Artificial intelligence is rapidly entering the finance function, but CFOs are still in a phase of discovery rather than large-scale transformation. While vendors are flooding the market with AI-driven solutions, most organisations are still testing use cases, particularly in transaction processing and basic automation.
In our conversation, Karan Marwah, Partner, CFO Advisory, Grant Thornton Bharat, discusses how CFOs are approaching AI adoption, where real value is emerging, and why governance and data readiness will determine success over hype.
Q: We are talking too much about AI in finance. Are CFOs really adopting AI, or are they just watching?
Karan Marwah (KM): Everyone is trying to understand what AI can do for them and then adopt what makes sense. At this stage, there is significant experimentation.
The market is crowded with vendors offering AI solutions, though in many cases it is still unclear how different they are from earlier technologies like robotic process automation (RPA).
Several use cases, especially basic transaction processing, are working well in terms of speed, efficiency, reliability, and reducing human error.
However, more complex applications are evolving and will take time to scale. There is both curiosity and caution in every CFO’s mind about how to leverage AI effectively.
Q: Can CFOs afford not to adopt AI in today’s digital landscape?
KM: The short answer is no.
However, I also see a level of FOMO (fear of missing out) — if you are not adopting AI, you risk falling behind. However, the bigger issue finance leaders face today is talent: hiring, retention, and getting the right skills at the right time while managing constraints for cost and growth imperatives.
Technology is increasingly seen as a way to address these constraints. It won’t replace humans, but it will take over certain tasks, allowing people to focus on higher-value work.
AI is also being viewed as a tool to enhance efficiency and generate insights from data, though data quality and availability remain major challenges.
So, the question is not whether CFOs should adopt AI, but how to adopt it and how to measure ROI, especially since experimentation is still ongoing.
Q: What KPIs should CFOs keep in mind during AI adoption?
KM: KPIs depend on the use case. If it is about efficiency, then metrics like throughput, accuracy, reliability, availability, resilience, and redundancy matter.
Cybersecurity and data safety are also critical, especially given the sensitivity of financial information and regulatory requirements.
Q: How should CFOs define ROI?
KM: At a basic level, ROI comes down to whether the investment is reducing cost or improving efficiency and output. Once that is established, more sophisticated measurement frameworks can be built. But the starting point should always be simple, practical outcomes. Common metrics can include improvements in cycle times, first-pass throughput, debtors’ days outstanding, creditors’ days outstanding, etc.
Q: Which finance functions will see the most tangible benefits from AI?
KM: We are already seeing large-scale adoption in transaction processing, where volumes are high and automation delivers clear benefits.
Another key area is insights generation and data analytics, though this is still evolving due to challenges around data quality and benchmarking.
Compliance is also a strong use case, where AI is helping automate controls, detect anomalies, and even flag potential fraud or errors before they occur. This is similar to predictive maintenance in manufacturing but applied to financial systems.
Q: What mistakes do CFOs make while adopting AI?
KM: It is more about experimentation and iteration. Many companies are still figuring out the right tools for their specific needs. So, I would not call them mistakes.
Off-the-shelf solutions often require significant customization. Without sufficient time, effort, and collaboration with technology partners, outcomes may not be optimal. It is less about error and more about the learning curve.
Q: Over the next 2–3 years, where will AI deliver the most impact?
KM: The biggest impact will be in the back office and middle office—transaction processing on one side and forecasting and insights on the other.
AI will also help leverage historical data more effectively and improve forecasting accuracy, as well as enable better comparison between actual and planned performance.
Q: Where can AI investments fail to deliver expected outcomes?
KM: Failures usually happen when organisations do not invest enough time in defining the problem clearly or selecting the right technology.
AI governance is critical, that is, knowing what you are trying to solve, why you are solving it, and how success will be measured. Without that clarity, outcomes can fall short.
Q: What final advice would you offer to CFOs?
KM: There is a misconception that AI will make people redundant. In reality, it will make certain tasks redundant, not people.
The key is to embrace technology with caution and discipline. Be sceptical, but not fearful. AI is the future, but success will depend on how thoughtfully it is adopted.
