Ep. 20 | Why Finance Leaders Need AI with ROI
Related Podcasts
AI investment is accelerating across financial services, but implementing more AI does not automatically create more value.
In Episode 20 of Finance Chief Uncut, Rishi Kapoor, Global Head of Partner Sales Engineering at Alteryx, joins Matt High to explore how finance leaders can approach AI investment, governance, agentic AI and return on investment.
With more than 20 years of experience in data and technology, including roles within investment banking, Rishi argues that organisations should start with a clear business problem rather than the technology itself.
His focus is on “AI with ROI”: understanding what a process currently costs, what needs to improve and whether AI can deliver measurable value.
🎧 To listen on Spotify click here
▶️ To listen and watch on YouTube click here
🍎 To listen on Apple click here
In this episode:
- Why AI investment should begin with the business problem
- How finance leaders can measure “AI with ROI”
- Where AI can support financial services and reconciliation
- Why business context and governance matter
- How organisations should approach agentic AI
- Why human oversight needs to be measured
- How AI costs and token usage affect ROI
- Why finance leaders should start small and scale with evidence
Start with the business problem
For Rishi, successful AI adoption begins with understanding the problem an organisation is trying to solve.
Finance leaders should establish the current cost, time and performance of a process before deciding whether AI is the right solution.
That baseline also makes it possible to measure whether an investment creates genuine value.
Rishi argues that organisations should ask what happens to any time AI saves. If a team saves 10 hours, can those hours increase capacity, improve service or avoid another cost?
This distinction is central to what he calls “AI with ROI.”
Use AI where it adds value
Rishi points to established AI applications in financial services, including fraud detection, document analysis and internal knowledge search.
He also sees opportunities within processes such as reconciliation.
When two numbers do not match, identifying the difference is only the first step. Teams still need to gather evidence, understand why the discrepancy occurred and decide what action to take.
AI can support this process, but Rishi argues that technology needs the right business context to make useful recommendations.
That means organisations must define what information means, which rules apply and who has the authority to act.
Build governance into AI adoption
As organisations move from AI experimentation towards production, governance becomes increasingly important.
Rishi cites a figure of more than 88% of AI solutions remaining in pilot mode, highlighting the challenge of moving AI into operational environments.
He argues that processes should be understandable, repeatable and auditable, with business experts involved alongside technology teams.
This becomes particularly important with agentic AI.
Rishi suggests organisations should treat AI agents as delegated authority. An agent gathering information or routing a case carries a different level of risk from one changing a ledger or releasing a payment.
Organisations should therefore expand an agent’s authority gradually, based on evidence and with clear controls.
Measure the cost of human oversight
Human involvement also needs to form part of the AI business case.
As Rishi puts it, “Human oversight is the real work.”
If reviewing and correcting AI output takes longer than completing the original task, organisations need to account for that cost when assessing ROI.
The same applies to the cost of AI itself.
Rishi shares an experiment involving a financial services partner and investment bank where, according to him, using AI alone consumed 80 times more token budget and still missed two or three reconciliation breaks.
Combining AI with other platform capabilities, he says, reduced token usage while achieving 100% accuracy in the experiment.
The lesson is not to apply AI to every part of a process, but to identify where it genuinely contributes value.
Start small and scale with evidence
Rishi recommends starting with low-risk, back-end processes such as month-end reconciliation before expanding AI into more complex or customer-facing areas.
Rather than approaching AI transformation as one large implementation, organisations can build incrementally and increase its role as evidence supports the investment.
For finance leaders, this means asking a more disciplined question.
Not simply “Where can we use AI?”, but “Where can AI solve a defined business problem and deliver measurable return?”
As Rishi advises, “focus on the business problem” and build from there.
Explore more from Finance Chief Uncut
Catch up on Episode 19, featuring Mahbub Chowdhury FCIPS, Senior VP & Head of Procurement at City Bank PLC, exploring how procurement can cut costs, reduce financial risk and strengthen profitability.
