Five things AI is already doing in financial management

Technology does a tremendous amount of work on our behalf. The goal of development is simply to make everyday life easier, and according to Sakari Jorma, the change over the past few years has been massive.

When you ask Finago’s Chief Technology and Product Officer Sakari Jorma what AI is actually doing in financial management right now, he promises to summarise recent software development into five key points. But first, he asks a counter-question.

“Do you remember when you had to go to the bank with your bankbook to withdraw cash, and pay bills at the counter? Now you get cash from an ATM with a card, if you even need cash at all, given that both bills and purchases are paid by phone. Accounting is going through the same kind of transformation.”

Sakari Jorma oversees product development and maintenance across Finago solutions, from AI to information security. Continuous development aims to make life easier for businesses and entrepreneurs.

Jorma describes himself as a technologist, always thinking about how things could be automated and done more easily.

“Technology has become an increasingly defining part of how people operate. Solutions already do a vast amount of work on our behalf, and their user experience keeps improving.”

1. Manual work has decreased dramatically

Bookkeeping is statutory, standardised, and repetitive. It follows the same rules for every company and is regulated to ensure comparability, which makes it an ideal candidate for automation.

“Over 15 years, we’ve gone from physically carrying receipts to an accounting firm and someone typing them in there, to data transferring automatically from a payment terminal to bookkeeping without anyone touching anything,” Jorma says.

The transformation is massive. You can see it with your own eyes. The shelves that used to be full of binders have simply disappeared from accounting firms.

Technological infrastructure, comprehensive wireless networks and affordable mobile connectivity, has made the Nordic countries leaders in applying technology to financial management. This level of digitalisation in accounting doesn’t exist elsewhere yet.

“We’re genuine pioneers. There are still European countries where people pay in cash and receive handwritten receipts as documentation,” Jorma says.

2. Work roles are gradually changing

The goal of development isn’t to take anyone’s job, but to make work more meaningful. AI is on the side of work satisfaction and professional growth, as long as you keep up with the development.

Jorma knows there are still plenty of clients who want to bring their receipts once a month to their familiar bookkeeper, have a cup of coffee, and share how they think their business is going. That will change eventually, if only through generational turnover.

“Scanning has been around for 20 years. At first we were pleasantly surprised to save on postage; now we’re saving working hours. Nobody spends time opening email attachments, scrolling through invoices, and typing in data anymore. AI does it in fractions of a second, transfers the data directly to the accounting system, where a person checks and approves the invoices for payment,” Jorma says.

That time saving frees the accountant for what matters: analysing the data that’s been collected. Accountants have shifted from office workers to financial advisors, and financial management has become the nerve centre of the business. That, Jorma says, is what accounting work should really be.

“Less clicking, more thinking. The system works for you. When AI reads data and runs queries, you get the current state of your business in an instant, forecasts for the coming months, or the answer to who your best customer is and how much you invoice them each month. That’s what actually matters. Not whether all the petrol receipts are accounted for,” Jorma says.

3. Repetitive tasks are disappearing

There’s no point developing anything just for the sake of it, Jorma says. The relevant questions are: what is the benefit, and who does it benefit?

The accounting charts in the Nordic countries are extremely logical. Travel expenses, rent, and phone costs are posted to the same accounts in exactly the same way across every company, often every single month. AI learns, pre-posts the invoices, and a person simply checks and approves.

“Recurring payments are the best candidates for automation. Why on earth would a person do the same job over and over again? Automation has been doing this for a long time. It was one of the very first tasks for AI and machine learning,” Jorma says.

Simply put, accounting is formulas, mathematics, and putting things in order.

“That’s why software needs to function simply too. Entrepreneurs keep the societies going; they need to be able to focus on their business. We free up time for entrepreneurship by handling the accounting routines,” Jorma says.

4. Errors are reducing further

However careful a person is, they’re never as precise as a machine. When it comes to finding, correcting, and flagging errors, a machine never loses focus. It picks up even small numerical discrepancies more reliably than a human.

The linguistic capabilities of AI are also limitless, and genuinely useful in financial software.

“Whatever language a receipt or invoice is in, AI can read it. You can run a query in our analytics tools in Dutch, for example, and the software responds in whatever language you’re asking in, regardless of what language the underlying data is in. The risk of misunderstanding goes down and the process speeds up, because you don’t need a separate translator,” Jorma says.

Development work is continuous. One significant AI feature currently in development, requested by Finago’s Finnish customers, is the interpretation of collective labour agreements. Finland has 180 different collective agreements in force, and no one knows every detail of each one. Yet nobody wants wages paid incorrectly.

“If an employee says they worked 32.5 hours this week and the entrepreneur asks how their pay should be calculated, or whether they’re accumulating additional leave days, the AI agent interprets the collective agreement and gives the payroll calculator the correct answer straight away,” Jorma says.

5. The amount of real-time data keeps increasing

Younger generations’ culture of wanting everything immediately has spread to older ones too. Answers need to be there instantly. Waiting a moment is already too long, let alone weeks.

Information is an advantage, and the more real-time it is, the more valuable it is. Reliable forecasts are worth more than anything.

Transactions tend to follow recurring patterns, and where there’s a pattern, you can predict what’s coming. That’s AI’s greatest capability: processing and analysing data.

“We think a lot about whether we can build automated actions on the basis of AI analysis and potential recommendations, giving alerts and warnings or even suggesting course corrections. For now, though, human judgement is still better than AI in these situations. Humans have a broader neural network and can question things,” Jorma says.

He says he always reads reports and analyses as a sceptic. If you don’t, you end up dumber than the machine.

Data is only valuable if it’s also reliable and correct. The challenge with large language models, Jorma says, is how convincing they are. AI is forced to give an answer. If it can’t find one, it’ll make one up.

Users need to keep asking for sources and reasoning and stay critical of the conclusions AI draws.

“The human checks and approves,” Jorma repeats, as he has done throughout the conversation. That remains the most important human role right now. Building too much automation on top of language model analysis isn’t advisable yet.

“When something sounds easy and natural, it becomes believable. What if I’m just telling stories and making it up?” Jorma asks with a big smile.