When AI takes over the work – what will customers still pay for tomorrow?

What happens to a business model when the service customers pay for today can largely be delivered automatically tomorrow?

For companies, this question could be more far-reaching than the question of which AI model they should use.

The first phase of generative AI was marked by experimentation: writing texts, analysing information, creating presentations, developing software. Companies wanted to understand how AI could make employees more productive.

A second phase is now emerging. AI systems are increasingly taking over not just individual activities but entire work steps. They research, analyse, program and coordinate tasks across different applications.

This changes not only a company’s cost structure. It also raises the question of what a company will actually sell in future.

The productivity paradox of AI

The figures are based on respondents’ self-reports. They are not independent evidence of a causal effect of AI on company profits.

If AI makes so many people more productive, why are more companies not benefiting from it?

One possible explanation: we optimise existing activities without questioning the underlying business model.

A consultancy can produce analyses faster with AI. A law firm can review documents more efficiently. A software company can develop new features in less time. But higher productivity does not automatically mean higher profits.

If customers continue to pay for hours, licences or individual services, a tension arises: what happens to the price of a service when producing it suddenly takes only a fraction of the previous time?

The real shift: from access to outcome

This development is particularly visible in the software industry. For decades, the business model of many software providers was relatively simple: companies paid for access to an application – often per user per month.

But what happens when an AI agent can operate the same application and take over the work of several users? The number of people logging into a piece of software may then lose significance. What matters is which tasks actually get done.

For providers, this raises a pricing question: should the customer pay for access, for a work step or for a verifiable outcome? This can only be decided for the specific offering. If you sell an outcome, you must be able to explain what counts as success and who is responsible for it.

One conceivable offering would be a verified monthly report instead of a software licence. Another would be the ongoing maintenance of a dataset at a fixed price. Such offerings also change the provider’s tasks: it must assure quality and correct errors, rather than merely providing a tool.

A software agent sells not just a seat for developers, but an implemented change. A customer service agent could in future be paid according to successfully resolved enquiries.

For some offerings, it can make sense to sell a completed task rather than access to a tool.

This does not mean that traditional subscriptions will disappear. But their economic logic is increasingly being called into question.

When customers can suddenly become providers themselves

A second development could be even more profound. In McKinsey’s 2026 survey, 32 per cent of respondents report that their organisation has already decided not to buy at least one software product or feature because it can be developed in-house with agentic coding tools. [1]

This changes a fundamental decision for many companies: buy or build? Until now, building your own software was often expensive, time-consuming and associated with considerable technical requirements.

AI lowers these hurdles. Security, maintenance, integration and scaling naturally remain demanding. Nevertheless, developing their own solutions is becoming economically attractive for more companies.

For established providers, this creates a new challenge. They no longer compete exclusively with other companies, but increasingly with their customers’ ability to produce parts of the service offered themselves.

This raises a strategic question:

How does a company defend its value creation when its customers can automate part of it themselves?

The answer is likely to lie less and less in technology alone. Competitive advantages arise where companies have unique data, deep industry knowledge, trusted customer relationships, integrated processes or capabilities that are hard to replicate. AI makes some services more interchangeable. At the same time, it can increase the value of what cannot easily be copied.

Real innovation begins with the business model

In many companies, the AI discussion still begins with an efficiency question:

Which of our current processes can we automate?

That is a sensible starting point. Strategically, however, it is not enough. The more interesting question is:

What offering would we develop today if we could found our company anew with the possibilities of AI?

Instead of individual projects, a consultancy could offer continuous, AI-supported decision support. An industrial company could sell not only machines, but guaranteed production output or availability. A software provider could evolve from tool supplier to operator of complete business processes.

Which of these options makes sense depends on customer value, costs and risks. New technology alone does not create a new business model. Only when the value proposition, the way value is created and the revenue logic change does genuine business model innovation emerge.

Three questions for the next strategy discussion

If you want to understand the impact of AI on your own business model, you can start with a simple thought experiment.

1. What if the cost of our most important service fell by 90 per cent?

Would we offer the same service more cheaply? Would we reach new customer groups? Or would our current pricing model lose its foundation?

2. What if our customers could produce 80 per cent of our current service themselves?

Which part of our offering would then still be indispensable? Where is the value created that customers cannot reproduce themselves?

3. What if we were no longer paid for our activity, but exclusively for the outcome?

How would we have to change our processes, capabilities, risks and customer relationships? The percentages are deliberately chosen scenarios, not forecasts. That is precisely why they are well suited to making existing assumptions visible.

The key insight

AI will change not only how companies work. It will change which services remain scarce, how value is distributed between companies and customers, and which business models are attractive in the long term.

If you regard AI purely as an efficiency technology, you may be optimising a business model whose economic foundations are already shifting.

Companies should examine what value their customers will still pay for even when producing a service becomes considerably cheaper.

This marks the start of a concrete task: reviewing your own offering and its pricing model.