Carla Aranda, a global finance executive with 25+ years of experience at companies including L’Oréal, breaks down her advice for measurable AI success
Embedding AI is the top priority for CFOs across the globe
Over three-quarters are aiming to move beyond basic applications to some form of autonomous decision-making, according to Deloitte. At the same time, nearly two-thirds expect AI complexity to rise in 2027.
This is a familiar story for finance leaders getting to grips with AI: A simple test succeeds, prompting it to be used for more nuanced tasks without careful implementation. This causes the opposite of the desired impact: more complexity, more confusion, more noise.
Carla Aranda, a global finance executive with over 25 years of experience at companies including beauty giants L’Oréal and Revlon, has observed this “trap” in action.
Speaking to The Exchange, she sets out how finance leaders make measurably faster and better business decisions with AI by emphasizing simplification and clarity.
1. Where do CFOs typically start their AI ‘journey’?
Most finance leaders start where it feels safest: repetitive, rules-based work like reconciliations, variance commentary, and first drafts of reports.
The trap comes right after. If you plug AI into a process that’s already complex, you get complex outputs, just faster. Before I add any technology, I look at what I can remove from the process first. Subtract, then enable.
2. Are they leaving more impactful applications on the table?
Yes, many are leaving the bigger prize on the table. The most valuable use is seeing what a human can’t: patterns across millions of rows, scenarios you’d never have time to run by hand, and early signals buried in the data. That’s where AI starts changing decisions.
One more thing. A process that previously worked with 100% humans in the loop needs a fresh look once AI is in it. Who reviews, who signs off, and what gets checked all need to be redesigned on purpose.
3. Finance leaders are often flooded with data and reports. What are the risks?
I’ve lived this many times in my career.
Finance has a habit of producing more reports and more analysis than anyone can use. We once ran an analysis of the number of business intelligence (BI) dashboards we were producing across the company. The number was astounding: 1,500 dashboards. I was probably on the receiving end of at least half of them. It was a great exercise to start sunsetting most of them. Just imagine the amount of manpower that was released by doing this.
The risks are real. Your best people spend their days producing reports, with little time left to think. Leaders get buried in information, and more data can pull you down a rabbit hole. The whole organization slows down, because every decision waits for one more cut of the numbers.
Before producing any analysis, ask what decision it’s meant to support. If there isn’t one, don’t build it.
The one thing I’ve learned is that 80% clarity today is more valuable than 100% certainty next month. Waiting for perfect data feels safe, but by the time it arrives, the opportunity may be gone. Numbers need a story to drive action, and a pile of dashboards rarely tells one.
4. What have you learned about using AI to simplify and enable better business decisions?
My biggest learning is that AI rewards clarity. Before I bring it into a process, I define the ultimate goal of that process and strip out the steps that don’t serve it. Only then does the tool become meaningful.
It’s just as important to decide upfront where a human reviews the work. AI is a fantastic tool, and it can also be confidently wrong. I learned that in a very ordinary way. I was helping someone map out a number of locations we wanted to visit in the city. Time was tight, so I used an LLM to create a map. Only when I arrived at the first location did I discover that the streets didn’t make sense, and neither did the directions or the blocks. We ended up pulling out our phones and using old-fashioned Google Maps. In that personal situation, the stakes were low. Imagine the same mistake in a forecast or a board deck. It could be expensive.
When complexity and uncertainty are high, those human checkpoints are where your team’s judgment protects the business.
5. What is the ‘banana test’ and how does it help finance leaders know when to use AI?
It’s called the banana test because of the shape your data takes when you plot it. Most points fall along a curve that looks like a banana, and the anomalies are the ones that sit off it.
With a manageable data set, your sharpest analyst can plot it and spot them quickly. That’s not an AI problem. But when the data runs into millions of rows across brands, customers, and promotions, there is no single chart she can look at. If they can’t spot the anomaly with their own eyes, you’ve reached the limit of what the human brain can do with it.
That’s where AI earns its place. Don’t use it to spot the obvious. Use it where the human eye will miss something.
6. Can you share an example?
I applied this when I was measuring the ROI of customer discounts.
We had hundreds of promotions running every month for each brand. Multiply that by brands and customers, and we were looking at millions of rows of data.
My smartest team member would have needed months to dissect it. So, we built a regression model that quickly showed which promotions were profitable, and which weren’t, and moved those dollars into what actually generated profit.
7. How can finance leaders make sure they don’t outsource their judgment?
I recently heard the term ‘AI laziness,’ and it’s so true.
When something makes life easier, it slips into our day-to-day until we depend on it. It happened with the first smartphones, and it’s happening with AI.
For a finance team, the risk is quiet. Nobody stops thinking overnight. The muscle just weakens, until one day nobody can explain why the number is what it is.
8. Are there practical steps to avoid knowledge atrophy?
A few habits help. Ask people to explain the output in their own words before it goes anywhere. Keep the human checkpoints you designed, and rotate who owns them. Have junior team members build some analysis by hand so they learn what good looks like. As a leader, ask ‘how do you know?’ often enough that it becomes part of the culture.
Above all, keep reminding your team that they are still part of the process, and that their judgment is what makes the checks and balances work.
9. When delivering simplicity, how do you avoid losing the nuance of global financial leadership?
I started my career in Argentina, then Brazil, before moving to the US, so I learned early on that the same metric can mean very different things in different markets, especially in hyperinflation environments. There, 40% sales growth can still mean you sold fewer units than last year.
I protect nuance by simplifying the structure and leaving room for the story. Globally, everyone speaks the same language: the same few KPIs, the same definitions, the same calendar. Locally, the team owns the explanation behind the number, because they know the customer, the currency, and the competitor down the street.
When I cut a report, I ask whether anyone would make a different decision without it. If the answer is yes, that’s nuance, and it stays. If the answer is no, it was complexity all along.
10. What fundamentals would you share with a CFO aiming for a simple but impactful AI transformation?
Start with the decisions that matter most to the business, and work backward to the processes that feed them. That’s where AI pays off.
Subtract before you automate. Every process carries steps nobody remembers asking for. Remove them first, so the technology is enabling something clean.
Start by picking one painful process and win there. An early win builds more trust than a roadmap with twenty pilots.
Design the human checkpoints from day one, along with the governance that goes with them.
And keep your people at the center. Tools are enablers. The transformation works when the team understands why it’s happening and sees what’s in it for them: less low-value work, and more time for the work that made them choose finance in the first place.
11. How should CFOs benchmark successful AI adoption? What does good look like?
Good, to me, is when AI takes over your low-value work (who likes it?) and frees up your time and your team’s time for strategic, high-value work.
To benchmark it, I’d look at four things:
Where the hours go: is the team spending less time preparing numbers and more time explaining them to the business?
Decision speed: is the time from question to decision getting shorter?
Output: are you producing fewer reports, and are more people actually using them?
Trust: does the business rely on what finance tells them, and can your team explain every number AI touched?
If those four are moving in the right direction, AI is doing its job.
____
Carla Aranda is a global finance executive with more than 25 years of experience building a finance career across continents, industries, and company sizes, from a small local business in Argentina to global beauty companies like L’Oréal and Revlon. She writes and speaks about how simplification and AI can help finance leaders turn information into faster, better business decisions.
