As if business disruptions and intensifying competition were not already keeping C-suite executives up at night, another pressure is emerging: corporate support functions—long a relatively stable cost of doing business—are becoming a drag on the bottom line. But this is not just a cost problem. It reflects a deeper shift in how companies need to operate.
The surprise is understandable. General and administrative (G&A)—the umbrella covering finance, HR, legal, compliance, procurement, and technology—has historically been treated as manageable territory. Companies trimmed it periodically, but volumes were predictable, growth was linear, and the work was overwhelmingly human led.
That model is now breaking down, due to the convergence of several forces:
- AI is compressing knowledge work while AI-enabled competitors operate with leaner cost bases and faster decision cycles.
- Regulatory and geopolitical complexity is increasing administrative burdens, even as corporate functions are weighed down by decades of organizational layering.
- Talent requirements are tilting toward digital and technology specialists, and decision speed has become a competitive advantage.
Together, these forces are pushing traditional G&A setups to a critical breaking point.
The deeper problem is that most corporate functions were designed for a slower, more stable business environment. They are increasingly incompatible with the speed, complexity, and integration required in AI-enabled organizations. What is required now is not simply applying AI tools to existing processes but rethinking how these functions run—making them more automated, efficient, and able to support the business with far fewer resources.
This is the premise behind functions of the future (FoFs)—the shift toward leaner, more automated corporate functions designed to perform at AI speed. Here we outline a FoF framework to help companies navigate that transition.
More Than AI
Best-in-class organizations with streamlined, tech-enabled corporate functions run G&A at roughly 3% to 5% of revenue, while the average for most other companies is near 8%. (See Exhibit 1.) In sectors such as financial services and industrial goods, the gap is particularly wide. For a company with $10 billion in revenue, the difference amounts to $300 million to $500 million a year—not in product development or go-to-market investment, but in overhead.
Stay ahead with BCG insights on people strategy
What distinguishes the leading companies on the right side of the ledger is not simply greater use of AI, but how it is applied. Rather than layering tools onto existing tasks, they have rebuilt functional activities around AI—reworking processes and how tasks are handled end to end, simplifying decisions, and redesigning how support functions operate. This helps explain why many left-side organizations deploying AI are not seeing comparable results: the technology is being added, but the underlying model remains unchanged.
Early examples illustrate the kinds of improvements companies are beginning to see. In many organizations, much of the pressure to add headcount now comes from corporate functions. Compliance, analytics, HR, and other knowledge roles are expanding as regulatory and business complexity grows. One large technology company addressed this by requiring “AI justification” before approving new hires; that is, managers must show that AI cannot perform or significantly augment the role before additional headcount is added. With AI literacy and AI agents, functional headcount has remained flat even as revenue grew about 25%.
Beyond headcount discipline, the impact of AI can be substantial. In some cases, G&A costs have been cut by half or more as the technology automates 60% to 70% of routine jobs, freeing teams to focus on activities that require judgment and decision making. Replacing periodic reporting cycles with always-on workflows can reduce cycle times by as much as 90% while significantly narrowing the number of direct reports and issues competing for the CEO’s attention.
Replacing periodic reporting cycles with always-on workflows can reduce cycle times by as much as 90%.
As these changes take hold, companies begin to change in more fundamental ways. Corporate functions shrink as automation reduces demand for transactional and coordination roles, with many organizations seeing 25% to 40% reductions in functional headcount over time, particularly in middle layers. Decision cycles collapse as real-time data replaces periodic reporting and reconciliation, in some cases reducing cycle times by 50% to 90%. Leadership structures simplify, with fewer functional silos and a smaller number of senior leaders responsible for coordinating across functions.
The scale and speed of these improvements are often unanticipated. Companies using the FoF framework can typically undertake the transformation with investment of roughly 0.3% to 1% of annual revenue over 18 to 36 months, with payback achieved in 12 to 30 months, depending on company size and starting point. In other words, this is not a decade-long rebuild of corporate infrastructure but a fast reconfiguration of the way corporate functions run.
For many companies, the opportunity rivals major supply chain or commercial transformations. Unlike those initiatives, which can take a long time to prove their worth, FoF programs typically pay back within two years while permanently reducing administrative complexity. As a result, they are not just efficiency efforts, but one of the most direct ways to improve both cost structure and how the company operates.
Why AI Alone Is Not the Answer
Those outcomes remain the exception. As a broad trend, BCG has observed that as many as 95% of companies adopting AI have failed to capture meaningful value because they deploy it on top of operating models that were never redesigned to support it. Most stop short of the deeper changes required, and many take an individual approach, implementing it in discrete, often segregated, functions rather than across multiple interwoven company activities and strategies.
Three issues explain most of the AI shortfall:
- Organizational Silos. Sub-functions use AI chiefly to optimize their own objectives with little incentive to align across boundaries. AI sharpens the individual silos without closing the gaps between them—and that is where much of the value disappears.
- Weak Governance. AI initiatives launch locally and operate under inconsistent standards. Without centralized ownership and coordination, what should become a shared capability is only a set of disconnected programs.
- Change Management That Doesn’t Change Anything. Many organizations lack clear KPIs for AI adoption, underinvest in capability building, and fail to redefine roles and decision rights. Tools get deployed, but behavior doesn’t shift.
The result is a striking disconnect between confidence and reality. In a recent BCG survey of more than 200 corporate executives, roughly 80% of CEOs said their organization was well positioned to adopt AI relative to competitors. But only 25% had put in place the foundational prerequisites to make that true. Leadership conviction is running well ahead of organizational readiness.
Those prerequisites matter more than most AI roadmaps acknowledge. Modular ERP, integrated data platforms, agentic workflow reconfiguration, and scalable global business services (GBS) approaches are not glamorous investments. But without them, AI deployment hits the same ceilings it always has. The organizations realizing durable returns built the foundation first.
What the Functions of the Future Company Looks Like
In many large organizations today, CEOs spend a disproportionate share of their time managing corporate functions rather than the business itself. Upward of 80% of the CEOs we surveyed reported that 40% or more of their direct reports sit in G&A functions, and 70% said those arrangements should be consolidated. As functional roles and sub-functions have multiplied, administrative oversight has expanded along with them.
The functions of the future model collapses that complexity. In an FoF organization, corporate functions run with much smaller teams as AI agents handle much of the routine work. Human roles focus on judgment, strategy, and business support, while more activity moves onto shared platforms and specialized partners.
At the same time, the responsibilities of corporate functions begin to shift more fundamentally. Instead of operating as separate departments, they increasingly act as platforms that run work end to end. Activities such as hiring, procurement, and reporting do not move through a chain of functional handoffs. They are executed through integrated systems, with automated workflows handling most of the underlying activity and functional teams providing oversight, expertise, and governance. In practice, this means, for example, that automation seamlessly would handle hiring end to end—from requisition through offer—without waiting for approvals and working through disjointed tools in four different departments.
Leadership structures evolve as well. As functional tasks become more integrated and automated, companies need far fewer senior leaders to oversee siloed sub-functions. Instead of multiple functional heads competing for the CEO’s attention, organizations move toward simpler reporting relationships—often anchored around a CFO and a consolidated administrative leader such as a chief administrative officer (CAO) responsible for coordinating functions. A general counsel, often reporting to the CAO, is also frequently needed to advise on governance, reporting and regulatory compliance matters.
Many organizations are already taking steps in that direction. Some CFOs, for example, are targeting 80% of finance transactions handled automatically without human intervention, with similar ambitions emerging in HR, legal, and supply chain operations. The objective is not marginal efficiency gains, but highly automated functional systems backed by hybrid teams of humans and AI agents.
The Seven Building Blocks of Functions of the Future
Companies making the most progress are changing the way functional activity is organized, executed, and governed. In our experience, the shift rests on seven building blocks (see Exhibit 2).
- AI-First Functional Strategies. Redesign work with AI as the starting point by determining which end-to-end workflows are best suited for AI—typically those with meaningful value pools, many manual tasks, and heavy data requirements. This produces a different workforce mix, with more prosaic duties handled by AI agents and human talent focused on higher-value concerns.
- GBS as an Autonomous Execution Engine. Shift 30% to 40% of functional activities into AI-enabled GBS platforms that operate continuously, manage exceptions automatically, and complete the lion’s share of transactions without human intervention. The strategic reframe matters: GBS runs as a scalable engine that drives performance, not as overhead to be minimized.
- Human Decision Making, Enabled by Centers of Excellence (CoEs). As AI takes on more analysis and routine execution, people shift to setting direction, defining parameters, and exercising judgment. Lean CoEs bring together expertise, standards, and governance, with much of the underlying work handled by AI.
- Modular ERP and a Single Data Backbone. Autonomous workflows depend on clean master data and integrated systems. Without a unified data architecture and modular ERP, AI output may be locally useful but globally inconsistent.
- Digital Fluency as a Talent Requirement. AI literacy becomes a requirement for hiring, promotion, and leadership, rather than a specialist credential or a training elective. Functional talent strategy must be rebuilt around AI-enabled roles rather than traditional positions.
- Leadership for a Human-plus-Agent World. Leaders must manage hybrid human-agent teams, oversee automated workflows, and continuously improve performance across both. This requires fewer fragmented leadership roles and stronger integrative management at the top.
- Outsourcing as a Capability Accelerator. Strategic partnerships with specialized providers can speed up automation and AI deployment far faster than building capabilities internally. Determining what to own and what to access through external ecosystems becomes a central element of functional strategy.
Many organizations are already implementing elements of this model—automating finance processes, deploying AI in compliance, consolidating shared services—and some are making real progress. But companies that apply only parts of the system rarely achieve the full impact of functions of the future. The building blocks sync as a system: the greatest results are achieved when they are implemented together rather than piecemeal.
It is important to note that in an AI-first functional world, the dashboards that companies have relied on for decades will mislead them. New gauges of success are necessary. For instance, in a finance firm’s shared services activities, traditional metrics focus on labor economics, designed for a world where humans touched every transaction, such as cost per invoice processed, full-time employee utilization, and error rates. But as AI absorbs the high-volume, routine work, these measures break down. (See Exhibit 3.) Cost per invoice approaches near-zero as overhead and marginal cost collapses, making it meaningless as an efficiency signal. What matters instead is the cost per exception resolved, since humans handle what AI cannot. Similarly, FTE utilization becomes a distorted lens when half the throughput is machine-generated; instead, capacity utilization across human and AI combined is a more salient data point. The finance function of the future doesn’t just need new tools, it needs a fundamentally different scorecard, one built for outcomes rather than inputs.
The finance function of the future doesn’t just need new tools, it needs a fundamentally different scorecard, one built for outcomes rather than inputs.
From Building Blocks to Implementation
These seven building blocks are the defining characteristics of the FoF model. Leading companies follow a consistent sequence. They begin by redesigning processes across functions, mapping how work actually moves through the organization and identifying where tasks can be eliminated, automated, or simplified. They then layer in AI and agents, determining what runs autonomously and where human judgment remains essential. With that in place, they align data and systems so information flows cleanly and consistently across the business.
The fuller reshaping of roles and leadership structures follows as AI-enabled processes come online at scale, reducing layers and redefining responsibilities around the new way of operating. Where companies most often stumble is when organizational changes or technology deployments move forward without the process redesign work underneath them, which is one reason AI programs often fail to scale.
In practice, companies execute this sequence in stages, with changes in processes, technology, and teams reinforcing one another (see Exhibit 4).
Most transformations begin with a short diagnostic and design phase lasting four to six weeks. During this period, companies identify the processes with the greatest automation potential, determine how data and systems will support the new workflows, and outline the organizational changes required. The goal is a clear plan instead of a collection of disconnected AI experiments.
In the subsequent four to five months, the first automated tasks are implemented, focusing on a small number of high-volume activities that can quickly demonstrate results. These early deployments show how the new approach works in practice and establish the foundation needed to expand further.
The final stage—usually the second six months of the project—extends the model across the organization. Automation is implemented in additional functional areas, shared service platforms are expanded, and AI is integrated into centers of excellence. In addition, the extent and success of AI adoption is tracked via KPIs and improved by targeted upskilling.
AI-Led End-to-End Workflows
As companies put these changes in place, the way corporate functions operate begins to change. Work does not move through a chain of departments. Instead, it runs through unified workflows that span entire activities—from planning to reporting or from sourcing to payment—with AI agents handling much of the underlying work.
People still set direction and supervise the system, but the workflows themselves run continuously. Agents move data across systems, coordinate tasks, and handle routine decisions automatically. Instead of periodic reporting cycles and manual handoffs, activity flows in near-real time, improving speed and accuracy while removing much of the repetitive effort.
This shift also changes the economics of G&A. In traditional models, cost control is reactive. Companies analyze their expenditures after they occur and then consider reduction measures in postmortems. In an AI-first model, cost management can be preventative. Autonomous workflows act more like a neuronal system, continuously monitoring signals, enforcing constraints, and intervening before cost is generated. As a result, companies can reduce G&A toward the 3% to 5% range, while operating with AI-based structural guardrails that govern costs in real time rather than after the fact.
Exhibit 5 illustrates how this shift would happen in finance. Traditionally, finance responsibilities have been divided among teams such as FP&A, budgeting, and accounting, with information moving between them through reviews and reconciliations. Over time, many companies have tried to streamline these activities into end-to-end structures such as record-to-report or order-to-cash.
Agentic systems extend this model further. Workflows run continuously through coordinating and execution agents, while people focus on areas that require judgment, such as exceptions, analysis, and regulatory reporting. They also supervise and refine the automated systems that carry out routine work.
Early Results from Partial Implementation
Few companies have implemented the full FoF approach. Most organizations are applying individual building blocks—modernizing data platforms, rethinking workflows, or deploying AI agents in specific functions. These efforts are already producing measurable gains, but they fall short of the full potential of a complete transformation. Most companies experimenting with individual levers improve performance, but reaching the 3% to 5% G&A levels achieved by leading organizations typically requires adopting the entire system.
Several examples illustrate the progress companies are making. A global bank facing rapidly rising regulatory costs redesigned its compliance function around AI-enabled monitoring and reporting. By embedding anomaly-detection agents into transaction oversight and automating large portions of regulatory reporting, the bank reduced compliance costs by 35% to 45%, cut breach incidents by 40%, and shortened reporting cycles from quarterly reviews to near-real-time monitoring.
An oil and gas major with selling, general, and administrative expenses above 8% of revenue focused on transforming its GBS platform into an AI-enabled execution engine. After consolidating shared-service operations, standardizing master data, and automating source-to-pay processes, the company automated roughly 70% of transaction clearing, reduced invoice cycle times by 80%, and significantly simplified leadership oversight as shared services moved onto a single platform.
Strategic Challenges Posed by FOF Models
As companies embrace FoF approaches, several practical questions will arise. One concerns the workforce. Automation will reduce demand for many transactional roles across G&A while increasing the need for people who design, supervise, and improve AI-driven systems. Companies will need explicit talent strategies combining reskilling, redeployment, targeted recruitment, specialization, and, in some cases, workforce reduction. In practice, this means redesigning roles and removing layers of management, reporting, and cross-functional coordination that no longer add value. This is often the hardest part of the transition.
Organizational design will also evolve. As workflows become more integrated and automated, companies will move toward flatter structures with fewer fragmented functions and simpler leadership layers. Middle management roles focused primarily on coordination, reporting, and oversight are likely to shrink, while spans of responsibility increase. The goal is to reshape the organization so that business unit leaders, functional executives, and the C-suite are not pulled into layers of management.
Finally, governance will have to keep pace. Boards and senior executives will need to know where AI is used in corporate functions and who is accountable for its decisions. This includes establishing clear ownership of AI-driven processes, defining who is responsible when systems fail or produce unexpected outcomes, and clarifying how decisions are made. Even highly automated operations require clear responsibility and human judgment for key decisions.
Conclusion
The implications are clear. Companies that rethink how corporate functions perform will gain speed, control, and efficiency that traditional models cannot match. More importantly, they determine how effectively and flexibly the company operates and competes.
For decades, corporate functions were treated as overhead to be managed. In an AI-driven organization, they become the functions of the future. They go beyond support roles to become critical systems and processes that define how the company runs, responds, and adapts. Each company will build them differently. The choice is not whether to improve efficiency, but whether to fully redesign and integrate corporate functions—or continue to absorb the cost, complexity, and slower performance of models that do not work. Those that move first will set the pace; others will be forced to catch up.
