The AI conversation in the finance world has shifted in the past 12 months from exploration to justification as it has become clear AI is here to stay, and the winners are those that successfully deploy these tools throughout their operations.
In this environment, the focus becomes less on visionary transformation and more on containing the surge in hidden charges and taxes that come from AI deployment. While initial software investment can be sizeable, the continuous investments in data hygiene, governance and cloud-heavy architectures can rival and exceed that initial investment if not identified early and handled correctly.
According to PwC’s 29th UK CEO Survey, 84% of chief executives believe AI will increase efficiency over the next 12 months, but less than half have seen financial benefit from deployment to date. “In the near term, CEOs are rightly focused on efficiency, resilience and cost control. But the longer-term prize is growth and competitive advantage. That requires scaling AI beyond pilots and embedding it into how organisations operate and deliver value,” says Umang Paw, chief technology officer at PwC, UK.
This shows the battle financial directors have on their hands, managing a technology with apparently limitless potential to reform the workplace while not, in many cases, providing any near-term financial benefits.
The sticker price fallacy
Most AI business cases begin with model licensing fees or SaaS subscriptions, but this is rarely where the real cost sits. Hidden ongoing costs, from ramping up AI deployment to ensuring the data is high quality and reliable, requires a large amount of investment both in more tools and services, alongside hiring skilled workers to ensure successful deployment and management of the AI project.
In cloud-heavy deployments, data egress and storage fees can drive between 30% and 50% overspend against initial projections, according to Flexera’s State of the Cloud report. Furthermore, managing cloud spend (84%) and security (77%) remain top of mind when it comes to cloud challenges. This could be even higher in data-intensive financial workflows, where large language models are integrated into the cloud service.
When it comes to AI deployment, a large portion of time is spent on data preparation and integration rather than model development. This can bleed into future AI deployments unless the business can improve its data pipeline, which is difficult to achieve while deploying new tools and systems.
Deployment of AI-native FinOps
FinOps was originally intended to reduce the ballooning costs and unpredictability of cloud compute and storage. In the age of AI, these tools need to be reimagined to utilise the power of AI while also optimising AI spend in real-time.
Instead of a thin spread of investment across dozens of piloted projects, chief financial officers are increasingly selective over which workflows to optimise with AI. Fraud detection in high-volume transactions, forecasting, and automated monitoring are three areas being targeted.
Return-on-investment gap
Finance teams are reporting return-on-investment (ROI) from AI deployment, but these have been uneven, with industry surveys suggesting an average of 10% ROI. That’s about half as much as businesses are targeting for 2026, but there’s no indication that ROI will improve this year without serious efficiency improvements.
One reason for this gap is dark data – the result of decades of legacy finance systems that hold poorly structured, duplicated or unclassified data entries. This can be difficult to import to an AI system, and if done incorrectly can result in inaccurate analysis. This is a data maturity issue many firms struggle with, and takes a concerted effort by all parties to fix.
According to research by Deloitte, most businesses report satisfactory ROI typically takes two to four years to materialise with AI, much longer than the typical investment and deployment cycle of under 12 months. Only 13% noticed returns in the first 12 months of deployment, highlighting the long-term thinking that needs to take place to make AI a financial success. Organisations should treat AI as a core organisational transformation and fund accordingly. The majority (95%) of AI ROI Leaders allocate more than 10% of their technology budget to AI. Moreover, they are more likely than other respondents to have significantly increased their AI spending in the past 12 months and are more likely to plan to do so again in the next 12 months.
A shift in capital allocation
The era of AI experimentation has hit the hard reality of the balance sheet. For the modern CFO and CTO, the challenge for 2026 is no longer proving that AI works, but proving that it can be industrialised without eroding margins through cloud sprawl and data debt. Bridging the ROI gap requires a fundamental pivot from treating AI as a discrete software expense to viewing it as a permanent structural shift in capital allocation. Those who fail to integrate AI-native FinOps and rigorous data hygiene today will find themselves funding inefficient legacy processes with expensive modern compute — a strategic mismatch that no amount of generative hype can resolve.
Benchmarking AI value realisation
The Path to AI Value: PwC Global AI Study
This analysis breaks down why only a fraction of firms are seeing bottom-line impact. It provides a “Value Realisation Framework” specifically for finance leaders to track soft productivity gains versus hard cost savings, offering a roadmap for shifting from efficiency to top-line growth.
Optimising the economics of cloud and AI: Flexera State of the Cloud Report
As AI drives unprecedented demand for GPU and storage, this report is essential for CTOs managing vendor lock-in and hidden egress fees. It includes data-driven strategies for FinOps teams to forecast AI-related cloud spend and reduce the 30% average overspend cited in current market benchmarks.
Scaling AI through data maturity: Deloitte, AI ROI: The paradox of rising investment and elusive returns
This report tackles the dark data problem head-on, providing a maturity model for enterprise data architecture. For IT heads, it outlines the specific technical debt hurdles that prevent AI from achieving ROI in the first 12 months and offers a guide to AI-ready data governance.
The AI conversation in the finance world has shifted in the past 12 months from exploration to justification as it has become clear AI is here to stay, and the winners are those that successfully deploy these tools throughout their operations.
In this environment, the focus becomes less on visionary transformation and more on containing the surge in hidden charges and taxes that come from AI deployment. While initial software investment can be sizeable, the continuous investments in data hygiene, governance and cloud-heavy architectures can rival and exceed that initial investment if not identified early and handled correctly.
According to PwC’s 29th UK CEO Survey, 84% of chief executives believe AI will increase efficiency over the next 12 months, but less than half have seen financial benefit from deployment to date. “In the near term, CEOs are rightly focused on efficiency, resilience and cost control. But the longer-term prize is growth and competitive advantage. That requires scaling AI beyond pilots and embedding it into how organisations operate and deliver value," says Umang Paw, chief technology officer at PwC, UK.
