For decades, selling an intellectual service meant selling time. The more hours a study, an assessment or a report required, the higher the invoice.
Artificial intelligence is shaking up this model: in a few minutes, it carries out part of what once took several days. If time no longer measures the value of a service, what is the client really paying for today?
To understand this, let us look at two concrete situations.
Two situations, one dilemma
Situation 1 (the augmented provider): a consultant writes risk assessment reports invoiced at €6,000 (6 days at €1,000). He now uses AI (subscription: €300 a year) which, in time, should save him up to half of his working time.
Situation 2 (the augmented client): a self-employed professional prepares his own accounts with the help of AI and passes them to his accountant for validation only. Should the accountant, who used to charge €1,000 a year to do everything, keep his price?
In both cases, AI carries out part of the work. But the value of the service does not follow the same path.
Situation 1: the consultant and the execution-time trap
When the consultant uses AI for his report, the work is redistributed:
- The consultant's role: gathering the material (interviews, fieldwork), configuring the tool (analytical framework, methodology), checking the results, settling questions of judgement and taking responsibility for every line.
- The AI's role: anonymising and structuring the data, drafting a first version, suggesting theoretical references and checking consistency with the sources.
At first, the time saving is small: the time gained on drafting is absorbed by configuration and checking. But once the tool has been mastered, the gain becomes real: studies already measure time savings of 25 to 40% on this type of task. Tomorrow, this report could take just 3 days.
The accounting calculation (3 days × €1,000) points downwards. But value follows a different logic: today's report can give the client more than yesterday's (tailored statistical analyses, year-on-year monitoring indicators), in half the time.
The 5 pillars of value in the age of AI
To set a fair price for an AI-assisted intellectual service, five key factors need to be analysed.
1. Real time and the “J-curve”
AI reduces execution but demands a new investment: learning, configuring and checking. In economics, this is the productivity J-curve: at first, performance drops while people learn, then it improves. The real cost of AI is not the €300 subscription, but this learning time. However, this argument only justifies keeping the price unchanged during the transition phase.
2. Human judgement versus automation bias
AI makes “prediction” and text generation very cheap (Agrawal, Gans and Goldfarb). What becomes scarce and valuable is human judgement: deciding which question to ask and what to do with the result.
This is the Baumol effect: what remains purely human gains value as everything else becomes automated.
This need for judgement also explains why experienced workers are currently protected in the job market in the face of AI, while the hiring of juniors is falling (Stanford study). Judgement is what clients pay for, and judgement comes with experience.
3. Quality and the gold-plating trap
With AI, adding options or enriching a report costs almost nothing. But beware of the Kano model: what pleasantly surprises the client today becomes their baseline expectation tomorrow.
In project management, giving more than was asked without making it visible is called gold plating. An invisible extra creates neither gratitude nor loyalty (Marcel Mauss). Any addition produced by AI should be explicitly named or offered as an option.
4. Competition and the Jevons paradox
An efficiency gain only brings prices down if competition forces it to. If your clients are buying specific expertise or a certified signature, you keep the productivity gain. Moreover, according to the Jevons paradox, when a service becomes more efficient and more accessible, demand may rise: organisations may commission more analyses.
5. Assurance and signalling theory
When buying a service, the client pays for reassurance (Michael Spence). The consultant is not selling words on paper: he guarantees the soundness of the diagnosis and puts his professional liability on the line.
Breaking down the €6,000: from time to value
Billing solely for time spent is becoming risky for the augmented expert. Where possible, it is better to move towards value-based pricing, grounded in a breakdown of the deliverable:
| Service component | What it covers | How it changes with AI |
|---|---|---|
| Execution | Research, formatting, raw drafting | Marginal value close to zero |
| Framing and direction | Defining the right problem, setting the angles | Rising strategic value |
| Critical judgement | Checking, cross-referencing, eliminating errors | Decisive value |
| Assurance and responsibility | Staking one's reputation, justifying every decision | The core of the value invoiced |
Practical recommendations for providers:
- Favour fixed pricing per deliverable, where the client and the contractual framework allow, rather than billing by the day.
- Separate execution from assurance: make your quality-control method explicit in commercial proposals.
- Charge for the extra: if AI lets you add a statistical monitoring module, present it as a valued option.
Situation 2: the augmented client and the changing role of the accountant
In the second case, the self-employed professional does the bookkeeping with AI, then asks the accountant to “simply validate” it.
If the accountant keeps charging €1,000 when his work is now limited to validation, the client will feel short-changed. Yet what is this professional actually paying for?
- Tax compliance and credibility with the tax authorities.
- Risk cover through the accountant's professional liability insurance.
- Peace of mind in the event of a tax audit.
The accountant can no longer sell data entry. He must transform his offer: lower the price of purely formal checks, but offer proactive strategic advice in return (tax optimisation, cash-flow management).
Value to be reinvented continuously
AI does not destroy the value of intellectual services: it shifts their centre of gravity. Tomorrow, value will lie in the following equation:
Value of the service = human judgement × mastery of AI
AI does not create competence: it amplifies it. On its own, it produces only an illusion of competence. Human judgement on its own becomes too slow and too costly. It is the combination of the two that creates real value in the market.
As the economist Carlota Perez has shown, every technological revolution goes through a period of friction before finding its new bearings. We are in it now.
AI does not replace expertise. AI changes what clients pay for. Tomorrow, they will not pay for the time spent, but for the judgement they can rely on.
Sources and references
- Agrawal, A., Gans, J. and Goldfarb, A. (2018). Prediction Machines. Harvard Business Review Press.
- Baumol, W. (1967). Macroeconomics of Unbalanced Growth. American Economic Review.
- Brynjolfsson, E., Rock, D. and Syverson, C. (2021). The Productivity J-Curve. American Economic Journal: Macroeconomics.
- Brynjolfsson, E., Chandar, B. and Chen, R. (2026). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence. Stanford Digital Economy Lab.
- Dell'Acqua, F. et al. (2023). Navigating the Jagged Technological Frontier. Harvard Business School, Working Paper 24-013.
- Kahneman, D. and Tversky, A. (1979). Prospect Theory. Econometrica.
- Kano, N. et al. (1984). Attractive Quality and Must-Be Quality. Journal of the Japanese Society for Quality Control.
- Mauss, M. (1925). The Gift (Essai sur le don).
- Noy, S. and Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science.
- Perez, C. (2002). Technological Revolutions and Financial Capital. Edward Elgar.
- Smith, A. (1776). The Wealth of Nations.
- Spence, M. (1973). Job Market Signaling. Quarterly Journal of Economics.
The figures in both situations are illustrative.
This article was written with the help of AI tools (Claude and Gemini), used for research, source checking and structuring. The ideas, experience and analysis are the author's own.