1. The uncomfortable reality
Operations, including service providers, global capability centers, and retained enterprise teams, were traditionally built around a simple proposition: take repetitive work, perform it efficiently and at scale, and organize people around the capacity required to do it. For years, the unit of value was often the hour, the seat, the FTE, or the cost base, even as operations teams delivered value through process improvement and automation.
AI now fundamentally changes that equation. A machine can increasingly absorb work that once required people to read documents, answer questions, process transactions or manage queues. For enterprises, this is a productivity story. For service providers, it is also a business model story because the unit AI displaces — the human hour — is often the same unit they have historically invoiced. For GCCs and retained operations, the pressure shows up through cost structures, productivity expectations, workforce models, and the need to prove value beyond activity.
This creates an uncomfortable reality for operations leaders: AI is being deployed to absorb the very work that has historically justified headcount, cost, revenue, or scale — at a pace not imaginable earlier. What many may not yet fully see is that the advantage will not come from AI itself, which is fast becoming accessible to all, but from what surrounds it: the data, domain context, and ability to execute with trust.
2. The question operations now must answer
The question, then, becomes how to redefine the value proposition. Take a customer-service interaction, a reconciliation, a claims intake, a document review, or an underwriting submission. For years, more work meant more people, more cost and, for service providers, often more revenue. As AI absorbs more of this activity, the old relationship between work volume and value comes under pressure.
The instinctive response is commercial or structural: Move from FTEs to transactions. From effort to outcomes. From capacity to gainshare or enterprise impact. These shifts matter. But is that the full answer? If AI increasingly performs the work, what exactly is operations being valued for — execution capacity, or the responsibility for what gets executed?
3. Changing the commercial model is not transformation
Moving from FTE-based pricing to transactions, outcomes, or gainshare is necessary for service providers. For GCCs and retained operations, the parallel shift is from activity measures to enterprise value. But changing the commercial or measurement unit does not, by itself, change the operating model.
Consider a provider paid per claim processed. It automates half the activity, but keeps the same handoffs, controls, management layers, and exception paths. Productivity improves, and the client reasonably expects the unit price to fall. The process has become cheaper, not fundamentally different. The same logic applies inside enterprise operations if AI reduces effort but does not change how work flows, decisions are made, or accountability is carried.
The deeper transformation begins when the workflow itself is redesigned. McKinsey’s State of AI 2025 research found that workflow redesign had the greatest effect on realizing EBIT impact from AI. BCG’s 2026 analysis of more than 600 US public companies makes a similar point: only 6% qualified as AI adoption leaders, but those leaders outperformed peers by 9 percentage points in industry-adjusted shareholder returns, driven by revenue growth and margin expansion rather than investor hype. The lesson is simple: changing the pricing or measurement model may improve alignment, but it is not transformation unless the work itself changes.
4. Where value moves: from work to results to decisions
Three forms of value now coexist in a modern operations model: work, results and decisions. Work is the traditional ledger of hours, FTEs and cost per unit - and the one under greatest attack from AI. Results shift attention to what gets delivered: a resolution, a transaction, a claim-handling step. Decisions are where value becomes harder to automate.
This is also where the answer is not obvious, and where domain knowledge matters most: what should be automated, what needs review, what should never happen without intervention, the classic human-in-the-loop vs. human-on-the-loop decision, how uncertainty should be escalated, and who remains accountable. These are not just technology choices; they decide who carries the risk, how customers are treated, and whether the model holds when exceptions rise.
5. P&C insurance examples: where decisions matter
P&C claims and commercial underwriting bring this to life. Both involve high-volume information processing, but both also contain moments where ambiguity, risk, and consequence require judgment.
Claims. After a major weather event, AI can extract details, organize documents, identify gaps, and recommend a triage path. But some claims will not fit the pattern due to issues such as a coverage question, conflicting evidence, fraud indicator, vulnerable customer, or severe loss. Those cases need a different path because the consequence of getting them wrong is higher.
Operations value is not just in processing claims faster; it is in the expertise to separate routine cases from consequential ones, and to manage the timing, quality, evidence, and data trail of escalation.
Commercial underwriting. AI can extract information, flag gaps and summarize risk. But where the exposure is unusual, information incomplete or the risk close to the edge of appetite, judgment still matters. AI can support the analysis; the underwriter remains accountable for the material risk decision.
In both insurance examples, the principle is the same: automate the predictable, make uncertainty visible, and apply judgment where it matters most. This lesson equally applies for industries like banking, healthcare, or finance operations
<span style="color:#FB4E0B">6. Building the operating model around AI</span>
Getting this right starts with three principles:
Redesign the work before the workforce. Start with what the work should look like, not how many roles can go: what AI can own, what it can support, and what still needs experienced judgment.
Preserve the learning. If AI removes all routine work from human experience, the organization may eventually lack people with the domain experience to rise to the occasion when AI fails.
Make trust operational. Trust cannot sit on a slide under change management. It must be built into how the work runs, so employees, clients, and stakeholders can see how any decision is made, examined, or challenged.
Responsibility then expands in a natural progression:
Automate the task - AI handles defined, repeatable work where rules are clear and risk is controlled.
Augment the person - AI prepares and recommends; a person applies judgment and stays accountable.
Orchestrate the workflow - agents, systems and teams pass information, evidence and work across functions and towers, instead of becoming new silos.
Own the outcome - operations take responsibility for how the process performs, including exceptions, controls and recovery.
As responsibility expands, measurement and accountability must follow. Contracts and scorecards cannot stay anchored to effort, activity and cost; they should reflect resolution quality, exception handling, compliance, leakage, customer outcomes and auditability — with explicit clarity on what AI decides, what people review, what operations own, what the business or client retains, and how value is shared when performance improves.
7. Sell less time. Own more responsibility. Earn the value
Operations do not need to become AI vendors to remain relevant. They need to become more adept at owning the work around AI — and the responsibility that comes with it.
As models and agents become available to almost everyone, AI alone will not differentiate. What is harder to replicate is domain understanding — knowing how a coverage nuance, a regulatory obligation or a client’s risk appetite should shape a decision — built by operating inside the workflow every day. It is what allows operations to own decisions, not just tasks.
But data, context and AI do not create value on their own. They need execution that clients and regulators can trust: deciding where AI acts, when people step in, how systems coordinate, and how every outcome is explained and owned. Put simply:
(Data + Context + AI) × Trusted Execution = Enduring Operations Value
Data gives AI something to work with, context tells it what matters, and AI brings speed and scale. Trusted execution is the multiplier — without it, the sum counts for very little. It is also where responsibility sits, and where operations are best placed to lead.
To truly transcend the hour, operations must be known not for the time they supply, but for the judgment they bring, the outcomes they own, and the risk they are willing to carry. That is when the hour stops being the measure — and responsibility becomes the real source of value.
Disclaimer: This paper reflects the author’s personal perspective and is shared as a thought piece. It does not represent formal advice, an official position of any organization, or endorsement by third-party sources referenced. External research and data are included for contextual support only; readers should refer to the original sources for full details.