AI in finance: Where expectations fail and what drives value

Finance teams today are under more pressure than ever to move faster, provide strategic insight, and support business growth. Yet, many are still held back by manual processes, fragmented systems, and outdated operating models.

To explore what is really slowing finance down, where automation creates the most value, and why AI adoption often falls short of expectations, we spoke with Ivana Kappenmann, Executive Partner and CFO, and Mario Gula, Partner and AI Lead. Together, they share their perspectives on the current state of finance transformation and what organisations need to get right before AI can deliver real business impact.

What is really slowing down finance teams today?

Mario:

The biggest slowdown is poor data foundations. Finance teams operate on unstructured data, siloed systems, and often conflicting numbers across sources, which means there’s no true single source of truth.
As a result, a significant amount of time is spent reconciling and validating data instead of analysing it. This delays reporting, reduces trust in numbers, and ultimately slows down decision-making across the business.

Ivana:

In my view, what holds finance teams back today is that the mandate of finance has become much more strategic, but the function is still too often organised around reporting cycles rather than decision support. Senior management expects speed, foresight, and business partnership, while finance is still spending too much energy on coordination, controls, and operational complexity.

So the issue is less about working harder and more about whether the finance model is fit for purpose. The teams that move fastest are the ones with clear ownership, standardised processes, and the ability to translate financial information into action, not just into reports.

Are finance functions truly digital or just digitised?

Mario:

Most finance functions today are still digitised, not truly digital. They’ve moved from paper to digital tools, but the underlying processes remain manual and document-driven, still heavily reliant on PDFs, emails, and human intervention.

True digital finance means redesigned, system-driven workflows with structured data flowing end-to-end, enabling straight-through processing. Many organisations are still transitioning toward this model, often constrained by legacy systems, fragmented data, and regulatory requirements.

Ivana:

I would say most finance functions today are digitised, with pockets of being truly digital. Many have implemented ERP platforms, dashboards, workflow tools, and automation in specific areas, but the operating model is often still built around period-end reporting, manual controls, handoffs, and exception handling. That tells me the transformation is still incomplete. Recent studies point in the same direction:

PwC says ERP upgrades and stronger control environments are still top priorities for CFOs, KPMG found that 45% of controls in banking and insurance finance functions remain manual, and 70% see their manual-versus-automated control mix as suboptimal, and Gartner reports that 87% of organisations with ERP already in place still plan to replace or upgrade it within three years. So in practice, many teams have digitised the surface, but they have not yet fully redesigned the finance engine underneath.

What truly digital finance means to me is something deeper: standardised processes, clear data ownership, integrated planning and reporting, and automation embedded end to end so that finance can spend less time producing information and more time steering the business. In regulated industries like banking, that transition is harder because legacy architecture, control requirements, and organisational complexity slow down change. So my answer would be: finance functions today are generally no longer analogue, but many are still not digital by design.

Why is document-heavy work still one of the biggest hidden inefficiencies in finance?

Ivana:

I would say the real problem is not document handling itself, but the amount of finance capacity that still gets consumed by manual intervention. When 40% of CFOs (source: PwC) still rely on manual data adjustments and nearly half of controls remain manual, it is clear that too much effort is still going into validating and moving information rather than using it. The business pays for that through slower reporting, less scalability, and weaker decision support.

Mario:

In many cases, finance teams still spend too much of their time working with documents, reviewing, extracting data, retyping, and cross-checking information.

The real cost isn’t just time, it’s opportunity cost. Highly skilled professionals are tied up in repetitive tasks instead of focusing on analysis, forecasting, and business support. On top of that, manual work introduces errors and limits how much the business can scale.

Where does automation deliver the fastest business value?

Ivana:

I would start with processes that create pain every month and where the ROI is visible quickly: reconciliations, close activities, control execution, and high-volume transaction exceptions.

Those are strong candidates because they affect reporting speed, control quality, and the amount of senior finance time tied up in repetitive work. Recent research supports that view: KPMG found 45% of controls in finance are still manual, while Gartner says common AI use cases in finance already include knowledge management at 49%, accounts payable automation at 37%, and error detection at 34%. I would also say the barrier is no longer only technology. A lot of organisations now have AI tools, but adoption is still uneven — Deloitte found only 13.5% are using agentic AI today, even though more than 80% believe AI will become standard. So the quickest wins come from combining automation with clear ownership, disciplined adoption, and redesigned workflows.

Mario:

Document processing is typically the fastest and most impactful place to start. Tasks like data extraction, validation, and report generation are repetitive, high-volume, and easy to measure in terms of ROI.

Automating this can reduce processing time from days or weeks to hours, while also improving accuracy. It’s a strong entry point because it delivers quick wins, demonstrates value clearly, and builds momentum for broader automation initiatives.

What about AI in the sector? Where does reality fall short of expectations?

Mario:

AI usually falls short when expectations are misaligned. It’s often treated as a plug-and-play solution, but in reality, it’s a probabilistic tool that depends heavily on data quality and context.

If the underlying data is inconsistent or poorly structured, AI will not fix that, it will amplify it. It works best as an augmentation tool, improving human output, not replacing sound processes or domain expertise.

Ivana: 

My view is that the biggest gap is not between AI hype and AI potential, but between AI ambition and organisational readiness. In finance, companies often expect AI to transform decision-making very quickly, but in reality, it delivers best today in narrower, supervised use cases like anomaly detection, commentary support, and knowledge retrieval. 

Gartner found that adoption in finance remained steady in 2025, with 91% reporting only low or moderate impact initially, and Deloitte found that while over 80% expect AI to become standard, only 13.5% are actually using agentic AI today. In a regulated sector, trust, governance, and data quality matter as much as the technology itself. So the real lesson is that AI works best as an accelerator of a well-run finance function, not as a substitute for one.

What actually blocks AI and automation adoption?

Ivana:

What actually blocks AI and automation adoption is usually not the technology itself. In finance, the real barriers are fragmented data, non-standardised processes, and the fact that organisations are often not ready to trust and scale these tools properly. 

Research shows that poor data quality, limited technical capability, and lack of trust are still bigger obstacles than access to the technology. So the real question is not whether the tools exist, but whether the function is ready to use them in a controlled way and connect them to real business value.

Mario:

The biggest blocker is choosing the wrong use cases and not aligning AI initiatives with real business value.

Many organisations start with visible but low-impact use cases, like basic chatbots, which creates the impression that AI doesn’t deliver results. At the same time, failed POCs are often treated as proof that AI doesn’t work, rather than a signal that the use case or approach was wrong.

Successful adoption requires focusing on high-impact areas, following a structured path from POC to production, and ensuring clear ownership and measurable ROI from the start.

Conclusions

Finance transformation is no longer just a technology agenda. Real progress comes from fixing data foundations, redesigning processes, and building an operating model that can support automation at scale. At the same time, AI adoption is fundamentally a change-management challenge: value is only created when the organisation is ready to work differently, trusts the outputs, and has the skills and governance to use them effectively. 

The businesses that move fastest will be those that combine the right technology with clear ownership, disciplined process design, strong user adoption, and a sharp understanding of where automation can create measurable value and competitive advantage.