detection systems. Supply chain optimizers. These tools were powerful, but they were also invisible. They sat in the back end, processing data, making predictions, never talking to the customer directly.

Generative AI changed that. ChatGPT, Jasper, and DALL-E put an interface on intelligence. The user typed a sentence and got back something that felt like a conversation with a knowledgeable human. The barrier between machine capability and human interaction collapsed.

Kanbach and his colleagues argue that this shift matters for a specific reason. Previous AI tools improved existing processes. Generative AI creates new processes. It does not just optimize a customer service script. It writes the script from scratch. It does not just analyze a market trend. It generates a marketing campaign based on that trend. The technology moves from being a tool to being a co-creator.

The authors found that this distinction has direct consequences for business model innovation. Companies can no longer treat AI as a cost-saving measure. They must treat it as a value-creating partner. That changes the logic of how a business operates.

The Six Propositions That Rewrite the Rules

Kanbach et al. (2023) developed six propositions that describe how generative AI impacts business model innovation. Each one challenges a conventional assumption about how companies should operate.

Proposition One: Value Creation Becomes Conversational

The traditional model of value creation is linear. A company develops a product. It sells the product. The customer uses the product. Generative AI inverts this. The customer interacts with the AI, and the value emerges from that interaction.

Consider a software company that builds a code generation tool. The customer types a request. The AI generates code. The customer refines the request. The AI improves the output. Value is not embedded in a finished product. It is created in the back and forth between human and machine.

This has implications for pricing, for customer relationships, and for the very definition of what a company sells. You are not selling a tool. You are selling a process of co-creation.

Proposition Two: Cost Structures Shift from Fixed to Variable

Software companies have long loved the subscription model. Predictable revenue. High margins. Low marginal cost. Generative AI disrupts this calculus.

Every time a user makes a request to a large language model, it costs compute power. Electricity. Hardware. The marginal cost of serving a customer is no longer near zero. It is real and it is variable.

Kanbach et al. (2023) found that this forces companies to rethink their pricing models. Some will charge per query. Others will offer tiered subscriptions based on usage limits. The old assumption that software costs nothing to reproduce no longer holds. Companies must build their financial models around a new reality: every interaction has a cost.

Proposition Three: Distribution Channels Get Compressed

Before generative AI, distributing expertise required layers. You needed experts to create content. You needed editors, designers, and marketers to package it. You needed salespeople to sell it. You needed support teams to answer questions.

Generative AI compresses this chain. A single model can answer customer questions, generate marketing copy, write documentation, and troubleshoot problems. The channel between the company and the customer becomes shorter.

The authors found that this creates both opportunity and risk. Companies can reach customers faster and more directly. But they also lose the human touch points that built trust and loyalty. The distribution channel is efficient. It is also colder.

Proposition Four: Customer Relationships Become Continuous

Traditional customer relationships are episodic. A customer buys a product. They use it. They maybe contact support. They buy again. The relationship has gaps.

Generative AI fills those gaps. A customer can interact with an AI assistant at any time. They can ask questions, request customizations, get advice. The relationship becomes continuous.

Kanbach et al. (2023) argue that this changes the nature of customer loyalty. Loyalty was once about brand preference or switching costs. Now it is about the quality of the ongoing interaction. If the AI is helpful, the customer stays. If the AI is frustrating, the customer leaves. The relationship is no longer based on the product. It is based on the conversation.

Proposition Five: Revenue Models Move from Products to Outcomes

The most radical shift Kanbach et al. (2023) identified is in how companies capture value. Traditional revenue models charge for access. You buy the software. You pay for the subscription. You own the tool.

Generative AI enables a different model: charging for outcomes. Instead of selling a code generation tool, a company charges per line of code produced. Instead of selling a design tool, a company charges per image generated. The customer pays for what the AI produces, not for the ability to use it.

This aligns incentives. The company wants the AI to be effective, because effectiveness drives usage. The customer wants the AI to be effective, because effectiveness drives value. Both parties are pulling in the same direction.

Proposition Six: Competitive Advantage Shifts from Data to Interaction

For the past decade, the conventional wisdom was that data was the new oil. Companies that collected the most data would win. Generative AI challenges this assumption.

Kanbach et al. (2023) found that the competitive advantage in generative AI comes not from the training data but from the interaction data. The models are trained on public data. Everyone has access to roughly the same foundation. What differentiates one company from another is the feedback loop of user interactions.

Every time a user corrects an AI output, rates a response, or refines a prompt, the company learns something. That interaction data is proprietary. It cannot be scraped from the internet. It is earned through use.

This shifts the strategic focus. Companies no longer compete on who has the biggest dataset. They compete on who has the most engaged user base.

Three Industries That Already Changed

Kanbach et al. (2023) examined three industries in detail to show how these propositions play out in practice.

Software Engineering: The End of Code as a Product

Software engineering was the first industry to feel the impact. GitHub Copilot, launched in 2021, showed that AI could generate functional code. ChatGPT accelerated this. Developers now use AI to write boilerplate, debug errors, and generate test cases.

The business model shift is subtle but profound. Software companies used to sell code. Now they sell the ability to produce code faster. The value is not in the code itself. It is in the speed and quality of the development process.

Kanbach et al. (2023) found that this changes the competitive dynamics. Small teams can now produce what used to require large engineering departments. The barrier to entry drops. The cost of experimentation falls. The result is more innovation, but also more competition.

Healthcare: Diagnosis Becomes a Service

Healthcare has been slow to adopt AI, for good reason. Mistakes can be fatal. Regulation is stringent. But generative AI is finding its way into clinical workflows.

Kanbach et al. (2023) highlight examples where AI assists in generating patient summaries, interpreting medical images, and suggesting treatment options. The business model shift is from selling diagnostic tools to selling diagnostic insights. A hospital does not buy a radiology AI. It buys the ability to read more scans, faster, with fewer errors.

The authors note that this raises difficult questions. Who is liable when an AI makes a mistake? How do you price a service that improves over time? The business model innovation is happening, but the regulatory framework has not caught up.

Financial Services: Advice Becomes Accessible

Financial advice has traditionally been a luxury good. Only wealthy individuals could afford a human financial advisor. Generative AI changes this. Chatbots can now answer questions about investments, taxes, and retirement planning.

Kanbach et al. (2023) found that this democratizes access but also changes the revenue model. Banks used to charge for advice by the hour or by the asset. Now they are experimenting with subscription models, per query fees, and bundled services.

The risk is that AI advice is not as good as human advice for complex situations. The opportunity is that AI advice is better than no advice at all. The business model must balance accessibility with accuracy.

What the Research Does Not Prove

The Kanbach et al. (2023) paper is a scoping review. It identifies patterns and proposes hypotheses. It does not test them rigorously. The authors are clear about this. Their goal is to map the territory, not to prove causality.

This means several things remain open questions. First, it is not clear which business models will survive. The paper describes possibilities, not certainties. Some companies will try per query pricing and fail. Others will succeed with subscription models. The research cannot predict which.

Second, the paper does not address the ethical and social implications in depth. Generative AI raises questions about job displacement, bias, and misinformation. The authors acknowledge these issues but do not resolve them. A business model revolution does not guarantee a good one.

Third, the research is based on early data. The paper was published in 2023, less than a year after ChatGPT launched. The landscape has already changed. New models, new regulations, and new competitors have emerged. The propositions remain relevant, but the specific examples may age quickly.

What This Actually Means

The Kanbach et al. (2023) paper is not a prediction. It is a map. Here is what it tells us to watch for.

• ▸If your company charges for access to a tool, prepare for the model to shift to charging for outcomes. Customers will increasingly pay for what the AI produces, not for the ability to use it. Start experimenting with usage based pricing now.

• ▸If your competitive advantage is based on data volume, rethink it. The advantage is shifting to interaction data. Build products that encourage users to engage, correct, and refine. That engagement is your moat.

• ▸If your customer relationship is episodic, make it continuous. Generative AI enables 24/7 interaction. The companies that build the best conversational experiences will win loyalty. The ones that treat AI as a cost center will lose.

• ▸If your cost structure assumes near zero marginal cost, update your financial model. Every AI interaction has a real cost. Build pricing that accounts for this. Do not subsidize heavy users at the expense of profitability.

• ▸If your business model has not changed in the last two years, it is already obsolete. The revolution is not coming. It is here. The companies that treat generative AI as a reason to rethink their entire model will survive. The ones that treat it as a feature upgrade will not.

Kanbach et al. (2023) showed that ChatGPT did not just introduce a new product. It introduced a new logic. The business model is no longer the water you swim in. It is the thing you must examine, question, and rebuild. The bottle is open. There is no putting the genie back.