Is the day rate doomed? What AI really changes about selling IT services

The day rate isn't disappearing. But it's losing some of its explanatory power. When an engagement is sold because it deploys ten people for three months, automating part of the work mechanically weakens the price's justification. When it's sold to access rare expertise, absorb heavy uncertainty, or temporarily reinforce a team, time remains, by contrast, a perfectly rational contractual unit.

The most likely transformation, then, isn't the end of the person-day. It's the accelerated diversification of the contract mix: day rate, fixed price, managed capacity, subscription, reusable assets, SLAs and outcome-based pay will increasingly coexist. The real question for IT services and consulting firm leaders is knowing which pricing unit still matches the value actually being bought.

In brief

• •Pricing pressure is real, but the available data doesn't let us attribute it mainly to AI. In 2026, 49% of IT services firms surveyed by Numeum-Xerfi cite pricing pressure among their top headwinds, but 54% also cite client caution.
• •AI is already changing delivery. In the 2025 KPMG-Numeum study, 72% of the IT services and technology consulting firms surveyed say they use generative AI in their delivery processes — a self-reported figure, not a productivity measurement.
• •The economic paradox is real: saving time doesn't automatically mean gaining margin. Numeum-Xerfi estimates AI-linked productivity gains could rise from 15% in 2025 to 22.3% in 2027 at IT services firms, while stressing they remain hard to convert into margin.
• •Capacity- and outcome-based models are gaining ground, but aren't uniformly replacing time spent. Deloitte observes outcome-based models rising in outsourcing, even as 80% of the executives surveyed still plan to maintain or increase their spend with external providers.
• •The day rate becomes fragile when it's used to price automatable output. It stays relevant when it pays for uncertainty, availability, accountability, or expertise whose result can't be cleanly isolated.

The false debate: "AI will make the day rate disappear"

The phrase is seductive because it seems to follow from unassailable logic.

An engagement needs 500 days. AI lets you produce faster. The same engagement now needs 350 days. If revenue equals the day rate multiplied by the number of days, the provider bills less. The model would then end up destroying itself the more productive it becomes.

That mechanism does exist. But it only works if several things stay constant: the scope purchased, the price, demand, the expected quality level, and the contract model. And those are precisely the variables that move.

A team can use the time saved to expand scope, test more, cover more applications, or improve security. A fixed-price provider can temporarily keep part of the productivity gain. An engagement can also shift toward a monthly capacity, a managed service, or an outcome commitment.

Staff augmentation is under pressure, but nothing shows it's disappearing across the board; the day rate is mainly becoming insufficient as the sole justification of value in certain heavily automatable scopes.

That's far less dramatic than "the end of the day rate". It's also far more structural.

What the data actually shows: the pressure is real

Let's start by separating three phenomena that are often blurred together: pricing pressure, productivity gains, and how contracts are evolving.

Pricing pressure is documented. The Numeum-Xerfi Observatory published in July 2026 puts the French IT services market at €34.6 billion, growing 1%. Among the firms surveyed, 49% cite price pressure as a headwind and 58% say they're repositioning, notably by deploying automation and AI. At the same time, 35% expect their operating margin to decline in the first half of 2026.

It would still be an overreach to conclude AI is driving day rates down. The same survey cites client caution as the top headwind, at 54%. The sector is also coming out of two difficult years: the IT services market had contracted 1.8% in 2025. The economic climate, supplier consolidation, budget trade-offs, competition and AI all overlap. The data doesn't let us cleanly isolate AI's causal effect on prices.

Delivery is already changing. KPMG and Numeum's 2025 Grand Angle study, run among nearly 200 IT services, engineering and technology consulting firms based in France, finds that 72% say they use generative AI in delivery, and nearly half say they have in-house AI or generative AI skills.

Finally, Numeum-Xerfi estimates AI-related productivity gains at IT services firms could rise from 15% in 2025 to 22.3% in 2027. But the same publication specifically flags how hard they are to convert into margin. That's probably one of the most important figures for understanding this topic: technical productivity and economic performance aren't synonyms.

The real shock is less about the rate than about what the client still agrees to pay for

In consulting, automating research, synthesis or prep work changes the perceived effort required and pushes clients to demand more proof, more results and more transparency — hence the growing exploration of outcome-based pricing, Consulting-as-a-Service, and asset-based consulting as complements to the "time × people" model.

Transposing this to IT services firms calls for more caution. A strategic analysis and a production IT system don't carry the same risk, the same acceptance criteria, or the same accountability.

But the economic causality is similar.

Take a highly standardised engagement: generating tests, documentation, initial ticket triage, or building repetitive components. If the provider justifies its price solely by having eight people on it for a hundred days, the client can legitimately ask why automation isn't reflected in the contract's economics.

Conversely, picture a specialist architect working three weeks on a complex transformation whose scope shifts daily. The buyer doesn't know exactly what will be uncovered, can't define a clean contractual result, and above all wants access to a specific skill. Billing by time still makes sense.

AI, then, doesn't make time intrinsically obsolete. It makes it harder to bill for time when a large share of it corresponds to standardisable output that the client now knows can be sped up.

That's a decisive difference.

Staff augmentation is most exposed when human headcount is itself the value proposition

The most sensitive case is staff augmentation where the sales pitch can almost be reduced to: profile × day rate × duration.

The more AI boosts that profile's output, the more three questions surface. Why do we still need this many people? Why do we still need this many days? And who should benefit from the productivity gain: the client, the provider, or both?

The ESN & ICT Forum organised by Numeum in June 2026, which brought together more than 60 sector leaders, reflects precisely this signal from the field: day-rate staff augmentation is described as increasingly challenged by capacity, outcome or shared-gain commitments. This is qualitative sector feedback, not a representative study that would let us forecast staff augmentation's disappearance.

The distinction matters because staff augmentation can serve an economic function that's nothing like obsolete: having the client absorb scope uncertainty while buying flexible human capacity.

The more unpredictable the problem, the more sense that makes. The more repeatable and measurable the problem, the less sense it makes.

The counter-example that should stop anyone declaring the "end of time"

Deloitte's global outsourcing survey is a good antidote to overly quick conclusions. Among the more than 500 executives surveyed, 83% already say they use AI in their outsourced services. Yet only 25% report a drop in supplier costs or an improvement in service quality attributable to this shift. Deloitte points in particular to governance and contracting difficulties.

Even more interesting: 70% say they've selectively brought certain scopes back in-house over the previous five years, but 80% also plan to maintain or increase their outsourcing spend. Outcome-based models are gaining ground, without replacing every other contract form.

In other words, AI isn't producing a single trajectory from "lots of consultants billed by time" to "no consultants at all, only outcome-based pay".

It's producing a more fragmented landscape: insourcing on some topics, outsourcing on others, managed services, fixed prices, dedicated capacity, software assets and staff augmentation keep coexisting.

In which cases does the day rate stay relevant?

The starting point shouldn't be the technology used, but the economic nature of the uncertainty.

The day rate stays rational when the client is essentially buying availability and expertise, and it would be artificial to claim the result can be known precisely in advance.

That's the case for a rare expert deployed on demand, a complex investigation, a crisis, a programme whose scope keeps shifting with the client's own decisions, or a team temporarily embedded in an organisation the client still steers.

It also stays relevant when attributing a result is too complex. If a programme depends on the business, three providers, a software vendor, imperfect data, and the client's own internal decisions, pinning a financial outcome on the provider alone can produce a contract that's theoretically modern but economically bad.

Finally, the day rate has a virtue that's often forgotten: it's simple to understand, compare, budget and administer.

Its problem, then, isn't that it's old. Its problem shows up when its simplicity masks the wrong unit of value.

And when does it become insufficient?

The situation flips when four conditions line up: the work is repeatable, the scope is stable enough, performance is measurable, and the provider can genuinely influence that performance.

ISG already observes this shift in Application Development and Maintenance: European companies are progressively integrating generative-AI agents into the development, testing and application-management cycle, while engagements move toward more performance-driven services. ISG's framework, in fact, distinguishes the traditional model — mixing time & materials and fixed price, among others — from emerging models combining outcome-based pricing, subscriptions and continuous performance management.

That still doesn't mean outcome-based is superior. The more commitment a provider carries, the more control levers it needs. Otherwise, the client transfers a risk the provider can't actually manage, and inevitably ends up paying for that uncertainty as a premium.

The right contract isn't the most innovative one. It's the one that puts the risk in the right place.

The next step isn't one new single model: it's a portfolio of models

This is probably where the "end of the day rate" thesis becomes most misleading. The same services company can need five different contractual logics.

The strategic question, then, isn't "what do we replace the day rate with?". It becomes: "how much of our portfolio should still be sold by time, and how much now has what it takes to be sold differently?"

That framing profoundly changes the job of a BU leader.

A point often forgotten: selling by outcome demands being better at measurement

It's relatively easy to decide you want to "sell value". It's far harder to define what that means contractually.

Imagine an IT services firm promising to cut time-to-production by 30%. What happens if the client's staging environment is down three days a week? If its Product Managers sign off on specs eight days late? If another vendor controls the infrastructure?

A fully outcome-based fee would put the provider on the hook for variables it doesn't control.

This is why a hybrid contract looks more robust in many cases: a base fee pays for the capacity, costs and accountability the provider actually carries; a variable component pays for a defined improvement on indicators whose drivers are sufficiently controllable.

Measurement difficulty and moral-hazard risk are, in fact, among the recognised limits of outcome-based pricing, which means proof, traceability and risk allocation need to be defined before making value the new contractual unit.

Value-based pricing, then, doesn't remove the need for contracting. It makes it more demanding.

The other mistake would be turning every AI gain into a client discount

For the provider, the mirror-image temptation is to promise straight away: "We're 20% more productive, so we'll be 20% cheaper."

This strategy raises at least two problems.

First, the observed gains aren't free cash. AI comes with its own costs: licences, models, infrastructure, governance, security, observability, asset maintenance, and above all human oversight time. The relevant economic margin has to account for technology and supervision costs, not just payroll cost.

Second, systematically handing back the entire productivity gain to the client would remove the provider's economic incentive to keep investing in industrialisation.

A more sustainable approach is to look at what the gain actually lets you produce: fewer days, more scope, higher quality, better availability, lower risk, or some combination of these.

AI, then, raises a new commercial question: how do you share the productivity gain without erasing the economic value of whoever made it possible?

This change hits sales teams and BU leaders first

The Business Manager used to building an offer around profiles, seniority levels and rates has to learn an extra grammar.

They need to be able to explain the client's baseline, the unit of value being sold, what automation genuinely changes, and what the provider does and doesn't guarantee.

That doesn't mean they stop selling people. In some engagements, the person is precisely the product: their experience, their availability, their sector knowledge, or their ability to take on accountability.

But when they're selling an industrialised service, continuing to think purely in FTE counts and day rates can become a commercial weakness.

On the leadership side, this also forces a rethink of certain metrics. A model that only rewards utilisation rate, billed days and margin rate can penalise a team that automates its work effectively. Conversely, measuring only recurring revenue or the number of agents deployed can mask unprofitable assets.

The pricing question, then, ends up becoming an organisational one.

Procurement, too, would do well to move past the day-rate-discount obsession

A 10% rate cut is visible in a spreadsheet. An improvement in delivery quality, a reduction in rework, or better accountability on run operations is less so.

That partly explains why the day rate persists: it's an excellent procurement metric, even when it becomes an imperfect metric of value.

Simply asking "can you cut your day rates?" can push the provider to claw back margin elsewhere — for example in team composition or supervision level. It's more useful to shift part of the negotiation toward proof, traceability, governance and expected results.

For an IT services firm, this can translate very concretely. Instead of comparing just two day rates, a buyer can compare average delivery lead time, rework rate, production defects, the level of accountability assumed, the assets included, and reversibility.

The day rate remains a data point. It simply stops being the whole story.

What the data especially doesn't let us claim

• •We can't claim AI is currently driving a general fall in day rates. Pricing pressure is documented; its precise cause isn't.
• •We don't have a universal AI productivity rate that would let anyone automatically recalculate prices. Even Numeum presents its figures as sector estimates and notes their conversion into margin remains difficult.
• •Nothing lets us claim outcome-based pricing will become the dominant model. Deloitte and ISG observe it gaining ground, not a full substitution for existing models.
• •AI isn't sealing the fate of service providers. Despite the selective insourcing Deloitte observes, four out of five executives surveyed in its study plan to maintain or increase their spend with external providers.

These counter-examples matter: they stop a real economic tension from being turned into an inevitable scenario.

Dryve's read: the day rate isn't disappearing, it's dropping a level

The day rate will remain an important billing unit. What it should progressively stop being, though, for a growing share of engagements, is the main argument explaining the contract's value.

That's the shift that matters.

A contract can perfectly well keep day rates in an appendix to handle change requests, one-off expertise or overruns, while selling, at the top level, a capacity, a service, a fixed price, or a performance commitment.

The market's transformation, then, isn't measured only by the percentage of revenue officially "outcome-based". It's measured by a subtler question: what now structures the sales conversation first?

"How much does your senior developer cost?"

Or: "What problem are you taking accountability for, at what service level, and how do we measure it?"

The shift will probably be gradual, uneven across roles, and sometimes reversible. But that's where the most credible economic transformation is playing out.

Self-diagnostic: does your day rate still measure what you sell?

For each question, score 0 if the answer is no, 1 if it's partial, and 2 if it's clearly yes.

• •Do you know what share of your revenue directly depends on the number of days billed?
• •Have you identified the engagements where AI genuinely reduces human effort, with a credible baseline?
• •Do you measure the gain across the whole workflow — quality and rework included — rather than on an isolated task?
• •Can your salespeople sell something other than a profile, a day rate and a duration?
• •Can you define a measurable capacity or service level for at least one offer?
• •Do you have engagements where a fixed price would make more sense than staff augmentation because the result can be specified well enough?
• •Do you know precisely which results can be attributed to your work and which depend too heavily on the client?
• •Do your contracts specify who benefits from the productivity gains coming from your assets and automations?
• •Do your margin calculations account for licences, models, AI infrastructure, asset maintenance and supervision time?
• •Can you explain to a buyer why some engagements should stay on the day rate and why others shouldn't anymore?

0 to 7: your model is still heavily indexed on effort. The priority probably isn't creating a new pricing scheme, but identifying where value can actually be measured.

8 to 14: your mix is starting to evolve. The risk is stacking up fixed price, staff augmentation, subscriptions and bonuses with no clear doctrine for allocating risk.

15 to 20: you have the foundations to actively steer your contract mix. The challenge then becomes checking that your sales organisation, your incentives and your margin management genuinely keep up with this shift.

The conviction

AI isn't dooming the day rate. What it's mainly dooming is the idea that a day sold is always, on its own, proof of value.

Time remains a good unit when it pays for rare expertise, availability, or uncertainty that can't be contracted for any other way. It becomes less convincing when the provider is really selling a stable, industrialisable, measurable process.

The services companies that make this shift probably won't replace their historic model with a new miracle model. They'll instead learn to choose the pricing unit based on what they genuinely take responsibility for.

It's a pricing change. But it's also a change of job for BU leaders, salespeople, delivery managers and buyers.