Billing by the hour amounts to making a client pay for my slowness. With AI, time spent is no longer a reliable indicator of the value produced. It can even move in the opposite direction: less production time can create more value for the client. The real question is no longer how long I took, but what the client can do, and when, thanks to what I deliver.
The thought experiment that should bother all of us
Take two providers and one client, who needs a market study to decide whether to launch a product. Not a cosmetic study: competition, market sizes, positioning, price scenarios, regulatory risks.
Let us state the assumption clearly, because it is what makes the experiment interesting: both studies reach the same conclusions, with the same level of source checking. Not a watered-down version against a serious one. The same result, checked with the same rigour. If this assumption seems impossible to you, this article does not concern you yet. It becomes a little more relevant every quarter.
The first works the way we worked in 2019: a team, a data collection plan, ordered data, interviews, a synthesis. Six months later, he hands over a solid document. Three hundred hours. At 150 euros an hour, he bills 45,000 euros and nobody is surprised.
The second masters her tools: processing chains already built, verification safeguards defined, knowledge of the sector. She delivers the same study in twenty-four hours. At the same hourly rate, she bills 1,200 euros.
The second produced exactly the same thing, only better, and she earns thirty-seven times less. Something is wrong, and it is not her.
Now ask the question from the client’s side, because it is the only one that really matters. What is the value of having this study in twenty-four hours rather than in six months?
It is enormous, and it has nothing to do with the document itself. The client decides this week. He launches before his competitor. He tests a hypothesis, finds out it is wrong, and tests another the next day instead of waiting for the autumn. He does not lose the market window that was open in the spring and closed in September.
The six months saved are not a discount on an invoice. They are an asset. And in the hourly model, nobody bills that asset because nobody knows how to measure it.
The hourly rate never measured value
The hourly rate has always been a convenient approximation. For fifty years, time spent was a reasonable indicator of effort, and effort a reasonable indicator of expertise. The equation more or less held.
It held above all because it suited everyone: simple to calculate, simple to control, and it shifts the risk of overruns onto the client. It also protects the mediocre provider, since in this system being slow pays more.
That last point has bothered me for a long time, well before AI. A billing model that rewards slowness and penalises mastery is a model that discourages exactly what a client should want to buy.
If I had to sum up what has just broken, it would be this chain. For fifty years, it stood up:
Before: time → effort → expertise → value → price. Each arrow was an acceptable approximation, and the whole thing more or less worked.
Now: time ≠ effort ≠ value. The three have come apart, and the price can no longer be deduced from the first term. It is built on four others: the stakes, the result, the deadline and the risk carried.
AI does not make the value of work disappear. It breaks the link between the time needed to produce it and the price it deserves.
What exactly AI breaks in the equation
AI did not invent this problem, it made it impossible to ignore. The hourly rate had survived the spreadsheet, email and online databases: each saved time, but gradually. This time the gap is an order of magnitude, and above all it is visible to the client, who uses the same tools at home in the evening.
Wolters Kluwer published an analysis in August 2026 that sums up the situation in a phrase I find hard to contradict: “The more efficient you are, the less you earn”. Other technologies have already produced this effect, from computerised accounting to computer-aided design. But rarely has a technology made the paradox so brutal, or so fast: the more efficient the provider becomes, the more their hourly revenue shrinks, and this plays out over a few quarters rather than a decade.
Three effects add up, and they all point the same way.
The first: the client now refuses to pay for avoidable time
The same analysis notes that the AI discount is becoming widespread: clients want to see efficiency gains on their invoice, not in the provider’s margin. They did not demand it two years ago, because they did not know what a model produces in ten minutes.
The second: the billable share of the work is shrinking
Again according to the Wolters Kluwer benchmark survey, nearly 40% of lawyers now spend less than half their time on billable tasks. What remains is hours of steering, checking and judgement, which do not fit neatly into a timesheet but which weigh on the profit and loss account.
The third: the cost of production can no longer be read in hours
When part of the work is carried out by tools billed by usage, the real cost of a file mixes human time, computing consumption and the amortisation of the set-up. A timesheet cannot model that.
The blind spot of the hourly model: it never bills the preparation. The months spent building your methods, testing your tools and learning where they go wrong appear on no invoice. Yet that is what makes it possible to deliver the study in twenty-four hours. The hourly model bills the delivery and ignores the investment that makes it possible.
Speed is not a bonus, it is part of the product
This is the point I spend the most time getting people to accept, including fellow professionals. A study delivered tomorrow and the same study delivered in three months are not the same product. The PDF is identical, and yet you have not bought the same thing.
The reasoning holds for everything we sell. A site launched next Monday and the same site launched in February have exactly the same pages. For a company that has to start selling on Monday, they are two different purchases, and one of them can generate several months of revenue before the other is even delivered.
So I propose a way of reading it. It is not a scientific formula, it is a way of structuring a pricing conversation:
Value = result × reliability × speed
The result is what the client gets. The reliability is the trust they can place in it, so the checking work and the responsibility taken on. The speed is the date from which they can use it.
It is a multiplication, not an addition. If any one of the three falls to zero, the product is worth zero: a wrong study is worth nothing however fast it is, and a perfect study that arrives after the decision is worth nothing either.
The hourly model measures none of these three terms. It measures a production time, precisely the variable the client is not trying to buy. Worse, it feeds it into the price with the wrong sign: the faster you go, the lower the invoice, even though the value has gone up.
Practical consequence: when you deliver fast, do not make people wait to give the impression of heavier work. Sell the deadline as a component of the offer, and accept that a client with no urgency pays less than one who has it. That is what express transport and insurance have always done.
Musk announces the end of money, I think he has the timing wrong
This debate took a more spectacular turn once Elon Musk took it up. His position, expressed notably in an interview with The Economist in the summer of 2026, comes down to one sentence: within ten to twenty years, work will become optional and money will lose its relevance, because robots will produce such abundance that currency will no longer have a purpose. At Davos in January 2026, he presented robotics and AI as “the path to abundance for all”.
I am not building my strategy on this prediction, and I advise you not to either. It is disputed by people well placed to judge, starting with the head of Anthropic, who describes rather a painful transition for the labour market. And above all, even if it came true, it does not help me decide anything on Monday morning. The real question is not whether money will exist in 2036, but how I bill in the first quarter of 2027.
Musk talks about the end of money. I talk about the end of a unit of measurement. It is infinitely less spectacular, and it is already happening.
Where I agree with him is on the substance: exchanging time for money is a convention, not a natural law. This convention has a birth date, and it will have an end date.
The four models that replace the hour
This is not a hypothesis about the future, these models are already running. Wolters Kluwer notes, for instance, that fixed-price contracts account for 41% of the transactional work of a mid-sized law firm. Here are the four approaches, with what each one really assumes.
| Model | Principle | What it demands of you |
|---|---|---|
| Fixed fee per deliverable | One price, one scope, no surprises. The efficiency gain stays in your margin. | Knowing your real cost. Without a track record, you guess, and you get it wrong. |
| Subscription | A monthly amount for available capacity, not for a volume of hours. | A predictable workload and a written scope, otherwise the subscription becomes mutual hostage-taking. |
| Value-based pricing | The price is indexed to what the result earns or saves the client. | Knowing how to put a figure on the stakes with the client, and accepting to carry the risk if you got it wrong. |
| Hybrid, base plus bonus | A base that covers the cost, a bonus triggered when a result is reached. | An indicator that is beyond dispute and measurable by both parties, defined before you start. |
Two safeguards circulate among those who have already made the switch. First, the risk corridor: a floor that protects the provider if the job gets complicated, a ceiling that protects the client from a runaway invoice. Second, transparent re-invoicing of technology costs, above an agreed threshold, so that the consumption of tools is neither hidden nor silently absorbed.
The point nobody likes to look at: the risk changes sides
There is a very simple reason why the hourly rate has survived so long, and it is not laziness. It is that in this model, the client carries the risk. If the job takes three times longer than planned, they pay three times more. The provider never loses.
Moving to a fixed fee means taking that risk on yourself. Misjudging a project no longer results in a longer invoice but in a destroyed margin. Those who present the end of the hourly rate as a cheerful given need to be told: you are not changing your calculation method, you are changing your position in the business relationship.
This transfer is only sustainable on one condition: knowing your real costs. Abandoning the timesheet without knowing what a type of project really costs is not charging for value, it is gambling. So you have to measure your time more seriously than ever, not to bill it, but to know what you can promise.
The misunderstanding to avoid
Charging for value does not mean charging more because you are fast. It means indexing the price to the client’s stakes rather than to the provider’s effort.
On a low-stakes project, value-based pricing often gives a lower amount than the hourly meter would have. That is precisely what makes the model defensible in a meeting: it goes up when the stakes go up, it goes down when the stakes go down.
How I go about it, in practice
I do not claim to have solved the question, and I am not going to give you figures on my margins. Here is simply the method we apply at AR Digital, and you can do what you like with it.
I start with the stakes, not the scope. The first question is no longer “what do you want us to do”, but “what changes if it is done, and what does it cost you if it is not”. Until I have the answer, I do not quote. Many business owners have never put it into words, and the conversation is already worth something in itself.
I bill the deliverable and the deadline, never the hours. A quote states what is delivered, in what condition, by what date, with how many rounds of corrections. The number of hours appears nowhere. So there is nothing to negotiate about time, and the discussion is about the scope, which is a much better discussion.
I measure my time internally, religiously. This is the apparent contradiction of the model. We track the real time of each project, but this data never goes out to the client. It is used to know what we can promise next time without putting ourselves at risk. It is also what shows which processes deserve to be automated first, because you do not guess where the time goes, you observe it.
I apply a safety coefficient to new projects. With no track record on a type of assignment, the first time is paid for in learning. Better to accept that than to discover it at the end of the project.
I own the speed instead of hiding it. This is the hardest cultural change, and the one that pays the most.
What it changes if you are the client
This shift does not only concern providers. If you buy intellectual services, it gives you power, provided you change your reflexes.
- Stop comparing day rates. Comparing two quotes at the day rate no longer makes sense if one takes four days and the other twenty. Compare what is delivered, by what date, with what guarantee.
- Ask for the deadline before the price. If a decision is waiting on the delivery, each week gained has a value that only you can put a figure on.
- Ask about AI directly. Ask how your providers use it, what they check and who answers for mistakes. A provider who uses it and owns it is more reassuring than one who swears not to touch it. The same question applies to your own teams, and that is often where to start: AI training for your company costs less than a year of bad habits.
- Negotiate the sharing of the gain, not the rate. It is legitimate for you to benefit from your provider’s efficiency. It is just as legitimate for them to keep a share, otherwise they have no reason to invest in their tools.
The objections I get
If you no longer bill time, how do you justify your price?
By the result and by the deadline, two things the client sees for themselves. Time spent was never a justification, it was a shared convention. A client who receives what they expected, on the promised date, in the promised condition, does not ask for a count of hours.
Is this not an elegant way to raise your prices?
On high-stakes projects, yes, the price goes up, and that is acknowledged. On repetitive, low-stakes tasks, it drops sharply, because the client rightly refuses to pay full price for what an automated chain produces in a few minutes. A model that only went up would not survive two calls for tender.
And what if the AI makes a mistake in the study delivered in twenty-four hours?
Then it is my responsibility, exactly as if a junior colleague had made the mistake. That is in fact the whole point of the model: with a fixed fee, the error costs me, whereas at an hourly rate, the rework gets billed. A fixed fee aligns my interests with quality, an hourly rate aligns them with duration.
What happens to billing for truly unpredictable tasks?
The risk corridor answers this case properly: a floor that protects the provider, a ceiling that protects the client. For the rest, litigation, crisis, urgency that cannot be scoped, the hour remains a legitimate tool. I am not saying it disappears everywhere, I am saying it stops being the default norm.
Are you not at risk of devaluing human work?
It is the opposite. The hourly rate measures the same thing for a beginner and for an expert, a duration, even though their contributions are in no way comparable. The coefficient applied can be significant, but the unit of measurement stays the same, and it stays the wrong one. Charging for value amounts to recognising that what counts is judgement, experience and the ability to pose the right problem, things that cannot be measured in duration. What AI devalues is repetitive execution, not discernment.
Sources
- Wolters Kluwer, “The end of the billable hour? How AI is redefining the business model of law firms”, 18 August 2026. Source of the figures on the billable share of time, the AI discount and the 41% of transactional work on a fixed fee.
- Fortune, on Elon Musk’s predictions and the debate on universal basic income, 1 February 2026. Source of the Davos quote, of Dario Amodei’s position and of the Morgan Stanley figure.
- Clio, law firm pricing strategies beyond the billable hour, 2026 edition.








