Outcome-based pricing vs day rates: examples, and what the contract has to say
The short answer
Outcome-based pricing ties part of a supplier’s fee to a measured result in the buyer’s business, such as a ticket closed or an invoice matched, instead of to the days worked. The clearest services examples: TCS bills AI-led engagements on outcomes before moving them to fixed price, a McKinsey managing partner says about a quarter of its fees come from outcomes-based pricing, with fixed components left in, and E.ON and Daimler Truck have reportedly signed deals that defer fees against efficiency gains or split AI savings.
Against a day rate, an outcome fee moves delivery risk and value risk (the system runs, but the number does not move) to the supplier, and adds measurement risk: whether the number moved, and who can prove it. That risk falls on the party that does not run the meter, so keep the system of record in your estate and expect the supplier to price its exposure. Where the work is discovery, the scope is unclear or there is no baseline, a day rate is the honest contract.
Key takeaways
- Check what stays fixed. TCS converts to fixed price, McKinsey keeps fixed components, and Intercom keeps seat fees or a minimum monthly commitment; E.ON reportedly pays HCLTech nothing in year one.
- Whoever runs the meter controls the bill. Intercom and Zendesk run their own; Sierra agrees criteria up front. In a services contract, keep the meter in your systems.
- Tie the fee to process outcomes. A ticket closed or a tender sent can carry a fee. Outputs reward volume, and “hours saved” is where disputes start.
- A pre-AI contract is negotiable now. TCS says AI saves up to 15% on its engagements. If you pay per head or for capacity, ask where yours went.
What is outcome-based pricing, and how does it work?
Outcome-based pricing pays for a measured change, not for effort. It can have two parts: a fixed or recurring part that covers the supplier’s cost, and a variable part that moves with one metric against a baseline over an agreed window.
A supplier can be paid at four levels. Effort is the days worked. An output is what the supplier produces and counts itself, such as documents processed. A process outcome is an event in your process, recorded in your systems, that the work causes: an invoice posted without manual touch, a ticket closed and not reopened, a tender sent. A business outcome is the number the board reads, such as cost per invoice or EBIT, which many causes move. A fee belongs on process outcomes: outputs reward volume, and business outcomes are hard to attribute to one supplier. The business number is still what the work is judged against; the fee attaches to a discrete process event that moves it.
None of it is new. Accenture’s annual report says many of its contracts already tie its pay to “agreed-upon performance standards, milestones or outcome-based measures such as cost savings, revenue enhancement or business goals”. What changed with AI is buyer pressure: Infosys’s annual report says AI disruption and tighter budgets led it to offer output or outcome-based pricing “in certain situations in order to remain competitive”. OpenAI’s chief financial officer, Sarah Friar, said in January 2026 that in fields such as drug discovery, “Licensing, IP-based agreements and outcome-based pricing will share in the value created” (PYMNTS): a direction, not a price list.
Adoption is still thin. Gartner’s Tom Coshow told CIO Dive that only 19% of services buyers and 13% of service agreements on the seller side use such arrangements today, and Gartner projects that through 2031 less than 25% of tech CEO services contracts will use them. He called the increase “more buzz than reality”.
What are real outcome-based pricing examples in IT services and consulting?
Five services arrangements have enough public detail to learn from, and two of the named buyers are European.
| Provider and buyer | What is tied to the outcome | What stays fixed | How solid the source is |
|---|---|---|---|
| TCS, AI-led engagements | Billing on outcomes rather than hours; AI savings of up to 15% passed to clients in phases | The engagement moves to fixed price | July 2026 earnings call, via Mint |
| TCS, finance, HR and business-services segment | About 80% of contracts based on outcome performance measures | Not disclosed | Chief executive to Reuters |
| HCLTech and E.ON, cloud management, June 2025 | No payment in year one; from year two, payments tied to efficiency gains and specific business outcomes | Not disclosed | Reuters, August 2026, two people familiar; E.ON declined to comment |
| Cognizant and Daimler Truck, AI and automation, February 2026 | AI-related cost savings split between vendor and client | Not disclosed | Reuters, August 2026, people familiar; Cognizant declined to discuss it |
| McKinsey, global | About a quarter of global fees, measured against a client scorecard | “Some fixed cost components remain” | Managing partner for the UK, Ireland and Israel, and the global technology and AI leader, November 2025, via Business Insider |
Only TCS’s AI-led work and McKinsey keep a fixed part on the record. The E.ON deal pays HCLTech nothing in year one and ties later payments to efficiency gains and business outcomes; for the TCS segment and the Daimler Truck deal, no fixed share is disclosed. Outcome terms are rarely published, so ask a supplier for its contract structure, not its case studies.
Is TCS moving to outcome-based pricing for AI?
Yes, for AI-led engagements. On its July 2026 earnings call, TCS management said it bills clients on outcomes rather than hours for AI-led work, then moves those engagements to fixed-price contracts, and that AI generates up to 15% cost savings on its work, passed to clients in phases (Mint). Chief executive K. Krithivasan called 10% to 15% “a rule of thumb”, and told Reuters that about 80% of contracts in TCS’s finance, HR and other business-services segment are based on outcome performance measures.
The phased pass-through is a supplier stating on an earnings call that AI lowers the cost of work it already does, the strongest argument a buyer holding a pre-AI contract has. Mint reads the phasing as TCS preserving its margins, and Krithivasan said TCS can offer “a front-loaded commitment” smoothed over the term of the project. What you negotiate is the pace.
The second model described on the call is “fixed price, fixed capacity”: a programme first, then multiple pods once the customer sees value. A capacity contract keeps any productivity gain with the supplier unless it says otherwise.
Outcome-based pricing in AI: what do Sierra, Intercom and Zendesk count as an outcome?
Each vendor counts something different, and the vendor, not the buyer, writes the definition. For a services buyer these software prices are a lesson in definitions, not a model to copy.
| Vendor | Published price | What counts | Window or exclusions | Who judges |
|---|---|---|---|---|
| Sierra | No published rate | A completed task, such as a resolved conversation, a saved cancellation or an upsell; unresolved conversations usually free | Criteria agreed up front; routing work priced per conversation | Criteria agreed with Sierra |
| Intercom Fin | $0.99 per outcome; a minimum monthly commitment when Fin runs on another helpdesk, seat fees on Intercom’s own | A confirmed resolution, a customer who does not ask for more help after Fin responds, or a workflow (Procedure) Fin completes, including handoffs | One outcome per conversation | Intercom |
| Zendesk | Per automated resolution | Since 18 May 2026, only verified resolutions count; assisted escalations and contained resolutions are free | Email: 72 hours after the first email; messaging: 2 hours by default | A language model reads the conversation |
Sierra’s pricing post dates from December 2024; its July 2026 Horizon announcement says customers “pay for business outcomes delivered”, not for tokens. Intercom counts a customer who stops asking; Zendesk counts only what the AI finishes alone: a Zendesk executive told CIO Dive that a conversation 90% handled by AI with 10% passed to a human is not counted. Salesforce’s per-resolution pricing is examined in the Agentforce post.
Intercom and Zendesk count on their own platforms, and at Zendesk the judge is a model the vendor runs. The buyer does not run the count and cannot rerun Zendesk’s judge, so the buyer carries the measurement risk. That is tolerable for a product you can walk away from. In a services contract, negotiate the definition and keep the meter in your own systems.
Outcome-based pricing in consulting: why is it harder than in software?
Because the supplier does not control the result. A software vendor owns the product, the event and the meter. A consultancy or integrator works inside your company, where the result depends on your people’s decisions, your data and other suppliers’ changes.
Anshuman Vedi of Techspire Consulting describes the ERP version in Consultancy.uk: the client controls operating decisions and data, the integrator configures the system, and when a data decision is delayed the outcome slips between the two contracts. If the fee depends on decisions outside the supplier’s authority, he writes, “suppliers will price that uncertainty into the contract” or narrow what they guarantee.
The buyer’s own organisation adds friction. “Procurement and finance want certainty,” NiCE’s Kevin Lee told TechTarget, so an outcome share needs a cap that finance can budget.
Outcome-based vs time and materials vs fixed price: who carries which risk?
A day rate sells effort and leaves the risk with the buyer, as the Accenture post sets out. Every contract allocates four risks: scope risk (the work turns out bigger than planned), delivery risk (the system does not work or never reaches production), value risk (the system runs, but the number does not move) and measurement risk (whether the number moved, and who can prove it).
| Time and materials | Fixed price | Outcome-based | Hybrid: fixed build, run fee, outcome share | |
|---|---|---|---|---|
| You pay for | Days worked | An agreed deliverable, accepted | A measured change in a process number | A deliverable, its operation and part of the change |
| Scope risk | Buyer | Supplier, within change control | Supplier, but scope changes reopen the baseline | Supplier for the build, shared after |
| Delivery risk | Buyer | Supplier, up to acceptance | Supplier, until the window closes | Supplier for the build, shared for the run |
| Value risk | Buyer | Buyer: a working system that moves nothing is still accepted | Supplier, for the fee at stake | Supplier only for a holdback; with a bonus only, the buyer |
| Measurement risk | Not priced: no fee depends on a number | Low: acceptance tests only | Supplier, if the meter is in your systems; you, if it is in theirs | On the outcome share only: supplier if the meter is in your systems, you if it is in theirs |
| Where the supplier’s premium hides | Rate and headcount | Contingency in the price | Base fee, or a narrow outcome definition | Build fee and caps |
| Honest when | Scope is unknown, or the work is discovery | Scope is known and the result is a working system | The result is a frequent, discrete, attributable event with a baseline | You want a working system and a number |
A December 2025 report on relational contracts that the law firm CMS Kluge prepared for Norway’s Directorate of Health puts the two ends in one line (chapter 9): supplier risk is highest under fixed price and lowest under hourly billing. Outcome pricing adds the risks neither end prices, value and measurement, and the party holding the data controls how they are settled. CIO Dive, reporting Gartner’s Tom Coshow, offers one test for the whole table: “If the vendor isn’t taking on the risk, why are you bothering with outcome-based pricing?”
What does an outcome-based contract have to specify to be real?
Four elements are the floor: a number agreed before the build, a measured baseline, a defined window, and agreement on who owns the data that proves the result. TechTarget’s reporting from Dreamforce put the open questions as “Who controls the measurements? How will disputes be resolved?”
- The metric, defined to the edge case. Write the unit and what counts, with examples. Is a ticket reopened within a set period closed? Was a customer who went silent helped? Does a case that ends in a handoff count? Intercom counts a silent customer, and a handoff when it is part of a workflow (Procedure) Fin completes. Zendesk does not count a handoff, and counts a silent customer only if its model judges the request resolved.
- The baseline, measured and signed. The period it covers, the source, who measured it and the date both parties signed it. A baseline estimated after go-live is a negotiation, not a measurement.
- The window and the calculation. When counting starts, how long it runs, how often the share is invoiced, and how events that straddle two periods are counted.
- The system of record. Name the system whose data counts and keep it in your estate, with read access for the supplier. Both sides should be able to recount the same events from the same table.
- Attribution. Prices, volume and mix, an ERP release, a reorganisation or the season can move the number in the same window. Agree in advance how they are netted out: a control group, an adjustment formula or a list of excluded events.
- Decision rights and dependencies. What the supplier controls and what you owe it, with dates: data access, decisions, test users. Miss one and the window extends. The ERP example above lacked this clause; Vedi’s remedy is a jointly appointed delivery lead who takes day-to-day decisions, and a sponsor who steps in only above agreed limits. Without that person, he writes, “outcome-based pricing is simply a new fee model attached to the same old delivery problem”.
- Re-baselining. A change request that alters the scope resets the baseline and the target through the same change control, so nobody relitigates the number by stealth.
- Caps and floors. The fixed part is the supplier’s floor. Cap the outcome share so finance can budget it, and say whether results above target earn anything.
- Handoff and acceptance. What “handed over” means, who runs the system during the window, and whether the outcome share survives if your team changes it.
- The dispute path. An independent measurer, a timetable and a cost split, agreed up front, so a disagreement about a number is settled by recounting it.
- Termination inside the window. What the supplier is owed if you end the contract, or change the process, before the window closes: pro rata on the count to date, or nothing.
Ownership, model portability and maintenance are separate terms, in what to settle before you sign and the ten questions to ask a partner. Baseline and window for an ROI claim are in the questions to ask before an ROI number goes into your case.
Which metric should an outcome-based fee be tied to?
To a process outcome: a discrete event your systems already record and that the supplier’s work can plausibly cause. Forrester’s Abhijit Sunil, writing about IT sustainability services, says outcome pricing works when the problem is “measurable, bounded, and controllable” and the buyer has a credible baseline (Forrester).
| Metric | Level | Discrete event? | Attributable to the supplier’s work? | Fit for an outcome fee |
|---|---|---|---|---|
| Documents processed by the system | Output | Yes | Yes, and the supplier controls the count | Poor: it rewards volume |
| Invoice matched and posted without manual touch | Process outcome | Yes | Mostly, once volume is adjusted for | Good |
| Ticket closed and not reopened within the agreed period | Process outcome | Yes | Mostly | Good |
| Tender sent, with the first draft written by the system | Process outcome | Yes | Mostly | Good |
| Cost per case | Business outcome | Calculated | Partly: wages, volume and mix move it too | Only with an adjustment formula |
| Hours saved | Productivity estimate | No, estimated | Hard to prove | Poor |
| Revenue, margin or EBIT | Business outcome | No | Many causes | Poor, except as a small shared bonus |
Discrete events can be counted from your own logs and disputed one at a time. Hours saved is usually a time study multiplied by headcount, and the people who save the hours rarely leave, so the saving never reaches the cost base the CFO reads.
Good metrics can still be gamed. If a silent customer counts as resolved, an agent that stops replying earns fees. Decagon, which sells both models, said in December 2024 that most of its customers chose per-conversation pricing so they were not arguing over whether a frustrated customer who stops responding was resolved (Decagon). The opposite trap is a metric that counts only full automation, as Zendesk’s does: it pays nothing for AI that does most of the work before a person finishes it, and pushes a supplier away from the handoffs your process needs. Pick the event your business would count anyway, and write down the exclusions.
What does a hybrid pricing model look like for AI delivery?
A fixed part pays for the build, a recurring part for running it, and a variable part moves with the outcome. The variants differ most in who carries value risk.
| Variant | What is fixed | What varies | Who carries value risk |
|---|---|---|---|
| By phase | Discovery on a day rate or a small fixed fee, then a fixed-price build | A run fee with an outcome share | Buyer, until the outcome share starts |
| Target price | A target cost | A bonus if the work comes in under it, a lower fee if it overruns | Buyer: the target is on cost, not on the number |
| Deferred fee | Little or nothing in the first period | Later payments, tied to results | Supplier, for the deferred payments |
| Savings split | The cost baseline, measured line by line before the build | The saving, divided between supplier and client | Shared, on the saving |
| Holdback | Most of the build fee | The held-back part, paid only when the count passes the baseline | Supplier, up to the holdback |
TCS starts on outcomes and moves to fixed price, the reverse of the first row. A supplier that defers its fee, as reported for HCLTech and E.ON, is financing your first year; expect that cost to appear in later payments. A bonus pays the supplier more if the number moves; a holdback pays it less if it does not. Of the two, only the holdback moves value risk, and a hybrid built only on bonuses is a fixed price with a success fee.
A worked example, in variables. The process is supplier-invoice matching. Before the build, both sides measure the baseline B, invoices matched each month without manual touch, from the ERP’s own logs. The build costs a fixed fee F, scoped against what the manual process costs today, of which a holdback H is paid only once the count passes B. Running the system costs a recurring fee R. In each window, every matched invoice above B earns the supplier an agreed amount p, up to a cap C. If volume falls, B is scaled by a formula; if an ERP upgrade lands mid-window, those weeks are excluded. The day-rate quote for the same work is D days at rate r.
| What happens | Day-rate contract | Hybrid, bonus only | Hybrid with holdback |
|---|---|---|---|
| The count stays at B | D days at r, in full | F and R | F minus H, and R |
| The count passes B | D days at r, in full | F and R, plus p for each invoice above B, up to C | F and R, plus p for each invoice above B, up to C |
| The scope grows | More days: D rises | Change control re-baselines B and reprices F | Change control re-baselines B and reprices F |
Only the holdback column pays less when nothing moves. Every term has to be written before the build, or it gets negotiated after the result is known.
When are day rates the honest answer?
When nobody can yet say what the work is, or what it should move. Discovery is the first case: the output is a decision rather than a system, and paying for time is correct if the phase is short and has a fixed end. Unclear scope is the second: when the work depends on systems nobody has opened or a decision nobody has taken, a fixed price is a guess with contingency on top, and an outcome fee is a guess with a premium on top. The third is no baseline: if nobody recorded the current cost, error rate or cycle time, measure it first, on a day rate if necessary, and price the build afterwards.
Day rates also suit capacity your own managers direct, bought from a supplier set up to supply it: in Germany, engineers integrated into your organisation and taking your managers’ instructions count as agency workers, which needs a permit, and Italian law applies a similar test (who carries the delivery risk). The mistake is paying by the day for work that has a scope and a number, and keeping a risk you could have transferred. Which kind of provider suits each case is in when an integrator or a consultancy is still the right answer.
Is there published consulting day-rate data for DACH, the Nordics and Italy?
The only recent published average we found is Germany’s. In December 2025 the consultancies’ association BDU reported an average billed day rate of 1.000 Euro in 2025 (German notation for one thousand), 1.600 Euro for a managing director or partner and 700 Euro for an analyst, with IT consulting fees down 3% on the year (BDU). These are survey averages, not what the largest integrators quote. Sweden’s state framework, run by Kammarkollegiet, sets fixed and ceiling hourly prices in its framework price annex, not averages, and we found no comparable primary source for Austria, Switzerland, Italy or the rest of the Nordics. UK public-sector rate cards for AI engineers are in what forward deployed engineering costs.
How do you renegotiate an outsourcing contract that was priced before AI?
Start with how the contract pays. If your incumbent is paid per head, per FTE or for fixed capacity, or runs a fixed-price managed service priced on pre-AI effort, AI lowers its cost and the supplier keeps the difference. If it bills days, you keep the saving in principle, and the question is why the day estimates in each new statement of work have not fallen. The three questions to ask before you sign cover a new contract. At renewal, three terms matter.
A productivity clawback. The rate falls or the deliverable grows, by contract rather than goodwill. TCS puts the savings at up to 15%. For per-head and capacity contracts, write a productivity factor on the rate for each contract year, with a schedule for how much of the saving reaches you in year one. For time and materials, cap the days per repeatable deliverable, or move that work to a fixed price per unit. If the contract has a benchmarking clause, invoke it now: a benchmark taken after AI changed delivery costs is the evidence the clawback needs.
A right to audit AI use inside the engagement. Which AI tools touch your data, whose models they call, where they run, and whether your data trains anything the supplier reuses. Where those tools process personal data, the model provider is typically a sub-processor under your data-processing agreement, and Article 28(2) GDPR says a processor “shall not engage another processor without prior specific or general written authorisation of the controller” (GDPR art. 28). Ask for the current sub-processor list, and require notice before new tooling enters the engagement.
An exit ramp. The drafting for termination assistance and data export is in the exit-clause checklist. A day-rate incumbent needs two additions: a handover of code, runbooks and know-how that a smaller team can run without re-hiring the supplier’s headcount, and transition-period rates agreed now, before notice is served.
Accountability that is not measured gets argued over later. In a services case, not an AI one, Zimmer Biomet alleges in a September 2025 complaint, announced by its counsel Loeb & Loeb, that it paid $94 million in fees for a Deloitte Consulting SAP S/4HANA implementation full of defects, and it seeks at least $172 million. Deloitte moved to dismiss in November 2025, according to the advisory firm UpperEdge, and we have seen no ruling. The dispute is over what was promised before the work began and what was delivered. A measured number gives both sides something to count first, though it does not prevent a misrepresentation claim.
How does Nucleo price, and when are we the wrong choice?
We do not sell days. Our terms, on the services page, have three parts: a fixed fee to build, scoped against what the process costs you today in hours, errors and delay; a recurring fee to run and evolve what we built and keep the context engine current as the work changes; and, where the outcome is a discrete event we can attribute, such as a ticket closed or a tender sent, part of the fee tied to it. There is no rate card, and we have published no client results, so the examples in this post are other firms’.
The structure has friction. The outcome part depends on an event we can attribute against a measured baseline, so without one it cannot be priced. Each engagement is scoped to the company. By the test above, ask any supplier, us included, whether its outcome part is a bonus or a holdback, because only a holdback moves value risk to the supplier. And the build fee is fixed only once the process cost is scoped: a scoping step comes before a price, which is slower than reading a rate card.
If you need open-ended capacity directed by your own managers, from a supplier set up to supply it, a short strategy diagnosis, or a programme whose result is a board scorecard rather than an event your systems record, a day-rate or fixed-fee firm is the more honest contract. Where the outcome is a discrete event we can attribute, part of our fee is tied to it; EBIT is not a discrete event, so it falls outside that structure.
FAQ
What are examples of outcome-based pricing? TCS bills AI-led engagements on outcomes before moving them to fixed price. A McKinsey managing partner says about a quarter of its fees come from outcomes-based pricing, with fixed components remaining. HCLTech with E.ON, and Cognizant with Daimler Truck, have reportedly signed deals that defer fees or split AI savings. Intercom charges $0.99 per outcome.
How does outcome-based pricing work in AI? The vendor charges per completed result and defines what counts. Intercom counts a confirmed resolution, a customer who asks for no more help, or a workflow Fin completes, including handoffs. Zendesk counts only verified resolutions without human escalation, judged by a language model. Sierra agrees criteria with each customer. Compare the definitions before the prices.
Is TCS moving to outcome-based pricing for AI? Yes, for AI-led engagements. On its July 2026 earnings call TCS said it bills them on outcomes rather than hours, then moves them to fixed price, and that AI saves up to 15% on its work, passed to clients in phases. Its chief executive told Reuters that about 80% of contracts in its finance, HR and business-services segment use outcome measures.
Does outcome-based pricing work in consulting? Only where the consultancy controls enough of the result. Outcomes depend on the client’s decisions and data, so suppliers price that uncertainty into the contract or narrow what they guarantee. It works for discrete, measurable results with a baseline. Gartner, via CIO Dive, says only 19% of services buyers use such arrangements today.
Outcome-based vs time and materials vs fixed price: who carries the risk? Under time and materials the buyer carries scope, delivery and value risk. Under fixed price the supplier carries scope and delivery risk up to acceptance, and the buyer keeps value risk: the system runs but the number does not move. Under outcome pricing the supplier carries value risk for the fee at stake, and measurement risk falls on whoever does not run the meter.
What should an outcome-based contract specify? A number agreed before the build, a measured baseline, the window, and who owns the data that proves the result. Then the metric’s edge cases, a system of record in the buyer’s estate, attribution rules, decision rights, re-baselining, caps, acceptance, a dispute path, and what is owed on early termination.
When are day rates better than outcome-based pricing? For discovery, where the output is a decision rather than a system; when the scope depends on systems or decisions nobody has examined; and when there is no measured baseline. Measure the baseline first, on a day rate if necessary, then price the build.
How do you renegotiate an outsourcing contract that was priced before AI? Start with how it pays. Under per-head, FTE or fixed-capacity pricing the supplier keeps the AI saving, so ask for a productivity factor on the rate with a schedule. Under time and materials, ask why day estimates per deliverable have not fallen, and cap them. In both, add a right to audit AI use and an exit ramp.
Related reading
- Accenture’s worst day on the market: what it tells a CIO about the vendor they are about to sign
- Forward-deployed engineer vs consultant vs system integrator: a decision rule for buyers
- What is a forward-deployed engineer (and why enterprise AI needs one)
- The ROI of enterprise AI agents: 2026 benchmarks
- Salesforce’s own partners are not seeing Agentforce ROI
- Build vs buy vs embed: choosing an enterprise AI delivery model
Nucleo is an AI-native transformation company for European mid-caps. Our forward-deployed engineers build the context engine your company runs on, in your estate and model-agnostic, and put systems into production on top of it. See how an engagement works, or write to us.
Sources: Mint on TCS; Reuters via MarketScreener; Business Insider on McKinsey; Accenture Form 10-K; Infosys Form 20-F; PYMNTS on OpenAI’s CFO; CIO Dive; TechTarget; Consultancy.uk; Forrester; Sierra on outcome-based pricing; Sierra on Horizon; Intercom pricing; Zendesk automated resolution tiers; Decagon; CMS Kluge for Helsedirektoratet; Kammarkollegiet; BDU, Honorare im Consulting 2025, December 2025; GDPR art. 28; Loeb & Loeb on Zimmer Biomet v. Deloitte; UpperEdge on Deloitte’s motion to dismiss.