Meta takes the lead on AI in finance

At Meta’s International Headquarters in Dublin, Majella Mungovan’s finance team is already reaping the rewards of using artificial intelligence in day-to-day operations

The hype surrounding artificial intelligence (AI) continues to gather pace, as professionals across all sectors consider its potential impact on our future working lives.

While many accountants grapple with the scope and reach of this emerging technology, however, Meta – the US-headquartered social media giant – is already several years into using AI in finance processes at its International Headquarters in Dublin. 

“Five years ago, we decided we were going to really focus on using machine learning to drive efficiency across our finance team,” explains Meta’s Majella Mungovan, FCA.

As Vice President of Financial Operations with Meta in Dublin, Mungovan leads a 150-strong global finance team.

“Meta has four large finance operations around the world, one of which is in Dublin, where we serve people globally supported by several hundred people at our outsourcing partner,” Mungovan explains.

“We are responsible for all activities relating to revenue as well as everything from accounting and reporting to financial operations, and risk and operational assessments to collections. 

“Running a very large-scale operations team means we deal with many millions of transactions every year and doing this in a really efficient and scaled way is very important.”

Growth, speed and efficiency

The move to incorporate AI into processes at Meta’s financial operation in Dublin coincided with a period of intense growth for the company globally, Mungovan explains.
“Meta achieved revenue of $100 billion faster than any other company in history, so we have gone through an enormous growth phase over the past 10 years,” she says.

“When your company is undergoing explosive growth like this, speed, efficiency and scale are essential. Tasks you might be able to do manually in a slower moving environment have to be done faster and, for us, this meant looking at new technology to help us reach our goals.”

Founded in February 2004, originally under the name Facebook, Meta opened its first Irish office in Dublin in 2009 with a team of just 25. 

Now, with over 2,000 employees across 80 teams, the company’s new International Headquarters opened in Ballsbridge in Dublin in October 2023. It has additional sites in Co. Meath, where its data centre is located, and in Co. Cork, home to Meta’s Reality Labs.  

Since its launch in Ireland, Meta has also undergone rapid growth globally, acquiring Instagram in 2012 followed by WhatsApp two years later.

Employing over 66,000 people around the world, the tech giant continues to record milestones, with Facebook’s daily active users reaching a mammoth two billion in February 2024.

Automation: first steps

“When our global finance team in Dublin started looking at how we might use technology to help manage the sheer volume of work we were dealing with, our first step was to consider very basic automation rules,” Mungovan says.

“We built some fairly rudimentary machine learning models that could make some decisions on our behalf. Each machine learning model is essentially an algorithm. 

“You ‘plug in’ a large number of criteria to assess whether or not a decision needs to be made and the model learns and improves over time. 

“Our first use case was the credit decisioning process, using internal and external data related to customer-recommended decisions. Over time, our machine learning models have become increasingly sophisticated to the point now where they can reliably cover the vast majority of our decisions where they have been rolled out.” 

Mungovan’s team has also been exploring how natural language processing might be used to automate some parts of the revenue processes used in customer support.

“We are operating in a very heavily automated world. In the last year in particular, we’ve been able to explore new technology Gen AI and this has allowed us to really accelerate the progress we’ve been making in automation over the last five years,” she says.

“We can now move to touchless processes and transactions in a much more complete and efficient way – for example, with the helpdesk for our finance team. 

“When a customer gets in touch and says, ‘I need help with my invoice,’ we can plug in different AI agents so we can see who the customer is and what kind of problem they are facing with their invoice. 

“The agents can read the customer’s messages and communicate with them, in many cases resolving the issue and, in others, ensuring the query reaches the right people so it can be resolved and the ticket closed out quickly.”

Accuracy and speed are essential when it comes to customer care. “Our priority is that the customer gets the answer they need as quickly as possible and that, at the same time, we are operating as efficiently as we can in resolving issues before they escalate or cause friction internally,” Mungovan says.

Her team is also now using machine learning for cash flow forecasting. “This helps us to understand the customer’s payment behaviour,” Mungovan explains. “If there is any deviation from that, we can very quickly and accurately predict what free cash flow will look like across the company.”

What to expect

Based on her experience working with AI and machine learning for the first time, Mungovan says that careful preparation is a must at the outset.

“It’s a huge learning curve for everyone involved, particularly those of us from a finance background who have to get to grips with a new technology that, in turn, can have a big impact on how we do our work, and on our capabilities,” she says.

“My advice is to get out there and find out what other professionals and organisations are doing. Attend conferences and other events, read papers and case studies. Keep an eye on what people are sharing and reach out and ask questions.”

Preparing to introduce AI for the first time will likely take “a lot longer” than you expect, she adds. 

“You’re going to have to bring a lot of people through the process and everyone will want to make sure that the new model is working and fit-for-purpose before it’s introduced into the ‘live’ working environment. You’re looking at a learning phase of at least 12 months before you can expect to see any kind of return-on-investment.”

Machine learning models need to learn and that takes time, Mungovan says. “They tend to generate a lot of false positives at the outset. It probably took us two years to get to a place where our models were really starting to generate a decent return-on-investment, but once we had some traction, they evolved very quickly after that. 

“Now, we regard the project beyond its ability to deliver greater efficiencies; we also think about it from an assurance perspective. We have monitoring programmes running continuously in the background, looking for anomalies, exceptions and errors. 

Now that Meta’s finance team in Dublin is using AI day-to-day, Mungovan is confident that the technology will play an even bigger role in the years ahead.

“Our large language models are becoming more and more helpful to us. The technology environment continues to evolve all the time. What we have now, we didn’t have six months ago. It’s quite extraordinary.”

North Star

Mungovan’s North Star is, she says, that all finance processes become “as ‘touch pointless’ as possible”.

“I want my team examining anomalies and fixing issues at root, rather than having to deal individually with problems as they arise – right across the board from calculation and booking accounting entries to reconciliations, preparing commentary, analysing the movement on a general ledger account or managing expense categories.”

Based on her own experience implementing AI, Mungovan believes the technology has great potential to elevate the role of Chartered Accountants and other financial professionals in the future.

“I think the way we’re using technology in our own finance team at Meta really convinces me that AI and technological advancements will create more senior and sophisticated roles across finance and other functions,” she says. 

“Meta is a very large company with ambitions to become an even larger company. We are growing so rapidly that we need to figure out efficient ways of scaling. AI and machine learning is delivering these efficiencies for our finance team. It is making us faster and improving our capabilities.” 

Opportunity for the profession

Mungovan believes the use of AI in accounting will become “the norm” in the years ahead as more and more organisations and professionals adopt the technology to support and enhance the finance function.

“I remember when I was training to become a Chartered Accountant, people then were asking the same questions about technology and how it would affect the future of the profession, the jobs we do and the way we work. At that time, the big focus was Microsoft Excel and how it was going to reshape accounting norms,” she says.

“I view AI in a similar light today. Over time, AI as a tool will fundamentally change the role of the accountant in the same way Excel transformed how things were done 20 or 30 years ago. 

“AI will have the same kind of impact. It won’t replace the role of the accountant, but it will become a widely used tool, which will allow us to be more effective in our jobs. 

“Ultimately, I think AI is something to be embraced rather than feared. I am really excited about the possibilities this technology will offer our profession. Rather than be frightened, people should see opportunity – and I think this opportunity will be immense.”