CFOs get up to speed with AI

Helene von Roeder, CFO, Merck, Brian Montgomery, Senior Director, Workday and Martin Rosendahl, Senior Partner, McKinsey & Company shared their thoughts at the FT webinar “AI and the changing role of the finance function”. RAID Director Ben Avison reports.

Finance has been a late adopter of AI as compared to some other functions in an organisation, but that is changing quickly, according to Martin Rosendahl, Senior Partner, McKinsey & Company.

“Despite overpromises historically, there is a recognition that AI is a significant opportunity. If you can make a decision three or four weeks quicker, that is invaluable for any organisation. We see CFOs rapidly engaging now and that’s an exciting evolution,” he said.

Helene von Roeder, Chief Financial Officer at Merck cited a range of applications where AI can bring benefits , especially when it comes to unstructured data. For example, automation, machine learning or mathematical statical models have a much better use compared to AI when you are working with structured data, she said. “AI is super helpful when it comes to unstructured data. It’s also great where you have the ability to correct as a human. But anything where we need 100% confidence, accuracy in the data, we cannot use AI there just yet.”

Brian Montgomery, Senior Director, Workday highlighted useful applications of AI, such as cash flow prediction. “AI is very good at bringing in additional external data to enhance finance teams’ ability to predict cash flow. AI is starting to look very powerful in areas like that,” he said.

Quality in, quality out

One thing all panellists agreed on what the importance of data quality. “The cleaner and more real-time our data is, the more we can spot abnormalities,” said von Roeder.

“Data is key to all AI – it’s the quality of data that’s going to make or break you,” said Montgomery. “As finance functions, we have a lot of high-quality data already in our systems. The risk is where interpretation is needed, or if you’re looking at unstructured data.”

The question of how AI might help make unstructured data useable is seen by all panellists as “the holy grail”.

Bespoke vs. off-the-shelf AI

Another big question is when to develop in-house AI and when to deploy readymade solutions.

“We have our own Workday AI that we’ve been working on for a very long time. We have the largest HR dataset of any company in the world,” said Montgomery.

“We also know that different companies are going to want to use different AI agents. A lot of big players are building some great tech. We’re not huge fans of bespoke solutions; however, something that is standardised and configurable can be very powerful.”

Von Roeder added: “I’m sceptical about the different embedded AI solutions because it doesn’t allow you to use best in class when it comes to using AI for your purposes.”

Data privacy is an important growing concern when selecting an AI system. Where Merck uses external AI code, the data behind it is stored on Merck’s own systems. They can’t use Deep Seek, for example, as it makes it clear the data will go on a server in China.

AI agents as digital employees

There is a lot of excitement around agentic AI, but there are caveats.

“An agent is like a digital employee,” said Montgomery. “With real employees you don’t let them loose across the whole organisation, and there’s a  lot of onboarding. AI agents will be onboarded and trained; there will be governance and a clear business process – this applies to a virtual employee just as much as to an individual employee. It’s about enhancing human beings in what they do in their work.”

A particularly exciting area is how AI agents might help to create data. “It’s actually humans that create data,” von Roeder pointed out. “Using agentic AI to help humans to put the right data in is going to be groundbreaking in my opinion. This will really push this adoption to the next level.”

Advice on AI adoption

Martin Rosendahl, Senior Partner, McKinsey & Company: “The financial data cube is really critical part of getting any gen AI initiative right, to ensure that financial data is accessible, and usable in the best way. Be sure to do your homework on how the financial data cube works.”

Brian Montgomery, Senior Director, Workday: “Use AI to assist with things you are already an expert, in so you’re not going to be caught out by hallucinations.”

Helene von Roeder, CFO, Merck: “I am relentlessly focusing on getting clean data and end-to-end processes as well as upskilling our organization in the right use of AI.” For new joiners and young talents entering the field, her recommendation was to “stay curious. It may seem cool and complex, but it is actually very simple. There is no magic in it.”

This article is based on comments made on the FT webinar “AI and the changing role of the finance function”, expertly moderated by Elaine Moore, Tech Comment Editor at the Financial Times