How AI is Disrupting Your Business Model

Walk into any large enterprise today and ask about their AI strategy. You'll hear about co-pilot rollouts, productivity pilots, use-case registries, and training programmes. These things are real. They produce real results. And they're not sufficient.

Laura Stevens calls this the "comfortable illusion": the idea that AI is primarily a tool for making your existing operating model more efficient. The problem isn't that this is wrong. The problem is that it's a category error. Optimisation is an operational question. AI is a structural one. Applying operational logic to a structural shift means you're refining a model that may soon stop making sense entirely.

The deeper issue: a structural shift cannot be addressed with incremental logic. If competitors are rethinking their business models from scratch, then being incrementally more efficient inside your current model is not a strategy. The gap between "we're running pilots" and "we're rebuilding how we compete" is not one you close with a task force.

Markets have always been built around human buyers: impulsive, brand-influenced, limited in how much they can process. Every aspect of how businesses win customers rests on that assumption. AI is dismantling it.

Customers increasingly begin their journeys not on brand websites but in ChatGPT, Perplexity, or AI-powered search. They ask "what's the best CRM for a mid-sized company?" or "which laptop is best under a thousand euros?" and the algorithm summarises, compares, filters, and recommends. Humans still make the final call, for now, but the consideration set and how options are framed is shaped entirely by a system that didn't exist five years ago. If you're not in that set, you're not in the race.

The disruption deepens as AI moves from shaping decisions to executing them. Agentic commerce is already live in the US, with AI assistants completing purchases inside chat interfaces. The trajectory: autonomous agents that manage the full journey from discovery to checkout, followed by agent-to-agent commerce where your AI negotiates with brand AIs, no human involved. At that point, the interface becomes the decision maker. Brands aren't competing for attention. They're competing to be selected by a system.

Competitive moats have traditionally relied on scale: large teams, capital, deep expertise, long development cycles. AI is compressing all of that. What once required 50 engineers and months of work can now be built by a small team in weeks.

This is already happening. Stevens points to B44: built by a developer as a side project using AI-assisted "vibe coding." In six months, with no outside funding, the company reached 250,000 users, hit $3.5 million in annual recurring revenue, and sold for $80 million. The team? Fewer than ten people. Basis AI tells a similar story: a $1 billion valuation with a fraction of the headcount a traditional services business would need, built on autonomous agents handling end-to-end accounting.

The implication isn't just that new entrants are more dangerous. It's that the cost of building alternatives to your product is falling continuously. More competitors enter, and customers become less willing to pay premiums when they know alternatives are cheap to build. The question is whether your strategy accounts for that.

Most business models rest on a simple premise: revenue scales with effort. You charge by the hour, by the seat, by the deliverable. AI doesn't just make that effort cheaper to produce. It breaks the logic that made effort valuable in the first place.

Stevens puts it simply: if AI can generate a market analysis in minutes, why would a client pay five thousand euros for one? The pressure shows up everywhere. In professional services, it hits the junior leverage model hardest: AI automates exactly the repetitive cognitive tasks that junior analysts were hired to do, compressing billable hours and shrinking the traditional pyramid. In SaaS, it hits seat-based pricing: if AI lets three people do the work of twenty, you're selling eighteen fewer seats.

The response is not to defend the old model. It's to understand what customers will actually pay for when effort becomes cheap. The answer, consistently, is outcomes. Rolls-Royce doesn't sell jet engines. It sells hours of thrust. That logic is spreading across services, manufacturing, and beyond.

The failure mode Stevens sees most often: companies respond to structural disruption with operational tools. They build use-case lists. They deploy co-pilots. They run pilots. None of this is wrong, but it's insufficient when the disruption is structural. The question is not which processes to automate. It's whether the model those processes serve still holds.

BOI's framework organises the response into three tracks. Strategic reinvention: where do you play, how do you win, what makes you hard to replace? Workflow redesign: which workflows are high-volume, labour-intensive, and repeatable enough that an AI-first redesign changes your cost base structurally? Operating model redesign: as delivery changes, so must structure, governance, roles, and culture.

Defensible positions, Stevens argues, cluster around three things: proprietary data that others cannot replicate, deep integration into customer workflows that raises switching costs, and differentiated experiences that make comparability harder. A Dutch entrepreneur turned down a ~$500 million offer from OpenAI for his gaming behaviour dataset. Data is a moat. Integration is a moat. Generic output is not.