Agentic AI is moving from an additional feature to an integral part of service delivery. In technology outsourcing, it is already influencing how work is carried out, contracts are competed for, prices are calculated and value is demonstrated to customers.
What Is Agentic AI?
An AI agent receives a goal, interprets the context, plans a sequence of tasks, uses tools and takes action within defined boundaries. Unlike a conventional assistant, it does more than generate a response: it intervenes in a process to achieve an outcome.
In IT operations, for example, an agent can detect an incident, consult logs, investigate possible causes, propose a fix and carry out authorised actions. If a decision has a significant impact, it can pause the workflow and request human approval.
This capability requires connecting AI to the company’s systems, protecting data and recording every action. Business value therefore depends not only on the agent, but also on the architecture, governance, observability and oversight surrounding it.
From Advantage to Requirement
Proprietary agentic platforms can still differentiate a provider, but in some large contracts they are also beginning to act as an entry requirement. Customers no longer ask only whether a provider uses AI: they want to know which processes it automates, what productivity it achieves, how it controls risk and how it integrates agents with its systems and teams.
An isolated agent handles a task. A governed platform coordinates multiple agents, limits their permissions, records their activity and produces auditable metrics. That difference can be decisive in an RFP: the process through which a company requests and compares proposals to procure a service.
Industry figures reflect this shift. Accenture ended fiscal 2025 with $42.98 billion in new managed services bookings, $37.64 billion in consulting bookings and $5.9 billion linked to generative AI. NTT DATA, meanwhile, launched Smart AI Agent as an offering combining technology, implementation and services, with a target of $2 billion in related revenue by 2027. In both cases, AI is presented as part of the delivery system, not as a disconnected tool.
AI Revenue Alone Is Not Enough
Growth in AI-related revenue does not mean the market is expanding by the same amount. Some agentic work replaces hours, tasks and contracts previously billed through people-intensive models. Providers must therefore distinguish between new AI revenue, displaced traditional revenue and net growth.
The latest results illustrate this clearly. In its second quarter of fiscal 2027, TCS reported $3.1 billion in annualised AI revenue, more than 10% of revenue, while sequential constant-currency growth was 0.5%. Infosys reported that AI already accounted for 10.5% of its quarterly revenue and was growing 29% quarter on quarter, but reduced its annual growth guidance to 1–2%. Genuine commercial traction in AI can coexist with moderate overall growth.
Risk arises when a company treats every agentic contract as expansion without subtracting the revenue it is replacing. That interpretation can create overcapacity, misguided hiring decisions or investments that fail to pay back. Customers also expect AI productivity gains to translate into a lower price, a faster response or a broader service.
Automation Is Already Delivering Savings
IT operations automation has moved beyond demonstrations. AIOps combines operational data, automation and artificial intelligence to reduce manual tasks and accelerate incident detection and resolution.
Kyndryl reported that its Bridge platform had freed capacity equivalent to more than 13,000 delivery professionals and generated approximately $775 million in annualised savings by the end of fiscal 2025. “Freeing capacity” does not necessarily mean eliminating jobs: some of the time previously spent on operational tasks can be reassigned to other work.
Results of this kind change the benchmark for first- and second-line support. If some alerts, classifications and resolutions are automated, it becomes harder to justify a fee solely by the number of people assigned to the service.
From FTE (Full-Time Equivalent) to Outcomes
Under an FTE contract, customers pay for work capacity equivalent to a number of full-time professionals. One FTE measures labour input equivalent to one full-time workload, not necessarily one individual. In an outcome-based model, pricing is linked to objectives such as incident resolution, system availability, faster response times or savings achieved.
The transition does not eliminate capacity-based contracts. Human availability remains critical in some services, while outcomes can be measured precisely in others. A common approach will be to combine components: professional capacity, platform use, technology consumption, governance and a variable element linked to performance.
The challenge is to agree a baseline, define how savings are attributed and establish who bears the risk when outcomes depend on factors outside the provider’s control. BCG notes that more than 70% of enterprise decision-makers prefer output- or outcome-linked models in certain services, while around 60% of providers still use time-and-materials or fixed-price structures in contracts involving agentic AI. Commercial intent is advancing faster than the ability to measure it and turn it into contracts.
What Each Side Should Review
If You Provide the Service
Classify the portfolio: identify which services will expand, which will transform and which will face declining demand.
Separate pricing components: talent, platform, consumption, integration, supervision and outcomes.
Measure substitution and net growth: do not present all AI revenue as additional business.
Integrate agents with processes and accountable people: automating an isolated task is not the same as redesigning a service.
Demonstrate value through auditable metrics: productivity, quality, savings, availability and risk.
If You Buy the Service
Demand clarity about the decisions each agent can make and those requiring human approval.
Request a record of actions, changes, outcomes and exceptions.
Define how savings are calculated and how much will be reflected in pricing or service improvements.
Assign responsibilities before automating sensitive operations.
Compare business outcomes, not just the number of agents or available features.
A New Service Economy
Agentic AI does more than add another outsourcing category: it changes the delivery mechanism. It can replace some revenue based on human effort, make operational savings visible and shift the commercial discussion from hours worked to outcomes achieved.
Accumulating tools without reviewing contracts, costs and metrics does not transform a business. The advantage emerges when agents, people and processes operate as a system, with clear responsibilities and verifiable outcomes. The question then is no longer whether a provider uses AI, but how it shares work, risk and value with its customer.
In B2B marketing and sales, this means reviewing which tools support the service and how their contribution is measured. Martech stack consulting can identify duplication and align platforms with the team’s processes; data and metrics analysis helps connect activity with business objectives and verifiable outcomes. If you work with a B2B marketing agency in Barcelona, it is also worth agreeing which part of the value comes from the tools and how it will be verified.
Frequently Asked Questions
Will Agentic AI Replace Traditional Outsourcing?
It will replace some tasks and change certain contracts, but it will also open up services in integration, governance, security and specialisation. The effect will depend on the process and the provider’s ability to redesign its offering.
Will FTE-Based Contracts Disappear?
Not necessarily. They will remain useful when customers need stable capacity or human availability. Their importance may decline in repetitive processes with easily measurable outcomes.
How Are Savings Verified?
Measure cost, time, quality and incidents before automation, maintain a comparable baseline and agree what proportion of the improvement can be attributed to the new model.
What Distinguishes a Platform From an Isolated Agent?
A platform coordinates agents, permissions, data, supervision and metrics. It enables a complete, traceable service, whereas an isolated agent generally performs a specific function.








