2025 – The Year of Agentic AI…or maybe not?

Part 2: Bringing Agentic AI to Finance

Last month, in Part 1 of this article, we discussed how agentic AI is becoming the next frontier in business transformation. We separated the hype from reality and explored real-life examples in companies like Alaska Airlines, Wiley, and Puma.

While these examples demonstrate impressive results, the question remains: “How do we actually make this work in finance?” This is what we will discuss in this part.

This technology is advancing rapidly, and – as we noted in Part 1 – both tech companies and traditional firms are making huge investments in its development and testing. The promise of agentic AI is that humans will have an agentic workforce at their disposal. Think of thousands of inorganic workers autonomously carrying out complex workflows, 24×7. The business reality, however, is much more nuanced than that.

Experimentation by tech companies to build the next unicorn model is one thing, but exposing businesses to such a technology without fully understanding it is another. In reality, business and finance functions are still grappling with years of complexity amassed in processes, disparate data, and a legacy tech stack. We still have a long way to go before we can realise any tangible and highly impactful benefits of AI agents.

Having said that, staying put is also not a choice. CFOs need to embrace and consistently move towards a future where technology is driving the core business operations. The cost of inaction can be as huge as the irrelevance of the finance function—a cost too big to be ignored.

Build a Roadmap

Do not make the mistake of jumping into point solutions and collecting use cases for agentic AI. We can surely find many candidates for automation, but we must take a step back before we get too excited about agents. Invest time in setting a clear strategy and roadmap for your digital transformation. Start with a compelling digital vision that demonstrates your ‘why’ behind the transformation and paints a vivid picture of the future. A reason for your function to transform and how agentic AI, or any other technology, fits into that future. Create a clear view of the current and future state and then build a roadmap to move from where you are today to where you want to be.

Fix the Data

A CFO I interviewed recently spent two years just fixing their data. While they continued to do smaller experiments and score quick wins along the way, the overarching direction remained around building a strong data foundation. Their strategy was to get brilliant at the basics in the first two to three years and then build advanced analytics capabilities from there onwards. Migrating to the cloud, retiring legacy systems, and building strong data pipelines enabled them to scale their digitalisation efforts much more quickly. Working on data infrastructure and foundation is like a ‘hockey-stick growth’. You do not see a lot of value in the beginning, but as soon as reusable datasets are in place, the business value shoots up owing to numerous use cases.

KPMG AI Quarterly Pulse Survey 2025 revealed that nearly half of organisations are exploring agentic AI, with around 37% piloting it while only 12% deploying it. The quality of data remains the biggest challenge (85%) to this adoption. It is imperative that in 2025, CFOs prioritise and co-create a data strategy with their C-Suite counterparts, which includes building data foundations, democratising data, and defining governance frameworks. Your AI agents will thank you for that.

Identify Momentum builders

The essence of digital transformation is experimentation. But that does not mean mindlessly investing in technology. It is especially not wise to invest too much and too soon in untested and still developing AI agents. However, as I mentioned above, not taking action is also no longer a choice.

Identify the least risky processes and workflows, which, if gone wrong, do not create a massive business disruption. The associated benefits or impacts will also be small, but it will give you a good opportunity to test and learn before you go all-in on AI agents. Keep the experiments controlled by keeping a human in the loop to validate the output. For example, you may start with workflow management of your financial planning or closing process. The agents will only be managing the coordination between different functions, tracking progress, and reporting process bottlenecks, etc. Humans remain in the loop for reviews and approval of journals, adjustments, or P&L rollups. In essence, the tasks impacting financials remain under the human domain, and a less risky collaboration piece is delegated to the AI agents.

Fix Processes

The quality of agentic AI is only as good as the underlying process it supports. Automating a complex process would only increase its complexity. We already face an explainability or a ‘black box’ algorithm challenge with AI. Using it to execute complex workflows autonomously can potentially expose the organisation to unknown risks. Make 2025 your process excellence year. Identify key processes that generate the most complexity and stakeholder noise. Carry out process excellence sprints, either internally or with external help, to identify improvement and redesign opportunities. Then, categorise these opportunities into people, process, system, and data workstreams and make them part of your digital transformation roadmap to track their progress .

Collaborate externally

The world of AI is ambiguous, complex, and rapidly changing. No organisation or finance function, on its own, can stay ahead of the trends. Collaborate with external vendors, join industry networks, and participate in tech conferences to explore what is happening in this space, how other organisations are leveraging the technology and the risks associated with it. Arrange and encourage your C-suite team to visit organisations that are ahead of the curve, not only to learn from their experience but also to get inspired. In addition, partner only with well-reputed external consultants who can help you navigate this journey. However, focus on building an in-house capability rather than solely relying on consulting firms to support the implementation.

AI is not a matter of retrofitting a few workflows to the new technology solution. It requires us to fundamentally redesign the current ways of working and the finance operating model. 2025 offers an opportunity to do just that and be ready for this revolution. We must balance ambition with return and avoid falling into the trap of the technology hype.

About the Author – Tariq Munir

Tariq Munir is a thought leader on AI, Data, and its intersection with business. He is also a keynote speaker, trainer, and monthly columnist at CFO Magazine A/NZ.

He can be reached at hello@tariqmunir.me or www.tariqmunir.me