How CFOs can make AI pay off

| Article

Business adoption of AI is widespread and growing rapidly. Nearly 90 percent of respondents to a recent McKinsey survey said their organizations now regularly use AI in at least one business function. Finance functions have embraced this technology enthusiastically: In a recent survey, 65 percent of CFO respondents said they would increase gen AI investment in 2025, up from about a quarter two years earlier.

However, in an increasingly recognized conundrum, only 37 percent of survey respondents said their organizations’ AI use had contributed positively to EBIT.1 Unfortunately, many finance functions are experiencing the same disconnect between efforts and results. Some teams still spend more than half their time on operationally intensive, low-value work, with the long-promised shift toward a more strategic, insight-driven function still on the horizon. This is despite the fact that the tools currently available, including predictive analytics, gen AI, and, increasingly, autonomous agents, have the potential to reshape work across almost every finance domain.

We believe many companies struggle to realize value from AI because they start with the wrong questions. Companies often begin by deciding what tools to use and which use cases to pursue, but ask about strategic goals later. A more effective approach to rewiring starts with leaders formulating a strategy focused on the greatest opportunities to create economic value for the business. They can then reimagine how AI could help achieve those goals. Supporting this vision requires the right AI tools and data, technology-savvy talent, and an operating model that integrates AI into how people work and make decisions.

Lessons from successful AI transformations suggest that, to translate AI adoption into measurable business value, CFOs should embrace three imperatives. First, set ambitious, quantifiable targets tied to business outcomes, and have the courage to launch the transformation without waiting for perfect data. Second, commit to end-to-end process reinvention, breaking free from the “pilot syndrome” that often prevents AI initiatives from scaling. Third, invest in change management, taking an active role as both executive sponsor and value gatekeeper while fundamentally rethinking how people and AI work together and how to put freed-up capacity to productive use.

The first wave of AI gave companies a chance to experiment and learn. The challenge now is to turn those early lessons into results. With their experience in rigorous planning and evaluating returns on investment, CFOs are well positioned to lead the way—first by transforming their own finance functions, then by helping the rest of the organization put AI to work.

Strategy: Target economic value with the data at hand

A successful AI strategy for finance requires that CFOs make two choices early on: what value they want AI to create and what they truly need to begin. Leading finance teams set measurable targets tailored for their organizations while resisting the temptation to make perfect data a requisite for getting started.

Set measurable targets up front

Implementing a value-oriented approach to AI adoption begins with the finance function deciding what it needs to accomplish and expressing the goal in measurable terms linked to specific business objectives. “Better, faster, cheaper” is a good mantra to keep in mind when setting these goals, which might include improving forecasting accuracy, shortening planning cycles, or reducing the time between closing the books and producing useful financial insights. Before committing resources, finance leaders can develop a business case that estimates the expected financial benefits and the costs of implementation and ongoing operations. They can also calculate the return on investment and payback period, and establish a way to track whether the anticipated value is realized.

Companies may use AI to address very different problems, but the starting point should be the same: identify the result they want, then work backward to determine how AI can help achieve it. One leading financial institution, for example, set a target of cutting its budgeting cycle by more than half, from more than 120 days to less than 60, which would allow it to update targets more frequently and respond faster to increasingly volatile market conditions. It then implemented an AI-enhanced budgeting system designed to accelerate the process and help meet the target.

A medtech company set the goal of closing its books within three days of each reporting period and automating more than 90 percent of reconciliations. Its AI efforts focused on low-value, labor-intensive accounting activities, with the aim of increasing productivity, freeing employees to spend more time analyzing the business, supporting decisions, and improving transparency.

In a different case, a utility focused on money lost due to discrepancies between supplier contracts and invoices, such as volume discounts that were not correctly applied. While manual efforts could capture only about 10 percent of the available value, AI solutions allowed the company to set a target of capturing the total, which is more than €20 million in savings.

Setting measurable targets also helps companies decide what to do next. They can measure how much the finance process has improved, determine what the improvement was worth for the business, and compare results with the cost of implementation. The analysis can then inform whether to invest further in the new process, apply the approach elsewhere, or reconsider an AI initiative that has not delivered enough value.

AI can improve efficiency and reduce costs, but it can also do more. We have seen finance teams use it to improve forecasting and support revenue growth, manage working capital more effectively, and identify risks or opportunities that might otherwise be missed. The opportunities will vary by company, but the larger ambition is to use AI to steer and improve the performance of the business, not simply to make the finance function less expensive to run.

The larger ambition is to use AI to steer and improve the performance of the business, not simply to make the finance function less expensive to run.

Start with available data, even if it’s not perfect

Once leaders define the outcome they want, the next question is whether they have enough data and technology in place to begin pursuing it. Unfortunately, the answer often becomes a reason to stall on AI adoption, as finance teams wait for all their data to be perfectly clean, centralized, and standardized. The teams that make progress do the opposite: They start generating value with the static data they have today, giving employees experience working with AI while letting the technology itself speed up data integration and improve data quality over time.

Large technology programs, such as enterprise resource planning (ERP) modernization, are often cited as reasons finance postpones AI adoption until the new systems and data are in place. We observed this difference in approach at two companies that were both updating their ERP systems. One treated the ERP program as a roadblock and put AI on hold until it was finished, causing a two-year delay in AI adoption. The other decided to leverage AI to determine which processes to overhaul first, test the changes, and shape the ERP transformation around what it learned.

The latter approach allows the company to learn, before its new ERP system is in place, how AI could change the way its finance processes work. It can use these lessons to redesign its planning and control processes around AI agents and shape the ERP transformation accordingly. The company expects its AI-enhanced processes to increase productivity by between 20 and 30 percent, optimize working capital by more than 20 percent through better forecasting, and reduce SG&A costs by roughly 5 percent.

AI can also help build support for large-scale data governance efforts, which can improve data quality and protection. These programs are often led by finance, but business teams may resist because they require significant effort without always providing immediate benefits. Companies can change this equation by ensuring that data transformation provides teams with immediate benefits, such as AI assistants that make it easier and faster to analyze customers’ data or access dynamic reports. This can generate support for the greater centralization and standardization of the data these tools require.

Execution: Commit to end-to-end process reinvention

Many finance functions began their AI efforts by experimenting with a profusion of pilots that let them explore the technology’s capabilities. But today, capturing real value hinges on choosing specific areas where AI holds promise and then rewiring them completely with AI. Among other benefits, this approach frees up teams’ capacity (rather than just improving individual productivity), enabling them to focus on more value-added business activities. Finance teams that develop expertise in end-to-end AI process transformations within their function can then lead the way for the rest of their organizations.

But today, capturing real value hinges on choosing specific areas where AI holds promise and then rewiring them completely with AI.

Move beyond ‘pilot syndrome’

Pilots are useful for testing new applications, but a large collection of them can consume time and investment without really changing how the function operates. Rather than spreading resources across many experiments, it’s preferable to choose a promising domain and commit to completely rethinking it with AI as part of the process. The point is to concentrate AI investment where leaders can trace process changes back to the value target.

For example, a global telecom company used AI agents to transform its financial planning and analysis (FP&A) work, automating much of the process of preparing forecasts and identifying potential gaps against financial targets, along with actions to address them. The rewiring accelerated the forecasting cycle threefold and freed up FP&A team members’ time (50 percent of which was previously spent on sourcing and aggregating data). They then used their new capacity to work with business leaders on evaluating options and making performance improvement decisions. By enabling finance to play a more active role in shaping business decisions, such transformations can strengthen the CFO’s position as a strategic partner to the business.

Redesign processes, not tasks

General-purpose assistants and chatbots are relatively easy to roll out and demo. But because dropping a tool into an existing workflow often automates a task while the surrounding process largely stays the same, it can be difficult to trace any economic benefit to the use of AI.

Finance can instead use AI to design better processes that reduce costs, increase speed and quality, and improve business outcomes. Successful process reinvention rests on three main principles: First, redesign workflows by defining activities and their owners (as in traditional process redesign), while also establishing decision rights, handoffs, and controls to determine where AI can help and where people need to exercise judgment. Second, develop specialized agents aligned with the company’s established ways of working, communication style, and tone of voice, and invest in their training to deliver ongoing accuracy and reliability. Third, make deliberate build-versus-buy decisions, leveraging off-the-shelf solutions for standardized, highly regulated activities such as tax filing, while investing in tailored solutions for more distinctive, business-specific processes.

Once finance team members’ workdays are significantly altered, leaders can decide whether to capture cost savings or redeploy freed-up capacity to higher-value activities. These could include identifying granular internal and external trends, proactively recommending business actions, and driving performance improvement initiatives.

In an example of an end-to-end rewire, a European bank completely redesigned its process for reviewing financial performance and deciding how to respond. The bank rebuilt the process around roughly ten AI agents, each specializing in a particular part of the workflow. Bank employees trained the agents to follow the institution’s established approach to performance analysis, including its logic for drilling down (for example, first by geography and then by product). The agents also learned the bank’s internal terminology.

Today, the agents prepare end-of-period reports, investigate the root causes of changes in performance, and suggest possible responses. They also prepare dashboards for management discussions in the format the bank prefers: the key message first, followed by a supporting table and recommended next steps. During these discussions, users can ask questions about the numbers and receive answers in real time. The system has achieved 99.5 percent accuracy across more than 2,000 KPIs and can answer questions in less than five minutes. The bank also reports highly positive feedback from its top management team.

Change management: CFOs lead and define roles

When it comes to technological change, the hard part is often not implementing the technology but rather updating the operating model. Deriving economic value from AI requires finance leaders to rethink the work people do (including how the CFO approaches the job), the skills and roles the function needs, and how people and AI will mesh.

Lead the change from the top

CFOs have two important roles in leading AI adoption. The first is to demonstrate their commitment by becoming visible sponsors of AI themselves. This might mean promoting AI-enabled dashboards, encouraging teams to answer questions through agents, and incorporating the technology into their own decisions and routines. When CFOs make AI part of how they work, they encourage others to do the same.

The second is serving as the gatekeepers of AI investments within finance and across the wider company. CFOs need to lead by example, applying a rigorous approach to continually evaluating whether spending on AI delivers sufficient value to justify costs. This requires accounting not only for the cost of building and running AI agents, including computing and token costs, but also for the human oversight they require, which can be a significant ongoing expense. Agents need ongoing tuning, and agent economics are constantly changing. Therefore, CFOs have to regularly reassess which applications merit further investment, which could benefit from changes in the balance between human and automated work, and which should be retired.

Decide what AI does and what people do

AI is particularly strong at pattern recognition, forecasting, optimization, automation, and functioning at scale. But it cannot exercise judgment or be accountable for a decision. When redesigning a finance process around AI, finance leaders have the opportunity to decide where the technology can operate independently and where people need to apply judgment and take responsibility for the result. Most activities will fall into one of three broad categories:

  • Fully automated: AI performs the task without routine human review.
  • Human-validated: AI performs the work, but a person reviews or approves the result.
  • Human-guided: A person remains actively involved, using AI to support the work.

These decisions can change how finance teams are built. Traditionally, an individual finance professional might perform a range of transactional and analytical tasks. In an AI-enabled finance team, that professional may instead oversee several agents doing portions of the task and step in where human judgment or accountability is required.

New roles can emerge as well. Some people may orchestrate the work of multiple agents, while others train and improve the models. Business unit leaders may spend more of their time interpreting AI output and using it to inform decisions. The skills finance needs will change accordingly, with less emphasis on manual reconciliation and more on judgment, scenario thinking, agent supervision, and the ability to explain complex analysis clearly.


Just a year ago, many finance functions were exploring how AI might fit into their work. Today, a growing number of CFOs and finance teams are targeting tangible results from their AI investments. Achieving these results requires a value-oriented approach to AI adoption, one that begins with the outcomes companies want to achieve, proceeds with focused and deep implementation, and leads to careful rethinking of processes and roles.

Rewiring finance does not need to happen all at once. Companies can start with one process where the potential value is clear, learn from the experience, and apply those lessons to the next process. This incremental approach can also reduce the temptation to wait for the technology to mature further. Although waiting for the next generation of AI may seem reasonable, getting started now gives finance time to learn what works, improve their data and technology along the way, and give their teams experience working with AI.

AI will continue to advance. Finance functions that learn how to put it to work today in ways that create value will be better prepared to capture what it makes possible tomorrow.

Explore a career with us