Recently, I had a conversation with a senior executive in a large digital consumer platform in the GCC about AI-enabled managed services. What caught my attention was not the technology itself, but the commercial question underneath it. Organizations are introducing AI across service desks, application support, customer operations and back-office processes. Providers are investing in automation, copilots and agents. Everyone expects lower cost, faster delivery and better service. Yet one question remains surprisingly unclear: if AI makes the provider more productive, who owns the value created?
The Provider Becomes More Efficient
The traditional managed-services model was built around familiar commercial units: people, hours, tickets, devices, transactions and service levels. Those units made sense when delivery costs were largely driven by human effort. More work required more people, and more people required more budget.
AI changes that relationship. A provider may now resolve routine incidents, generate code, analyse documents or support service requests with fewer people involved. The client may receive the same service, or potentially a better one, while the cost of delivery changes underneath. Providers are also carrying genuine new costs: platforms, integration, security, model operations, evaluation and governance. This does not automatically mean that the client should receive an immediate price reduction. It does mean that the old economic logic deserves to be revisited.
The Contract Still Prices Yesterday’s Delivery Model
This is where a familiar execution problem appears. The capability changes, but the commercial model does not. A provider introduces AI, productivity improves and some work is absorbed without additional headcount. Yet the contract may still be priced around the workforce, volumes and assumptions of the previous delivery model.
On paper, the service continues to perform. In practice, the economics have moved. Recent BCG research found that nearly 60% of enterprises have not yet seen measurable total-cost improvement in deals that include agentic AI. The research also found strong buyer interest in output- and outcome-linked models, while many providers still use time-and-materials or fixed-price structures. The technology may be new, but the contract may still be old.
Tokens Are Not the Real Question
Token consumption has become a prominent concern in AI discussions, and rightly so. Agentic workflows can involve models, prompts, retrieval, tool calls, retries, orchestration and human escalation. Their cost is not always visible through one simple invoice.
But the client should not need to manage tokens as its primary commercial unit. The client needs predictable pricing, service quality and accountability for outcomes. The provider, by contrast, needs visibility into its internal delivery economics in order to understand and control the cost of the service it is offering. A fixed price may give the client predictability, while internal cost telemetry gives the provider control. Both are needed, but they serve different purposes.
The Productivity Dividend Needs a Baseline
The challenge with many AI productivity claims is not that they are necessarily wrong. It is that they are difficult to prove. A provider may say that AI reduces effort by 30%, but compared with what? Which workflows changed? Did quality improve or decline? Did rework increase? Did human escalation fall? Did the client receive a better service, or only the same service at a lower internal cost to the provider?
Without a credible baseline, productivity becomes a presentation rather than a commercial fact. That baseline should include more than cost. It should cover service quality, resolution time, rework, customer experience, incidents, human intervention and the business outcome the service is meant to support. Otherwise, an organization may save effort in one part of the operation while quietly creating more complexity somewhere else.
Where Value Starts to Drift
Execution drift begins when a capability is deployed but the organization does not redesign the accountability around it. One team may own the technology, another the service contract and a third the business value expected from the investment. The provider owns the delivery model, finance owns the budget and procurement owns the commercial terms. Everyone owns part of the picture, but nobody owns the productivity dividend.
That is where value starts to leak. The organization may have deployed the technology successfully and the provider may have improved its delivery model. Yet the connection between technical capability, service performance and economic value remains blurred.
Turning AI Productivity into Shared Value
The answer is not to replace every managed-services contract with token-based pricing. That would simply transfer technical uncertainty from the provider to the client.
The answer is to create a clear value-realisation mechanism inside the contract.
First, both parties need an agreed baseline before AI changes delivery. This should cover cost, service volumes, resolution times, quality, rework, escalations and customer experience. Without that baseline, neither party can distinguish genuine productivity from normal operational variation.
Second, the contract should define which AI-enabled improvements matter. A lower number of tickets is not necessarily valuable if customer satisfaction falls. Faster code generation is not valuable if defects and rework increase. The measures need to reflect the business result, not only the provider’s internal efficiency.
Third, pricing should combine predictability with shared accountability. The client can retain a stable base price for an agreed service level, while measurable and validated productivity gains can be shared through a gainshare, outcome-based component or periodic price reset. The provider retains an incentive to invest. The client participates in the value created.
Finally, the arrangement needs regular governance. AI models, usage patterns, service complexity and delivery methods will change. A quarterly or semi-annual review of consumption, quality, performance and realised value prevents the contract from becoming disconnected from operational reality.
This is less about negotiating a better AI clause.
It is about creating a commercial model that makes value visible, measurable and shared.
The Most Important Question in the Next Contract
The next generation of AI-enabled service contracts will need to define the outcome being bought, the baseline against which improvement is measured, the quality controls that protect the client and the mechanism through which productivity gains are shared.
The question is not whether AI will make providers more efficient. It already is.
The question is whether clients and providers will build contracts that turn that efficiency into measurable value for both sides.
#ExecutionDrift #AITransformation #ManagedServices #ValueRealisation #AIGovernance