GBS must own AI outcomes; here’s how it can claim value

This HFS Point of View is for GBS leaders, global process owners, COOs, and transformation executives converting AI accountability into authority over applications, governance, and decision rights.

AI is changing the unit of value from transactions processed to business outcomes orchestrated, and Wall Street has already repriced the service providers that missed the shift: pure-play IT/BPO services firms trade at roughly 1.5x sales, while the S&P 500 averages 3.7x and AI-native companies around 46x (see Exhibit 1). Global business services (GBS) leaders face the same repricing internally, at a lag, as budget scrutiny and functions pulling work back.

GBS is uniquely positioned to own the shift because it already spans enterprise processes, data flows, and functional boundaries. At the same time, it cannot be held accountable for AI outcomes without authority over the applications, governance, and decision rights that determine them. HFS’s view is that GBS must claim the space between business and IT within two years, or the functions will reclaim the work.

If your GBS organization had a market cap, what would its price-to-sales ratio (PSR) be? The providers listed in Exhibit 1 are not failing businesses; they posted mid-single-digit growth and stable margins. The market has simply stopped paying for revenue that scales with headcount, and GBS runs a similar linear economic model, only with internal customers.

Exhibit 1: Wall Street has already repriced linear, people-driven services; internal GBS and GCC organizations could be next, at a lag

Note: The ~1.5x IT services average covers the seven pure-play firms and excludes IBM, whose revenue mix spans hardware and software; Nvidia appears as an AI-native reference point.
Source: HFS Research analysis based on publicly announced statistics, June 2026

A 2010 operating model cannot capture 2026 AI value

The repricing is the symptom. The cause is structural, as HFS President Saurabh Gupta framed it for the room:

We are in 2026, trying to capture value from AI. But we are stuck in a 2010 operating model, especially for GBS.

— Saurabh Gupta, President, HFS Research

To identify the challenges enterprise face and the changes required to capture business value from AI at scale, HFS convened an executive roundtable with IBM in New York in July 2026, attended by senior operations, finance, and transformation leaders from banking, property and casualty insurance, mortgage servicing, pharmaceuticals, medical devices, beauty, consumer packaged goods, automotive, media, and industrial services (pictured below).

The consensus was that the main problem is an accountability and authority gap, not a technology gap. GBS is asked to deliver AI outcomes without any control over the data, platforms, or decision rights needed to deliver them. Until the middle layer between business and IT gets real authority over applications, context, and P&L, agentic AI will keep producing pilots that never reach production.

Senior enterprise leaders debating the GBS agenda at the HFS-IBM roundtable in New York

Source: HFS Research and IBM, 2026

Four signals show the labor-arbitrage GBS model breaking under agentic AI

Industry leaders across the room surfaced similar structural failures. Each pointed back to an operating model built for labor arbitrage, now asked to deliver autonomous execution.

• Personal productivity is real, and it is not reaching P&L. Nearly every delegate had rolled out a copilot, and leadership could see the value in meeting notes and summarization. Yet few delegates could attribute the benefits directly to a financial statement. HFS’s field research shows that fewer than 10% of POCs reach production, leaving most organizations funding experiments that never touch P&L.
• The blocker is authority, and everyone is measuring accountability instead. Global process owners are accountable but cannot act on their own. They negotiate for configuration capacity and wait for approval, while a systems owner in one country can veto a model writing back to a source system and stall a program for a quarter. Small battles compound into a stalled operating model.
• Cost did not become hygiene. It changed shape, and nobody is governing the new shape. A 2026 HFS study of 510 Global 2000 enterprise leaders found that nearly 68% disagree that back-office cost reduction has been demoted to hygiene. One delegate burned through a two-year token commitment in four months, and another cited a technology firm capping developer token budgets at $1,500 a month. GBS has 20 years of governance muscle around FTE costs, but the cost line has moved to tokens, where few enterprises have a consumption model, caps, or chargeback logic. That makes it a GBS problem: GBS is the function that has historically governed the cost of shared processes, and if it does not claim token governance, AI costs land on functional budgets with no one held accountable.
• Reverse migration has started, and it is a verdict rather than a compliment. Several delegates described functions asking for centralized work back now that the digitization is done. Delegates traced the cause to trust: functions reclaim work when they believe GBS cannot move fast enough or deliver the AI outcomes they need, so the migration is a verdict on GBS’s credibility. The counterargument was decisive: functional teams can automate up to 90%, but the last mile lives at the intersections between functions, where the exceptions sit. Re-verticalizing guarantees you never close it.

Our global process owners are accountable, but they cannot act. We keep mixing up accountability with authority.

— GBS leader at a global industrial services company

Uneven maturity, a missing mandate, and low trust keep AI value from scaling

The three threads build on one another: maturity is uneven, a missing mandate stops the mature pockets from scaling, and even a granted mandate fails without adoption and trust.

Pockets of maturity do not add up to organizational readiness. One automotive leader rated his platforms at 7 or 8 out of 10, but his organization at 1. Islands of excellence built on a disconnected operating model do not compound into an end-to-end capability.

The mandate is political, and few CEOs have granted it. The authority gap described above is the symptom. The cause is that cross-functional decision rights are conferred from the top, and participants consistently described that grant as absent. Agentic work cuts across silos by design, so GBS needs a mandate from the CEO rather than permissions negotiated function by function.

Change management and cultural readiness weigh as much as technical readiness. The leaders furthest along spent more effort on adoption, trust, and narrative than on the technology. Process debt can be engineered away; a workforce that does not trust an agent to act cannot.

If work is going back to the functions, it is not because we succeeded. It is because we failed to earn their trust

— Shared services transformation leader at a global consumer products enterprise

Three structural patterns separated the few organizations from the rest

Three patterns recurred across the more advanced examples discussed, and all three are structural rather than technical.

• Reporting line: Where GBS genuinely owns transformation, its leader reports to the CEO rather than the CIO or CFO. That single line on the organizational chart converts a service provider into a peer.
• Application ownership: One delegate moved the IT platforms into GBS, giving their HR leader ownership over the people budget, the transformation budget, and the Workday platform. Applications, data, and process converge by design rather than by committee.
• Sequence: IBM applied the same discipline internally through its Client Zero program. It reengineered workflows across HR, IT, finance, procurement, supply chain, sales, and customer support for roughly 270,000 employees, eliminating steps before automating rather than automating the complexity it already had. Three years in, the firm delivered $4.5 billion in annualized productivity savings, more than double its original $2 billion target, with HR operating costs down 40% and an AI assistant resolving 94% of roughly 16 million annual employee inquiries without human intervention.

You do not start with technology. You start with where the workflows are broken and where the connective tissue between functions is missing. GBS leaders created shared services by co-locating functions, but they never connected finance to supply chain to marketing to operations. IBM Business Orchestration Services was built to close that exact gap, and when somebody comes to me for simple labor arbitrage, I am learning to say no.

— Yogendra (Yogi) Goyal, Global Managing Partner, Business Orchestration Services, IBM Consulting

IBM’s client evidence points the same way. At a multinational consumer goods company, a single-accountability model with IBM Consulting reduced the cost of finance from 100 to 62 basis points between 2023 and 2025. Automation reached 70% to 90% across high-volume transactions, cash application, SAP payments, and intercompany invoicing, and reconciliation productivity improved 50% in more than 60 markets. The larger point is positional: finance moved from a transactional back office toward business partnering, a shift that GBS must now make at an enterprise scale.

GBS already holds three of the four capabilities the middle layer needs

Business sets intent; IT enables and executes it. GBS work sits between them and belongs to neither, which is why so much of it stalls (see Exhibit 2). One delegate launched an agentic finance pilot with finance on one side, IT on the other, and the AI expertise missing from both, so they went outside for it. HFS describes this layer as the OneOffice operating model: business intent and IT execution sharing one accountability and one scorecard, with AI agents working inside a single governance loop.

Exhibit 2: The middle layer between business intent and IT execution is where GBS must operate, which is HFS’s OneOffice model

Source: HFS Research, 2026

HFS’s view is that the middle layer needs four capabilities. GBS already has three and needs to build on the fourth one.

• Enterprise context: Frontier models are heading toward utility economics. However, agents fail to scale because nobody has trained them on the enterprise’s own context, and no other function matches GBS’s simultaneous access to customer, employee, vendor, product, and financial data flows.
• A portfolio logic beyond centralize versus localize: The real question is whether the work is AI-executable or AI-augmented. Transaction-centric work is largely AI-executable, run by a provider and governed by GBS through SLAs and audit trails. Decision-centric work is AI-augmented, with GBS retaining human judgment on decisions the model should not make alone.
• Governance most enterprises lack: GBS can own token economics, model selection, audit trails, override rights, and a single intake process for AI use cases. Several delegates described governance vacuums, where teams experiment with no visibility into consumption or control.
• A talent profile that no longer exists: Most GBS organizations do not have enough people to translate a business need into a solution design without being asked a thousand questions on requirements. The most successful programs were led by functional people who made themselves technical, not technologists learning the process.

Four moves convert GBS accountability into authority within a year

Each move transfers a decision right into the middle layer, and none waits for better technology.

• Take the applications or accept that you are a recommender. Move the platforms that run your processes into GBS and create single accountability for people, process, and technology in at least one function. This converts accountability into authority.
• Sequence elimination before automation. Run one end-to-end flow through eliminate, automate, simplify, and elevate before deploying a single agent on it. Agents do not fix broken processes; they compound the problem.
• Claim token governance this quarter. Build the consumption model, caps, chargeback logic, and model selection policy before finance discovers the bill. GBS spent 20 years governing cost per FTE; it now needs to apply that same muscle.
• Rebuild the process owner profile, then prove the outcome. Audit global process owners against system design thinking rather than continuous improvement credentials and deliver one flagship end-to-end process with a P&L impact that the CFO will recognize. Few outside the function can say what GBS contributes to the bottom line; the price-to-sales problem in Exhibit 1 is partly a narrative problem.

The Bottom Line: The technology is ready; the operating model is not. GBS has roughly two years to claim the middle layer between business and IT before someone else does.

Almost every blocker raised at the three-hour roundtable was structural: fragmented ownership, applications in the wrong function, decision rights that exist nowhere, process debt, an obsolete process owner profile, and no governance over the new cost line. All can be solved through authority, which is the one thing GBS keeps asking for rather than taking.

This creates one test for every GBS investment: does it move authority into the middle layer, and does it move the enterprise up the HFS AI Trust Curve from model confidence toward decision reliance? If not, you’re just buying a faster version of 2010.