What happens to the price of AI tools once companies can no longer work without it?

← All articles
Artificial intelligence · EN

What happens to the price of AI tools once companies can no longer work without it?

Imagine an AI service reliably taking over work that previously cost a company €100,000 a year HR costs. At €75,000, it could still look attractive, provided the costs of integration, oversight and errors remain manageable. That is a thought experiment. But it raises a question I find more interesting than another prediction about when AI will replace white-collar workers.

What happens to the bargaining power of the companies buying it?

A field experiment across 66 firms and 7,137 knowledge workers found that users of an integrated AI tool spent around two fewer hours on email each week. The researchers did not detect changes in the quantity or composition of their tasks. That finding alone gives us no replacement threshold (Dillon, E. W., Jaffe, S., Immorlica, N. amp; Stanton, C. T. Shifting Work Patterns with Generative AI. American Economic Review: Insights, forthcoming).

Now consider a possible next stage: businesses automate entire workflows, reduce recruitment and gradually lose some of the expertise needed to operate without their AI systems. How easily could they then walk away from a provider or negotiate a price increase? In that scenario, providers could have room to price against the economic value of the work performed. A substantial share of the expected savings might end up on the supplier’s invoice. Competition, alternative models and the ability to switch providers could change that equation.

There is also a governance dimension. For high-risk AI systems, Article 14 of the EU AI Act addresses effective human oversight, including the ability to override outputs and interrupt the system. Article 26(2) requires deployers to assign oversight to people with the necessary competence, training, authority and support. These provisions concern high-risk systems, rather than every workplace AI tool.

If we gradually remove the expertise behind a workflow, who will still be able to challenge the system or take over when necessary? For me, responsible AI adoption should include preserving expertise, maintaining credible alternatives and understanding the cost of becoming dependent.

How much of the productivity dividend will businesses and employees retain, and how much could become revenue for AI providers?

Further reading: Anthropic Economic Index report: Cadences. Figure 2.5 is particularly worth a look: across 26 of 31 output types, AI autonomy was higher in Claude Code than in chat or Cowork. It provides useful context for the delegation question.

Discuss on LinkedIn