Making AI Pay Off in Finance

How CFOs, acting with urgency, are reinventing finance through process-led transformation, disciplined prioritisation and AI-enabled execution that delivers measurable results

At a recent executive roundtable hosted by The Hackett Group^® in London, senior finance executives from a cross-section of industries met to discuss the ways in which they believe artificial intelligence (AI) is transforming the finance function.

In some companies, dramatic progress is being made already, while at others, pilot projects are underway, but the transformation is taking more time to arrive. Interestingly, unlike some past technological advances, their relative progress appears to depend more on commitment than scale. Success, panellists said, demands a deep understanding of the technology, clear goals and a mandate to drive adoption.

In the free-wheeling discussion, we asked the panel about the essential elements needed to build an AI-augmented finance function. Panellists weighed in on everything from the build-vs-buy debate to the skills finance teams will need to succeed tomorrow. They agreed that the most important first step in the integration of AI into finance is to reimagine a single process from beginning to end.

1. AI in finance: where are we now?

In transactions, insights and controls, AI systems are making dramatic advances. However, successful AI integration also requires leadership, courage and executive support.

Last year, four out of five finance organisations had begun their AI journey and were in the piloting phase, but few had rolled out full-scale solutions.

Since then, many have made real progress building on the insights they gained in those pilots last year. They know more about the limitations of their own data, the governance needed for new AI solutions and what we call the “art of the possible.”

For many chief financial officers (CFOs), forming a realistic picture of what’s achievable may be the most challenging aspect of adopting AI. The scale of aspirations varies dramatically. While one financial planning and analysis team is letting Claude “run all over our numbers” in Excel, other finance executives believe AI will soon “blow up our world,” not just in finance but in the whole front and back office.

Overall, while some organisations are achieving great gains, others have made little headway, suggesting the gap between leaders and laggards is widening when it comes to AI use. A number of participants noted that much progress still depends on a relatively few innovative individuals. There are also often several obstacles, including the front office taking scarce AI resources, change resistance among experienced staff, and difficulty making the business case for AI projects.

Another challenge is the rapidly evolving nature of the technology. AI is an umbrella term that encompasses a growing range of solutions supported by AI assistants, AI-enabled analytics, AI-enabled automation through agents and much more. In addition, success has varied by category, depending on whether you are talking about transactional activity, insight-driven activity, controlling activity and specialised finance activities.

Transactional activity has a clear head start, benefiting from tangible use cases, easily quantifiable business cases, and the kind of standardised, high-volume workflows that AI solutions are built to handle. According to The Hackett Group’s 2026 Finance Key Issues Study, accounts payable leads the field with 33% of organisations having scaled deployments of AI, and 34% firmly in the piloting and planning stages. Similarly, with customer-to-cash, 24% of organisations have scaled deployments in AI, and 25% are in the piloting and planning stages. Following the trajectory from last year, we can reasonably expect the majority will be using AI solutions in these processes by early 2027.

Insight-driven activity is closing the gap quickly, according to the Finance Key Issues Study, with 19% of finance functions reporting scaled use cases and 40% more in the piloting or planning stages. However, the race for second place may have a surprise contender with controlling activity coming up quickly.

Controlling activity, despite low levels of currently scaled AI solutions, is now being piloted and planned in 60% of organisations in our Finance Key Issues Study. Controlling activity lends itself well to using this advanced technology as the “eyes and ears” of the accounting process. For example, whilst overall there is reluctance to move beyond augmenting human decision-making, many finance leaders said they thought it might be permitted in use cases of intercompany matching and blocking suspicious transactions.

Tax, treasury, compliance and other specialised finance activities are lagging noticeably. For example, our Finance Key Issues Study also noted the percentage of organisations with scaling, piloting and planned use of AI in compliance management totals 15%, less than the current scaled deployment in planning and forecasting. This may be a consequence of finance’s current reluctance to accept fully autonomous decision-making and preference to see it largely as augmenting human capital. This may change if AI assistants are rolled out to support quick interpretation of complex regulation and policy documentation.

Size, it turns out, doesn’t matter to the speed of adoption of AI. We have found companies worth just a few billion whose finance functions are operating on the cutting edge of AI. The key differentiator, in the view of some on our panel, is that the CFO has a mandate from senior executives to take charge of the initiative, a clear picture of the goals in mind, and the strength of leadership needed to implement it.

2. AI: Should you build or should you buy?

In the finance function, buying AI software is the preferred option right now, but it’s not clear how long this lead will last. As no-code development grows increasingly capable, the advantage of store-bought over home-brewed may become less clear-cut.

The build or buy decision for finance, with regards to AI, typically comes down to adopting functionality that is increasingly already embedded in existing business applications – such as enterprise resource planning (ERP) or other best-of-breed tools – or building AI agentic solutions that are layered over existing applications and data infrastructure.

At present, our research suggests that “the buys” have it. For example, in accounts payable, the process that leads the adoption statistics for scaled AI solutions – the ratio of solutions embedded within ERP/best-of-breed applications vs custom-built – is around 6:1. This clear preference to adopt functionality from current vendors replicates itself in all processes, except performance reporting and analysis, which approximates a fair fight between build and buy. However, the buy preference is only on average. Our entrants to this year’s Hackett Digital Awards showcase the potential of build, including numerous multi-agent solutions at vast scale, with significant cognitive and orchestration capabilities, acting independently of enterprise applications except to write back outcomes.

Some participants noted a variation in vendor speed of AI integration, with ERP vendors perceived as slower than SaaS specialists. “I’m seeing more coming through those software companies and the tools they’re spinning up, and they’re much more agile”, one executive said. Others preferred enterprise-hosted, firewall-protected copilots for sensitive data, and most preferred ring-fenced services and internal large language models.

Despite the current preference for buying solutions, leaders who set finance transformation strategies and roadmaps continue to wrestle with the build/buy choice. Some are concerned about the longevity of their software platforms. Others wonder “how much longer will my software suppliers offer strategic value when their core competence of software coding is eroded by no-code development, and does that time period justify the costs I’m paying in licences, upgrades and implementation?”

This erosion of the barrier to building poses a direct challenge to the strategic value proposition of best-of-breed vendors. Historically, these vendors justified premium licencing costs through proprietary AI models trained on vast industry datasets, deeply embedded process logic, and the promise of continuous innovation delivered through regular product releases. However, as no-code platforms make building accessible to anyone, and large language models supply the intelligence, the differentiation that once resided in vendor-built AI is being absorbed into the broader technology landscape. Already, some executives have reportedly run their own experiments: one CFO mentioned a colleague “who spent the whole weekend writing his own timesheet tool, and it scans all his emails… and writes up his timesheet. And that is now being rolled out”.

All this suggests that today’s verdict on the advantage of buying may change. Planners are right to question whether the licensing, upgrade and implementation costs associated with best-of-breed solutions will remain justifiable over a five-to-seven year transformation horizon – particularly if no-code alternatives can, with greater agility and lower total cost of ownership, deliver comparable solutions tailored to the organisation’s own data and ways of working.

3. What should an AI-augmented finance operating model include?

Integrating AI into finance requires answering five difficult questions that have less to do with the technology than with your vision of how the function should operate. Reimagining the finance function comes down to thinking through how AI-augmented corporate finance will operate in the future, and to do that successfully, the CFO will need to ask – and answer – five important questions:

How can we optimise value from our AI investments?

A strong ideation process is critical to prioritisation, tying ideas about how to use AI back to enterprise goals. Effective and proactive process ownership within the operating model will remain a key advantage because understanding current workflows is key to spot prime AI opportunities. Finance must have the capability to map its processes and decisions to surface bottlenecks, repetitive tasks, and high-value opportunities. To avoid wasting time and money, the function must have a structured, scalable approach to evaluating use case feasibility and potential value and be able to outline expected return on investment upfront.

One key that participants noted was the need to define value broadly and consider not just full-time equivalent reductions but other gains such as improved decision support and better business partnering, which make it possible, for instance, to shift to scenario-driven decision support during periods of volatility, such as geopolitical shocks.

To break down barriers to change, finance’s operating model should be supported by a clear CFO vision of the road to an AI-augmented function. This roadmap will need to include clear objectives that support aligning resources, justifying investments and motivating the team. There should be a concerted focus to shift mindsets by communicating openly about why AI is introduced and how roles will evolve. A dedicated finance transformation office will highlight successes, incentivise use of new tools, and reassure staff that AI is there to support and empower them.

How do we maintain alignment among process owners, product managers and enterprise architects?

The finance operating model should be designed in a way that maintains alignment between process owners, product managers and enterprise architects, keeping them focused on continually reducing complexity in current processes and technology landscapes.

High-quality business data is an essential prerequisite for unlocking AI’s potential in finance, as it has been for multiple waves of technology beforehand. As with all functions, finance must continue to play a leading role in the governance and management of data in its domain. CFOs must continue to invest in master data management and data quality programs, even as AI assistants put powerful tools into the hands of business users to undertake tasks such as data cleansing.

Finally, finance must partner with IT to ensure finance can benefit from enterprise strategies that include centralised data infrastructure, scalable cloud-based technology infrastructure, and controlled environments, for piloting on test data without risk of disrupting operations.

However, the finance operating model needs to draw clearer lines – and clearer bridges – between finance and IT. A typical grey area is the overlap between process and product managers. Finance processes benefit enormously from proactive product managers – people who track what strategic AI vendors can actually deliver today, not just what their roadmaps promise.

How can we enhance governance for AI-augmented functions?

Clear governance will be just as important in AI-augmented finance functions as it has always been. Process owners, who lead the design of AI solutions, will need to work closely with the ultimate accountability owners as both operations and policy adapt to accommodate AI with financial controls. Process owners will need to make sure that:

1. Traditional controls continue to apply to processes run by AI agents. For example, with AI agents, companies must ensure that automation doesn’t unintentionally knock down critical segregation of duties, say by having a single AI agent that can initiate, authorise and record a transaction end-to-end. To address this, organisations will need to continue to implement segregation of duties within system design and user/agent access management.

2. New controls are added to mitigate new AI risks. These might include tracking AI model performance over time, monitoring for bias and setting thresholds notifying gatekeepers about when to retrain models; periodically spot-checking AI outputs for accuracy or running “data poisoning” tests; or building in audit trails and logging to ensure that AI agents’ activities are recorded in detail.

3. New AI-native agentic controls. Specialised AI agents can be designed to bolster monitoring and auditing activities by automating continuous checks. AI tools in finance can monitor transactions and flag anomalies in real time, strengthening fraud detection and financial compliance.

How should we train people to use these new tools?

Equipping people to work in an AI-augmented finance operating model requires a deliberate expansion of skills beyond traditional functional expertise. While technical proficiency with AI tools is necessary, it will be insufficient on its own. Finance organisations must cultivate strategic leadership capabilities that enable professionals to align AI investments with enterprise objectives, prioritise initiatives based on value, and make informed trade-offs between automation ambition and operational risk. As AI increasingly influences forecasting, decision support, and resource allocation, finance leaders will need to be able to look critically at AI-enabled insights, ensure accountability for outcomes, and position finance as a strategic partner rather than a transactional service function.

A second critical dimension is understanding data and AI technology. The focus should not be on turning finance professionals into data scientists, but on ensuring broad data literacy, AI fluency, and responsible oversight. Finance roles must be redesigned to reflect clear separation between machine execution and human judgement, with individuals equipped to interpret AI outputs, understand model limitations, and apply appropriate governance and ethical standards. As AI becomes embedded in core finance processes, the ability to translate data into decision‑ready narratives – and to maintain trust through robust data governance – becomes a foundational capability rather than a specialist skill.

Overall, participants foresee the need for a wide range of skill sets, including data engineering, machine learning, prompt engineering, prompt and vendor management, and change management. Both re-trained veterans and fresh talent should play a role.

Equally important is strengthening organisational change, culture and collaboration capabilities. AI adoption in finance is inherently disruptive, reshaping workflows, decision rights, and role expectations. Finance professionals must be equipped to operate in environments of continuous change, contribute to enterprisewide transformation efforts, and collaborate effectively across functions such as IT, human resources, and operations. This requires strong change leadership, problem-solving, and program management skills, ensuring that AI tools are not only deployed, but adopted, trusted and embedded into day-to-day workflows.

Finally, an AI-integrated finance organisation must preserve and elevate traditional functional expertise and advisory skills. As automation absorbs more transactional work, the value of finance increasingly lies in judgement, consistent policy application, risk awareness, and the ability to influence decisions without formal authority. Professionals must be equipped to act as advisors – listening, facilitating and coaching – while applying deep functional knowledge with execution discipline. Whatever AI delivers, accountability remains a human responsibility – and so does the judgement that earns trust.

How can we reorganise to better manage our new AI-related services?

In response to the rapid onset of AI transformation, finance will need to reorganise itself. Our model structures the finance function around six distinct capability areas, shifting it from a process-execution system into an organisation that builds and governs its own intelligent systems:

AI centre of excellence – finance’s internal app foundry. Citizen developers use no-code tools to rapidly design, deploy and iterate AI-powered finance processes in sprint cycles – without dependency on IT or vendor release schedules.

AI data & model governance – ensures AI models remain accurate, unbiased, explainable and compliant over time. Owns model performance monitoring, retraining, data quality and regulatory obligations – treating models as live assets, not static software.

Finance value architects – the commercial face of finance embedded in the business. With routine analysis automated beneath them, they focus almost entirely on interpretation, challenge and strategic decision support.

Finance transformation office – manages the finance change portfolio, including technology roadmaps, build/buy decisions, process redesign, and benefits realisation, treating transformation as a continuous campaign rather than a periodic programme.

Product managers – holds finance’s technology vendors accountable for outcomes, not just delivery. Manages contracts, product roadmaps and upgrade cycles – a role that bridges finance strategy and supplier performance.

Specialist finance – the technical core – tax, treasury, statutory reporting and complex accounting. Freed from high-volume execution by automation, specialists focus on exception handling, model oversight, and judgement that AI cannot replicate.

The future is agentic.

Organisations that act with urgency to reimagine work and the workforce with AI will begin to realise their agentic potential: cost savings and positive improvements in productivity, quality, and customer and employee experiences.

If you haven’t begun the agentic journey, start now with a clear-eyed assessment of your organisation’s AI opportunities and threats. This will help target specific process areas where AI can drive business value. Study how successful organisations, such as our Innovation Award winners, have done it. And leverage experience and proven approaches, such as The Hackett Group’s Agentic Enterprise Operating Framework, to reimagine work with AI assisting, augmenting and acting autonomously.

It’s time to act. Contact us for more information or to schedule a demonstration.