As organisations accelerate the adoption of agentic AI into their finance departments, many are struggling to demonstrate clear returns. Aidana Zhakupbekova, CFO at Rydoo, explores how the priority for CFOs isn’t rapid deployment, but strong discipline to ensure that these systems deliver value within robust governance frameworks.
There are already signs that enthusiasm for agentic AI is outpacing results. Research suggests that more than 40% of agentic AI projects will be cancelled by 2027, driven by unclear value, rising costs and governance challenges.
This projection is significant, particularly given the current level of momentum behind agentic AI as the next phase of enterprise adoption. Organisations are moving beyond tools that support simple decision-making and towards systems capable of planning, executing and adapting actions autonomously. In principle, these technologies offer the potential to accelerate processes, improve responsiveness and enhance operational efficiency.
However, this potential has yet to translate into consistent, measurable outcomes.
With adoption on the rise, we’re witnessing a shift in focus from whether companies should invest in agentic AI to how it can be implemented in a way that delivers tangible returns while maintaining appropriate oversight. From finance to sales and marketing, the challenge across all business functions is to move beyond playful experimentation to establish value and, crucially, accountability.
The ROI challenge
One of the primary barriers to realising return on investment from agentic AI is how it is deployed. In many cases, it is treated as a broad upgrade to existing systems, rather than being applied to clearly defined use cases, making it difficult to measure impact or demonstrate value.
Without specific objectives or success criteria, it becomes difficult to assess whether these systems are delivering meaningful outcomes. The result is low confidence, minimal outputs and rising costs.
This reflects a broader misconception that more advanced technology will inherently produce better results. In practice, successful adoption depends less on the sophistication of the system and more on the precision of its application.
In finance departments for example, while 59% of finance functions are already using AI in some capacity, according to Gartner, many are still waiting for this to translate into consistent, strategic value. Not every process will benefit from increasing AI’s autonomy. Stronger outputs are typically achieved in targeted areas where rules are well established, data quality is high and results can be clearly measured. In these contexts, agentic AI can support more efficient execution and improved decision-making, without introducing unnecessary complexity. Realising return on investment from agentic AI is less about the pace of deployment and more about the discipline of implementation. This requires clearly defined use cases, robust performance metrics and a measured approach to scaling based on demonstrated results.
The intersection of governance and accountability
Precision of deployment must be matched by strong governance and accountability frameworks.
This is because the shift to agentic AI isn’t purely technical. Systems that act autonomously introduce a whole new layer of complexity, particularly in environments defined by human control and accountability.
The onus of responsibility, however, does not diminish. Decisions made within finance continue to shape reporting, compliance, capital allocation and investor confidence, meaning accountability must firmly remain with leadership.
Without clear controls, agentic systems risk exacerbating the very issues they are intended to solve. An AI agent approving expense claims, for example, may increase speed, but if policies are inconsistently applied, it will replicate those inconsistencies at scale.
The issue isn’t just accuracy, but also traceability and visibility. Finance teams need to understand why an agent made a decision.
Governance must therefore be built in from the outset. This means selecting tools with the right level of transparency and auditability as well as maintaining defined escalation points for human intervention.
As regulatory expectations around AI continue to evolve, these controls will become increasingly important for compliance.
While the EU has just pushed back the deadline for compliance on high risk AI systems as part of its EU AI Act, these regulations will eventually come into force. Laying the groundwork now will ensure that companies don’t fall foul of new regulations.
Agentic AI in practice
For agentic AI to deliver genuine value, finance teams need to move beyond experimentation and focus on how where these systems can add value in day-to-day processes. This means understanding not just outputs, but how decisions are generated, where intervention is required, and how exceptions are handled.
This places a greater emphasis on the role of leaders. As agentic systems take on more execution, responsibility shifts towards setting controls and ensuring decisions remain explainable. The question is no longer whether a process can be automated, but how it should be governed once it is.
