AI Agents Don’t Matter Without Adoption
Sibaranjan Patnaik explains why AI succeeds in finance only when teams redesign workflows, automate routine work, and keep human judgment at the center.
Sibaranjan Patnaik
Director of Finance and Strategy
AI Transformation: Sibaranjan emphasizes AI's role in redefining finance workflows, beyond mere presentation changes.
Strategic AI Use: AI helps finance teams operate at higher strategic levels, closing gaps between teams and headquarters.
Workflow Automation: AI automates mechanical finance tasks, allowing teams to focus on analysis and decision-making.
Adoption Challenges: Effective AI integration requires redesigned workflows and overcoming resistance from finance professionals.
Tool Exploration: Claude excels in finance use cases, but the exploration of various AI tools is recommended.
Sibaranjan Patnaik is the Director of Finance and Strategy at the India-based GCC of a large technology company. With 16 years of experience, he has worked across roles in both accounting and FP&A, giving him a comprehensive understanding of how finance functions operate.
We caught up with Sibaranjan to learn about the internal finance agents he's building — and why AI success is defined by the people adopting it. Here's what he said.
The Technological Transformation of Finance Is Finally Happening
I'm Siba — Director of Finance and Strategy with about 16 years of experience.
Continue Reading for Free
Create a free account to finish this article, plus get ongoing access to timely insights and practical resources.
Have an account? Log In
I started in accounting, moved into FP&A, and that journey across both sides of the function gives me a complete picture of how finance operates daily — not just the planning and forecasting layer, but the transactional and reporting engine underneath it.
Today, my remit has expanded beyond finance — I lead technology teams, revenue functions, and multiple verticals simultaneously, operating as a cross-functional leader as much as a finance one. That breadth changes how you think about transformation — you stop seeing it as a finance problem and start seeing it as an organizational one.
In hindsight, I find this almost funny: Throughout my career, I've heard about technology-driven finance transformation. Better tools. Smarter systems. The death of Excel. Yet, at every role I walked into — whether at a large FMCG, a global retailer, or a GCC — the core workflow remained the same. Pull the data, reconcile it, build the model, format the deck, send it up. Excel, every time.
We had dashboards built on manual data pulls. We had planning tools that still required someone to babysit assumptions. Technology existed in pockets, but transformation never quite arrived.
What's different now — and I say this as someone skeptical of the finance tech hype — is that AI, specifically agentic AI, is the first thing I've used that truly changes the underlying workflow, not just the presentation layer. I'm using it today to automate commentary, accelerate scenario analysis, and handle the repetitive analytical work that used to consume most of a team's week.
For the first time in my career, I genuinely feel the promised transformation is finally moving in the right direction. And that's what I try to bring into conversations when mentoring young finance professionals — because the window to get ahead of this is now.
Using AI to Operate at a Higher Strategic Level
I work for a mature, publicly traded organization with global operations. I won't share specific details publicly — I want it to be clear that what I share here are my personal observations only — but I can say I lead a finance organization within a GCC setup in India. This means we are a strategic delivery hub for a much larger global function.
The complexity is real; we do more than back-office processing. The team spans FP&A, revenue, and technology functions, operating across multiple verticals with significant cross-functional interdependencies. In this environment, we constantly balance global standardization with local execution realities.
From a transformation standpoint, GCCs in India have historically been seen as cost centers — built for efficiency, not influence. I am largely trying to shift that perception, both internally and in how the team sees itself. As a finance partner for a global business unit within a GCC, you need to operate at the same strategic level as anyone in the headquarters. AI is a big part of how we close that gap.
Why AI Overhauls in Finance Require Patience
I don’t have a fully deployed end-to-end AI-powered finance workflow running in production today. I’d be skeptical of anyone who claims they do at scale, with controls, inside a regulated public company environment.
I don't have a fully deployed end-to-end AI-powered finance workflow running in production today. I'd be skeptical of anyone who claims they do at scale, with controls, inside a regulated public company environment. I have a clear direction, a set of productivity tools that are genuinely changing how my team and I work daily, and a pipeline of agents in various stages of building and testing.
On the productivity side, I use AI tools daily for drafting communications, summarizing lengthy reports, structuring analysis, preparing for business reviews, and thinking through scenarios faster than I could on my own. These aren't glamorous use cases, but the cumulative time savings are real, and they've changed the quality of how I show up in conversations.
My current progress is more ambitious. I'm building toward a workflow where agents handle the mechanical parts of the monthly close cycle — data consolidation, variance flagging, commentary generation — and humans step in at the review and judgment layer. The architecture exists, and testing is ongoing. Getting it right, with the right controls in place, takes time.
I'm sharing this as a work in progress rather than dressing it up as something complete because I think the finance profession has enough people selling the destination. What's useful is hearing what the journey looks like in real time — what's working, what's being figured out, and what's harder than expected.
That's where I am. Building carefully, moving steadily, and not waiting for perfection before I start.
How AI Improves Variance Commentary
A recent change I've made is with variance commentary. Every finance team does it. Every month, without fail, someone sits down and writes two to three pages explaining why actuals were above or below plan. Line by line. It's important work in theory, but in practice, about 80% of it is mechanical. The numbers are already there. The variances are already calculated. Translation is what takes time — turning the data into sentences a non-finance stakeholder can read and act on.
Before we changed this, a senior analyst would spend the better part of a day on commentary alone. Not because they were slow — because that's genuinely how long it took to do properly across multiple cost centers and business lines. And by the time it was done, reviewed, and formatted, the window for the conversation it was supposed to enable had already narrowed.
I built an agent to handle the first draft. It pulls the numbers, identifies the material variances, and generates structured commentary in plain business language. The analyst's job is now to review, refine, and add the judgment layer — the context that only a human with business knowledge can provide.
That affects speed, but it also changes the quality of the conversation. When your analyst isn't exhausted from writing, they show up to the review meeting thinking about what the variance means for the next quarter — not just what happened last month. That's the shift I was after. One day of mechanical writing can become two hours of thinking.
A Framework for What Stays Human
My framework is simple: if it's mechanical, automate it. If it requires judgment, keep it human. That's it.
It sounds obvious, but most finance teams don't operate that way. Humans do mechanical work and call it analysis. AI tools sit on the shelf because no one has clearly defined the boundary.
AI handles data consolidation, reconciliations, variance flagging, first-draft commentary, forecast model updates when assumptions change, standard reporting packs, and cash flow projections based on historical patterns. These are repeatable, rules-based, high-volume tasks. Humans should not spend meaningful time on them anymore. An agent does it faster, more consistently, and without the fatigue that creeps in at the end of a close cycle.
Humans handle everything that requires context, relationships, and judgment. What does this variance mean for the business? Should we reallocate capital to this initiative or hold, given market conditions? What's the right way to present this risk to leadership? Is this forecast directionally credible, or are we just fitting a number to a target? AI can inform all these conversations; it cannot own them.
I draw the line here: AI gives my team information faster and in better shape than before. But the decision — and accountability for it — stays with the person. I'm not working around a limitation; that's the design.
How AI Improves Finance Team Efficiency and Confidence
Positively, the most reliable outcome is time reallocation. Automating mechanical tasks — data pulls, reconciliations, standard reporting, first-draft commentary — gives each person hours back every week. Not theoretical hours, but real ones. If you're intentional about how people use that time, you start seeing better quality analysis, faster responses to business questions, and finance professionals engaged in their work rather than grinding through a close cycle.
Forecast quality also improves — not because AI is smarter than your analysts, but because analysts are less fatigued and have more time to interrogate assumptions rather than just populate the model. Tired people make bad calls. Rested people with clean data make better ones.
Qualitatively, the most meaningful shift is confidence. When a finance team stops spending most energy producing and starts interpreting, they show up differently in business conversations. They have opinions. They push back. That's the cultural change that's hard to measure but impossible to miss.
AI's Biggest Failure Modes in Finance
Now the bad. The biggest failure mode I see is teams automating bad processes. If the underlying data is messy, workflows are broken, or teams don't understand the numbers, AI just accelerates the problem. Garbage in, garbage out, but faster.
The second failure mode is over-reliance on outputs people don't fully understand. AI can produce a beautifully structured forecast, a confident-looking variance analysis, and a clean scenario model. And if the person presenting it can't explain how it was derived or where the assumptions came from — that's a risk. Not a technology risk. A judgment risk. AI should make your team sharper, not more passive. If it's doing the opposite, that's the risk nobody is talking about loudly enough.
The third failure mode is adoption. You can build the most elegant agent, but if the team doesn't trust it or doesn't change how they work with it, it sits unused. Change management is not optional.
Why AI Adoption Requires Redesigned Workflows
When AI doesn't work, it's a people problem.
I've built agents that work. I've seen them tested, validated, and signed off on. Then, I watched teams quietly bypass them, returning to their old methods. Not out of malice, but from habit, comfort, and a genuine belief their manual version was more reliable, even when it wasn't.
That's the gap nobody talks about enough in AI transformation conversations. Everyone focuses on what to build. Very few focus on what must change in the human using the tool for it to truly matter.
Finance professionals, in particular, are trained to be precise and risk-averse. That's a job strength. It becomes a barrier when you ask them to trust a system they didn't build and can't fully see inside. The moment an AI output looks slightly off — even if correct — they default back to doing it themselves. And once that happens a few times, they stop using the agent.
I've learned adoption is not a communication problem. Sending an email saying "we have a new tool, please use it" does nothing. It's a behavior change problem. You must redesign the workflow around the tool, not just drop it into the existing workflow and hope people adjust.
Why Claude Stands Out in Finance Use Cases
The tools are changing fast, and whichever one becomes part of my long-term workflow will have to earn it. Right now, I’m staying curious across all of them and letting the use cases decide.
I use multiple tools, and I think that's the right approach right now. We're in an exploration phase, and every tool is still evolving rapidly. Picking one today and ignoring the rest would be the wrong call.
But if I had to lean one way, Claude has stood out for finance work specifically. The ability to reason through complex scenarios, work with numbers in context, draft structured analysis, and handle nuanced business language makes it particularly useful for the kind of thinking-heavy work that FP&A demands.
The tools are changing fast, and whichever one becomes part of my long-term workflow will have to earn it. Right now, I'm staying curious across all of them and letting the use cases decide.
Don't Confuse Buying AI Tools with AI Transformation
Here's my advice:
- Stop waiting for the perfect enterprise strategy before moving. Most CFOs I speak to are sitting on a transformation roadmap that has been under review for six months. Meanwhile, their teams still spend half their week on work they could automate today. The window is open right now. Use it.
- Don't confuse buying tools with achieving transformation. I've seen organizations spend significant money on platforms, licenses, and consultants and come out the other side with the same workflows they started with, just more expensive. Technology is not the hard part. The hard part is honestly assessing how your team works and being willing to change it.
- Invest in your people before investing in tools. A finance professional who understands what AI can and cannot do, who prompts well, who critically reviews AI output and adds the judgment layer — that person is worth more than any platform you buy. Build that capability in your team. It compounds.
- Get your data house in order. AI will expose every crack in your data foundation faster than anything else. If your numbers aren't clean, your processes aren't documented, and your systems don't talk to each other — fix that first. Otherwise, you're automating chaos.
- The CFO's job is changing whether you engage with AI or not. The question is whether you shape that change or react to it. The finance leaders who will matter in five years are the ones curious enough to experiment, humble enough to learn from what doesn't work, and decisive enough to build something.
Don't just have an opinion about AI. Do something with it.
Follow Along
You can follow Sibaranjan Patnaik's work on LinkedIn.
More expert interviews to come on The CFO Club!
