Most large organisations have spent the past three years running AI pilots. They have deployed productivity tools, launched automation programmes, and identified use cases. A reasonable number have achieved genuine results within the boundaries of the project. Very few have made AI work at enterprise scale. The problem is rarely the technology. The problem is the organisation around it.
AI does not scale through pilots. It scales through shared platforms, disciplined governance, and consistent execution across the enterprise. That requires an organisational layer that spans functions, owns shared capabilities, enforces standards, and integrates intelligent workflows into how work actually gets done. In most large organisations, that kind of cross-enterprise execution capability already exists. It is called Global Business Services.
GBS is not, at first glance, an obvious candidate for the role of enterprise AI backbone. It was built for labour efficiency, not technological leadership. Its identity has been transactional, its status in most organisations more operational than strategic. But it has structural advantages that no other part of the enterprise possesses, and those advantages matter more in an AI-enabled world than they did in the world for which GBS was originally designed.
This article argues that GBS has the potential to evolve from a service provider into something more structurally significant: the backbone that enables an intelligent enterprise to execute, govern, and improve. Realising that potential requires deliberate redesign. It also requires leadership to make a choice about GBS's fundamental purpose.
Built for a different era
GBS emerged in the 1990s as a rational response to a labour-driven operating model. The design logic was straightforward: consolidate common processes across functions and geographies, standardise them, and execute using lower-cost labour in centralised or offshore locations. Finance, HR, procurement, IT support and legal operations were the natural candidates. The result was a new class of organisation, neither purely internal nor fully outsourced, with global process ownership structures that had not previously existed.
The model delivered real value. P&G's GBS operation, one of the most studied in the world, now delivers 85 standardised services across more than 80 countries with a customer satisfaction score of 8.9 out of 10.1 The Hackett Group estimates that benchmark-performing GBS organisations achieve administrative cost levels 40 per cent below industry average.2 Over three decades, GBS has moved from experiment to mainstream. Today, approximately 80 per cent of Fortune 500 companies operate some form of shared services or GBS model.
But the economic logic that justified the original model has been eroding for some time. Labour cost differentials between markets have compressed. Offshoring has become ubiquitous, reducing its strategic distinctiveness. Complexity has grown as processes have become more interconnected and exceptions have become harder to standardise away. The Hackett Group notes that GBS organisations are increasingly being asked to do more than run processes: to manage data, support digital transformation, and provide analytical insight. The original mandate and the actual demand have diverged. AI is making that divergence impossible to ignore.
What AI changes
AI does not just make GBS processes faster. It automates the work GBS was originally hired to do. Accenture estimates that up to 80 per cent of finance department transactional work is automatable with current technology.3 Gartner predicts that 90 per cent of finance functions will deploy at least one AI-enabled solution by 2026.4 Invoice processing, accounts reconciliation, payroll calculation, query resolution, report generation: these are the activities that have defined GBS workloads for thirty years, and they are all on the automation curve.
The same pattern applies in HR operations. Talent acquisition screening, onboarding workflows, benefits administration, case management: McKinsey estimates that more than 60 per cent of HR service centre activities could be automated with currently available tools.5 In procurement, AI is enabling autonomous supplier matching, contract analysis and spend classification at scale. The routine execution that justified the original labour arbitrage model is, in meaningful part, on its way to automation.
This does not make GBS redundant. It makes it available. The same cross-enterprise infrastructure that handles transactional volume can, with the right redesign, handle something more valuable: the governance, integration and stewardship of AI-enabled processes across the organisation. As processing automates, the scarce capabilities become enterprise data stewardship, platform ownership, model governance, exception judgement, and the continuous redesign of cross-functional workflows.
These capabilities are not the same as IT infrastructure. They are not the same as corporate audit. They sit at the intersection of process, data, risk and operations. That is exactly where GBS already operates. The difference is the role GBS is asked to play and the organisational authority it is given to play it.
As processing automates, the scarce capabilities become governance, stewardship and architectural control. GBS already sits where these things live.
From service provider to enterprise spine
An enterprise spine is not a cost centre with a better brand. It is a structural layer that owns shared platforms, governs enterprise data standards, integrates AI into operational workflows, sets process architecture, and ensures risk oversight across functions. It is the execution backbone through which intelligent systems operate, and through which their outputs are monitored, corrected and improved. It is accountable not just for whether work is done, but for whether the systems doing it are working as intended.
GBS is the most credible candidate to become that backbone. No other organisational unit has the same combination of structural properties: it already spans functions, already standardises processes, already interacts with data flows across the organisation, and already holds the global process ownership constructs that make enterprise-wide coordination possible. This is the strategic territory of Global Business Services and Enterprise Platforms. IT owns infrastructure, not business processes. Corporate functions own policy, not cross-functional execution. Business units own performance, not shared capability. GBS sits at the intersection of all three.
Unilever's GBS is instructive. It has accumulated more than €1 billion in operational savings through centralised procurement and automation of shared services.6 More significantly, it has repositioned itself not just as a delivery engine but as an enabler of enterprise-wide capability. Its procurement GBS coordinates category management, supplier risk analysis, and spend intelligence across more than 190 countries.
The gap between today's GBS and tomorrow's enterprise spine is not primarily technical. Most GBS organisations already have the data access, the process reach, and the global infrastructure. What is missing is mandate and architectural authority: the organisational legitimacy to design how AI-embedded processes work, set the standards that govern them, and be held accountable when they fail. That is a leadership decision, not a technology implementation.
Governance and proximity to the centre
When AI is embedded in operations, governance failures do not stay contained. A flawed credit decision model in finance, an erroneous HR classification algorithm, a biased procurement scoring tool: each can propagate across thousands of decisions before a human notices. The speed at which AI executes is precisely the speed at which poorly governed AI causes harm. This is a different risk profile from the world GBS was designed to manage.
In a labour-driven operating model, process mistakes are local and correctable. A wrong entry is reversed. A mis-sorted invoice is re-processed. In an AI-enabled model, errors are systemic and potentially reputational. The EU AI Act, now in active enforcement, requires documented human oversight for high-risk AI applications in hiring, credit, access to services, and law enforcement.7 It requires audit trails, explainability, and identifiable accountability. Regulators are not satisfied by the explanation that governance was delegated to a peripheral shared services unit. Design authority and risk integration must sit close to where decision logic is designed and deployed.
This changes the organisational argument for GBS's positioning. The traditional case for centralising GBS at a distance from the corporate centre was efficiency: lower cost locations, separation of policy from execution, independence from business unit politics. The emerging case for bringing it closer is risk: accountability for AI-embedded process decisions needs to be traceable, and traceability requires proximity to the governance structures of the enterprise.
J&J Global Services has navigated this deliberately. It now deploys JAIDA, an enterprise AI assistant, across 138,000 employees in 60 countries, supporting HR, finance, procurement and IT operations. The deployment was not managed as a technology project. It was structured as a governance redesign, with GBS holding accountability for data quality, model monitoring, and exception escalation. That structural choice—integrating delivery and governance into the same organisational unit—is what made enterprise-scale AI deployment tractable.8
If GBS is to become the enterprise spine, it cannot remain structurally peripheral. It must operate with delegated enterprise-wide authority, with a direct reporting line to the CFO or COO, and with explicit accountability for the governance of AI-enabled operations. That is not how most GBS organisations are currently structured. It is, however, how the most advanced ones are beginning to reposition.
AI governance failures scale at machine speed. Proximity to the centre is a structural requirement, not an organisational preference.
Rethinking the outsourcing question
The outsourcing and offshoring model that sits beneath much of GBS was built on a straightforward economic rationale: the work was routine, skills were available at lower cost in offshore locations, and the value destroyed by distance and coordination overhead was acceptable relative to the savings achieved. For three decades, this logic held. Labour cost was the primary variable, and offshore centres in India, Eastern Europe, the Philippines and Latin America delivered genuine efficiency gains.
AI is changing the primary variable. When processing automates, what remains is governance, integration and architectural control. These are forms of work that are fundamentally harder to govern from a distance, harder to enforce at arm's length, and most consequential when they go wrong. Industry analysis from the Deloitte GBS Survey 2025 suggests that by 2027, technology-driven efficiency will largely replace traditional labour arbitrage as the primary cost lever in GBS. The arbitrage logic is not disappearing overnight. But its relative importance is declining, and the work that is growing in importance (AI governance, process architecture, data stewardship) demands a different organisational model.
This does not mean bringing everything back onshore. It means disaggregating what was bundled together for efficiency reasons and reconsidering each element. Volume processing, where AI can handle scale, may be best managed through technology platforms regardless of location. Specialist capabilities, where skills are scarce and model quality matters, may need to draw on global talent regardless of location. Governance and architectural control, where proximity and accountability matter most, may need to sit closer to the centre. These are three different questions that the old model answered as one.
The emerging model is one of capability provision under enterprise architectural control. External providers can deliver technology platforms, specialised skills and operational capacity. The enterprise retains design authority, data standards, accountability for outcomes, and the structural ability to govern what it has built. The spine stays internal. Specific capabilities can be external. That distinction is strategic, and most outsourcing contracts written before 2022 do not reflect it.
What GBS must become
The transition from service provider to enterprise spine is not a single event. It requires changes across the work GBS does, the capabilities it holds, and the organisational authority it carries. Some of what has defined GBS will diminish. Some will evolve into something more valuable. And some things will need to be genuinely new.
What diminishes
The purely transactional identity of GBS, defined by volume metrics, cost per transaction and labour utilisation, becomes a diminishing part of the value proposition as automation absorbs the work. Labour arbitrage as the primary strategic driver ceases to be the right frame. The structural peripherality that allowed GBS to operate at arm's length from corporate governance, tolerated when the risks were operational, is no longer compatible with the decisions it is now accountable for.
What evolves
Global process ownership becomes enterprise process architecture: not just running processes but designing how they work, how they embed AI, how exceptions are handled, and how outcomes are measured. Service management becomes platform orchestration, managing AI tools, data pipelines and automated workflows as integrated systems rather than isolated services. Compliance management becomes AI-enabled governance, with real-time monitoring, model auditability, and audit trails that meet regulatory requirements without relying on manual review.
What is genuinely new
Enterprise data stewardship (owning the quality, lineage and governance of the data on which AI decisions are made) is a new capability that GBS is structurally well-placed to hold because it already touches the data flows across every major function. AI workflow integration, connecting AI tools to live operational processes in ways that are safe, auditable and continuously improvable, is a new design discipline. And risk and ethics coordination across AI deployments, ensuring that embedded decision logic is monitored, challenged and corrected, is a cross-functional accountability that needs a structural home. GBS is the most logical candidate.
Building these capabilities requires investment in people as much as technology. The GBS workforce of the future needs AI literacy across all roles, specialist expertise in process automation, data engineering, and model monitoring, and leadership capable of operating at the interface between technology and business strategy. The Deloitte GBS Survey 2025 identifies talent transformation as the single biggest challenge facing GBS leaders: 67 per cent cite the inability to attract and retain digital talent as a top constraint.9 That is a hiring problem and a positioning problem. GBS will not attract the talent it needs if it is still perceived as a back-office function.
GBS will not attract the talent it needs if it is still perceived as a back-office function.
Making it happen: where to start
The transition does not require a single transformation programme. It requires a sequence of deliberate choices that, over time, change what GBS is structurally responsible for. Most of those choices are not technology decisions. They are decisions about mandate, accountability, and organisational design.
Clarify the mandate. The most common reason GBS fails to evolve is not a lack of capability. It is ambiguity about the scope and focus of its role. If GBS is mandated only to deliver services at minimum cost, it will optimise for exactly that. If it is mandated to govern enterprise-wide AI-enabled operations, it will develop the capability to do so. The mandate needs to be explicit, publicly stated, and backed by the authority to act across functions.
Reposition structurally. Enterprise spine status requires direct access to the corporate centre. For most organisations, this means elevating the GBS leader into the C-suite, establishing a direct reporting line to the CFO or COO, and making GBS a formal participant in enterprise risk and technology governance. It also means giving GBS explicit design authority over shared platforms and process architecture, not just execution responsibility.
Build the new capabilities deliberately. Data stewardship, process architecture and AI governance will not grow organically from a transactional base. They need to be built through targeted hiring, structured capability programmes, and partnerships with the functions that GBS serves. The organisations doing this well, including J&J, P&G and Unilever, have not waited for AI deployment to create demand. They have built the capability in advance and used it to lead enterprise AI adoption.
Reform outsourcing contracts. Most existing outsourcing arrangements were structured for a labour-driven model. They need to be renegotiated to reflect the new division of responsibility: external providers deliver capability and technology; the enterprise retains architectural control and governance accountability. Contracts that cede design authority to providers, or that make it structurally difficult to monitor AI-embedded processes, represent a governance risk that compounds over time.
The strategic choice
GBS has three plausible futures.
The first is continuation: a high-efficiency transactional utility, increasingly automated, increasingly commoditised, and increasingly dependent on technology vendors to remain relevant. It survives. It does not lead.
The second is adaptation: an automation support function, helping business units deploy AI tools, running centres of excellence, providing digital expertise on request. It adds value. It does not transform the enterprise.
The third option—enterprise spine— requires a deliberate choice by leadership. It means redesigning GBS from the outside in, starting with what an intelligent enterprise needs to execute safely, consistently and at scale, and working backwards to the organisational structure, capabilities and authority that are required. It means accepting that GBS must be elevated, not just improved, and that the governance of AI-embedded operations is too important to be left at the periphery of the enterprise.
The case for making that choice is not primarily about GBS. It is about how AI creates value at enterprise scale. AI will create an execution engine inside the enterprise, automating decisions, orchestrating workflows, and managing data at a speed and volume that no human organisation can match manually. The question is whether that engine is fragmented across business units and technology silos, or governed through a coherent, accountable structural layer with the authority to set standards, enforce them, and course-correct when things go wrong.
GBS is the most credible candidate for that role. No other organisational unit has the cross-functional reach, the process ownership infrastructure, and the execution capability to fill it. But the role will not be claimed by default. It will be claimed by the GBS leaders and their executive sponsors who make the argument clearly, build the capability deliberately, and are willing to accept the accountability that comes with it.
That is the strategic choice. And the window to make it is not indefinitely open. Explore how we work with GBS leaders on this transition.
References
- P&G Global Business Services (2024). Service Delivery Model and Performance Data. SSON / McKinsey & Company. www.ssonetwork.com/global-business-services/articles/creating-a-brand-new-strategy-for-shared-services
- The Hackett Group (2024). GBS AI Adoption Accelerating, With 63% Seeing Early Gains. www.thehackettgroup.com/the-hackett-group-report-reveals-gbs-ai-adoption-accelerating-with-63-seeing-early-gains/
- Accenture (2024). CFO Forward Study: 2024 Edition. www.accenture.com/us-en/insights/consulting/cfo-forward-study-2024
- Gartner (2024). Gartner Predicts That 90% of Finance Functions Will Deploy at Least One AI-Enabled Tech Solution by 2026. Gartner Press Release, September 2024. www.gartner.com/en/newsroom/press-releases/2024-09-12-gartner-predicts-that-90-percent-of-finance-functions-will-deploy-at-least-one-ai-enabled-tech-solution-by-2026
- McKinsey & Company (2023). Generative AI and the Future of HR. McKinsey Global Institute. www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/generative-ai-and-the-future-of-hr
- Unilever Global Business Services (2024). GBS Operating Model and Cost Outcomes. PA Consulting. www.paconsulting.com/client-story/unilever-creating-a-global-business-services-organisation
- EU Artificial Intelligence Act (2024). Regulation (EU) 2024/1689 of the European Parliament and of the Council. Official Journal of the European Union. Full enforcement from August 2025. eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
- CDO Magazine (2024). J&J Global Services GenAI Playbook: Data Quality, Stakeholder Centricity and "Time Back" for Employees. www.cdomagazine.tech/podcasts/jj-global-services-genai-playbook-data-quality-stakeholder-centricity-and-time-back-for-employees
- Deloitte (2025). Global Business Services Survey 2025. Deloitte LLP. www2.deloitte.com/us/en/pages/operations/articles/2025-global-business-services-survey.html