The new reality of outsourcing contracts in the age of AI
By the end of the first half of 2026, artificial intelligence has stopped being a peripheral option in the IT outsourcing industry and has become the backbone of operational delivery. According to McKinsey, 88% of organizations already use AI regularly in at least one business function. According to Deloitte, two thirds (66%) report productivity and efficiency gains from adopting it.
For IT service providers, this shift is an unprecedented opportunity: lower operating costs, automate low-value tasks and widen margins. For clients (the CIOs, CFOs and procurement leaders who manage outsourcing contracts) the critical question is a single one: who captures the value AI creates inside the contract?
The evidence gathered in 2025 and 2026 suggests that, in most active contracts, the answer is unequivocal: the provider. This article examines the phenomenon, quantifies its impact, identifies the warning signs and proposes a framework for action for leaders managing active outsourcing contracts.
What is the Captured AI Dividend?
The Captured AI Dividend is the difference between what an outsourcing provider charges the client and what it should charge given the actual level of automation and artificial intelligence already operating inside the contract.
It is, in essence, a saving that has already been generated in the provider’s production chain but never reached the client’s balance sheet. The provider automated processes, reduced assigned staff and deployed generative AI tools that speed up development and support, but the contract still reflects the old cost structure.
When those efficiency gains go entirely to the provider’s margin, the client is subsidizing a competitive advantage it paid for itself. This is not a theoretical projection. It is a transfer of value happening today in thousands of active contracts.
From labor arbitrage to Intelligence Arbitrage
For three decades, the logic of IT outsourcing rested on labor arbitrage: find jurisdictions with lower wage costs, contract by FTEs or hours, and deliver a proportionally linear service. More volume required more people, and more people meant more billing. The relationship between effort and cost was predictable.
Artificial intelligence broke that linearity. Today a provider can do with a team of 5 what used to take 15, and the marginal cost of AI tends toward zero as transaction volume grows. Yet if the contract is still structured around FTEs, hours or tickets, that efficiency becomes direct provider margin, not client savings.
The game has changed for good. It is no longer about who has the cheapest labor. It is about who controls the intelligence that runs the process, who owns the logic and data that intelligence generates, and who captures the resulting economic value.
This new paradigm, which we call Intelligence Arbitrage, demands a fundamental review of the contract architecture that underpins the client-provider relationship.
The evidence: what research firms say
Productivity promises vs. delivery reality
Outsourcing providers promise productivity gains of 40% to 70% with AI-enabled services. However, according to a joint panel by Morgan Lewis and Boston Consulting Group, the delivery reality is “significantly more complex”, requiring operating-model redesigns that most contracts do not provide for. In practice, the provider captures the gains it can implement unilaterally, and the client absorbs the complexity of the transition.
Deflation that never reaches the invoice
Gartner and Everest Group document real deflation of 10% to 15% a year in commoditized services: Level 1 support, manual QA, transactional BPO. These are the service layers where AI has the most immediate impact. If the contract rate for the service desk or basic support has not gone down in the last 12 months, the difference has become additional provider margin.
The commercial model gap
BCG reports that 60% of providers still bill under time and materials (T&M) or fixed-price models, precisely the models that let them keep all the productivity gains from automation. At the same time, more than 70% of enterprise clients say they prefer models tied to results or outcomes. This gap between what the client needs and what the provider offers is where the AI Dividend accumulates undetected.
Savings promised, not delivered
ISG (Information Services Group) states the disconnect bluntly: providers promise double-digit savings from AI, but ISG has yet to see real contracts that deliver that level of savings to the client enterprise. The savings are generated in the provider’s production chain. They are not passed on to the client.
The governance gap
ISG’s State of BPO report from June 2026 adds another dimension: only 21% of companies believe they have the in-house skills to govern AI-enabled services. At the same time, two thirds of those companies expect the provider to lead AI adoption. This asymmetry creates a governance vacuum that structurally favors the provider.
The four signs of a Captured AI Dividend
In three decades of working exclusively on the client side, I have found that Value Leakage does not announce itself; it is detected through signals. These are the four signals that point specifically to the AI Dividend:
Signal 1: The invoice didn’t move, but the service “improved”
If the provider boasts of operational improvements (tickets resolved faster, fewer escalations, higher availability) but the contract cost has not gone down, the obvious question is where that efficiency came from. If the answer is AI, automation or autonomous agents, the saving exists but it is not on the client’s balance sheet. It is on the provider’s income statement.
Signal 2: New surcharges have appeared
Surcharges of 15% to 20% for “AI infrastructure”, “technology licensing” or “digital innovation” are proliferating on outsourcing invoices. This is a double capture of the Dividend: the provider does not share the savings AI generates in its operation and, on top of that, passes on to the client the cost of the tool it used to generate them. Modern contracts should stipulate that AI infrastructure costs are billed as a net pass-through, with no markup, so the client benefits from the hyperscalers’ economies of scale.
Signal 3: The team changed, but the rate didn’t
The documented shortage of specialized AI architecture talent has created a phenomenon I call “Juniors with Copilot”: junior profiles equipped with AI assistance tools (GitHub Copilot, coding assistants, test automation tools) who are presented and billed at a Senior architect rate. The client pays for seniority and gets a prompt operator who shifts technical debt onto the organization. According to market data, rates for specialist AI roles have risen 18%, which encourages title inflation. The contractual answer is to require Key Skill Proofs (KSPs), demonstrable proofs of skill, instead of accepting the job title as a proxy for competence.
Signal 4: No visibility into the production model
If the organization cannot determine, with hard data and per contracted service, what percentage of the work is done by people and what percentage by machines, it has no visibility into its provider’s real production model. Without that visibility, any discussion of price, efficiency or rate adjustment lacks a factual basis. The first step is always to ask the provider for a formal AI inventory. In my experience, when the answer to that request is vague, generic or evasive, the likelihood of finding significant Value Leakage exceeds 80%.
The contract architecture that enables the Dividend
The Captured AI Dividend is neither an accident nor an act of bad faith by the provider. It is the predictable consequence of a contract architecture designed for a production model based on human labor, and never updated to reflect the arrival of intelligent automation.
Within the IT Experts Value Leakage framework, which documents seven dimensions through which outsourcing contracts lose value (see the anatomy of contract Value Leakage), AI amplifies at least three at the same time:
Automation arbitrage (Dimension 6). The provider automates the base of the talent pyramid (L1/L2 support, basic coding, unit testing, monitoring) but contract rates are not adjusted. What used to be a pyramid with a massive base of junior profiles doing repetitive tasks has turned into what the industry calls a “Diamond Model” (for IT services and consulting) or a “Flask Model” (for BPO and high-volume operations). In both cases, AI “ate” the base, but the invoice still reflects the old structure.
Paying for effort, not outcomes (Dimension 5). The pricing model keeps measuring human activity (hours billed, FTEs assigned, tickets processed) when production is already partly artificial. In the age of AI, hour volume is a dead metric as a cost indicator. A provider that bills 100 FTEs but runs on 60 people plus AI agents is generating an undeclared margin equal to the cost of 40 FTEs.
Outdated rates (Dimension 3). Rates negotiated when the contract was signed do not reflect the real deflation in delivery cost. If commoditized services fall 10% to 15% a year and the contract rate stays frozen, the gap builds up quarter after quarter. Over a five-year contract, this can mean a cumulative deviation of 30% to 50% from the current market price.
On a contract of $15M to $25M a year, the combination of these three dimensions can amount to 8% to 15% cumulative leakage, and the figures grow with every quarter that passes without intervention. To estimate the range for your own contract, use the Value Leakage Calculator.
Framework for action: three moves you can execute before quarter-end
Recovering the AI Dividend does not require waiting for the contract to expire or redesigning the agreement overnight. The following three moves can be executed immediately and change the dynamics of the relationship:
Move 1: Request a formal AI inventory
Formally ask your provider to declare the artificial intelligence tools it uses to deliver each contracted service, the percentage of work performed autonomously, and how AI infrastructure costs (processing tokens, GPUs, model licenses) are reflected in billing. Outsourcing contract experts already recommend including “AI tool disclosure” clauses as a standard component of contracts with an automation component. If the provider resists providing this information or gives generic answers, that in itself is a warning sign.
Move 2: Challenge every technology surcharge
AI infrastructure costs (processing tokens, GPUs, platform licenses) should be billed as a net pass-through, with no provider markup. The client is entitled to benefit from the hyperscalers’ economies of scale (AWS, Azure, Google Cloud), from falling memory-chip prices and from the natural deflation curve of compute. When a provider applies a 15% to 20% surcharge on AI infrastructure, it is turning its own operating cost into a profit center at the client’s expense.
Move 3: Require a breakdown of the production model
Ask for transparency on the real composition of delivery: what percentage is human work, what percentage is AI or automation, and what each component actually costs. If the provider cut its assigned human team by 30% thanks to autonomous agents, that saving must show up on the client’s invoice, not only on the provider’s income statement. This breakdown is the factual basis for any later renegotiation: no data, no case.
These three steps are not a formal renegotiation. They are an X-ray of the contract’s current state. And what that X-ray reveals will determine the scope and urgency of the intervention needed. To see the broader picture of where your contract stands, read about the four planes of contract equity.
The necessary evolution: new contract models
Beyond the immediate actions, the industry is converging on new contract structures designed for a world where AI is an integral part of delivery:
Pricing by autonomy level. Segment the price by who performs the task: a premium rate for human intervention in critical decisions, an intermediate rate for human + AI assisted work, and a processing or subscription fee for 100% autonomous tasks. This keeps the client from paying a consultant’s rate for a process run by a machine.
Inverse scalability. Stipulate in the contract that, as transaction volume grows, the unit cost falls automatically and progressively. Since the marginal cost of AI tends toward zero, the provider cannot justify linear growth in its invoice when the client’s volume grows. This mechanism has proven more robust and less prone to disputes than traditional gainsharing.
Productivity give-back clauses. A mandatory rate reduction of 5% to 8% a year, based on the provider’s technology learning curve. If the provider invests in automation (as it should), that investment must translate into progressive contract deflation, not into a unilateral build-up of margin.
AI portability and intellectual sovereignty. Guarantee in the contract that, when it ends, the client owns the optimized prompts, vector databases and automated workflows developed during the relationship. Automated operational knowledge cannot be held hostage on the provider’s platform.
Diagnosis: assess your contract’s maturity level
| Is your IT contract leaking value right now? Start with the free Contract Maturity Test, or book a 30-minute Governance Maturity Assessment to identify your critical leakage points and quantify the value currently being left on the table. No cost. No commitment. Just evidence. |
About the author
William D. López is an independent IT consultant with more than 30 years of experience negotiating and governing technology contracts in Latin America. Founder of IT Experts Consultoría, he has led optimization processes in more than 8 countries, with cost reductions between 9% and 35%. He publishes weekly analysis on Value Leakage, strategic sourcing and AI for CFOs and CIOs on LinkedIn.
