)
)
Digitalisation, automation, and artificial intelligence are now a fixture of modern corporate life. We now expect to see process automation in product, marketing, and sales teams - and obviously in most factories and engineering sites.
But we don’t talk enough about how valuable it can be for finance teams. Most companies still rely on a huge number of manual processes to reconcile payments and close the books.
Accounting, in particular, has huge potential to become more efficient and do more “in real time.” Fully automated closing is the dream for many CFOs and finance teams, and thanks to smarter tools, some are nearly there.
In this article, we’ll explore machine learning - one very useful subsection of artificial intelligence. I’ll explain what makes it a step up from rule-based automated tasks, and how it’s already making life easier for thousands of finance teams.
What is machine learning?
Machine learning is a kind of artificial intelligence which asks machines to think, behave, and learn the way that humans do.
Computer algorithms identify and learn patterns in data and use them to make more accurate decisions in similar scenarios. Machine learning is used to build automation systems and perform repetitive tasks faster and more efficiently, in a way that mimics human actions.
This is in contrast with a standard piece of software. Normal software will repeat the same commands over and over, hopefully without making mistakes along the way. Every time the same, always predictable.
The performance of machine learning tools improves with time. It learns the patterns from historical data; the more data there is, the better its performance. We give them feedback as they go, and they improve as a result.
Why is machine learning so valuable in finance?
As a basic rule, any repetitive task is just begging to be automated. The more tedious, the more people can benefit from automating these tasks.
A finance team’s mission is by no means tedious - it’s critical to a healthy, growing company. In fact, finance departments are increasingly working on strategic tasks that inform key decision makers within companies. But today, achieving this mission typically involves a lot of repetitive, frustrating busy work.
Let’s use a specific bookkeeping example: categorising payables. Every payable needs an associated expense and VAT account, both for tax reasons and to help the business see which activities are the most costly. This is very important.
But the actual work of categorising each payable takes time and adds very little value to the business. What’s more, there are clear patterns: this supplier is typically associated with that expense account and this VAT account. The same goes for cost centers and descriptions!
95% of these payments are pretty clear - you don’t need highly skilled, nuanced-thinking finance team members to make these decisions. But because there are right and wrong answers from time to time, machine learning is a perfect fit.
ML models recognise and learn from patterns to make mathematical predictions for the expense and VAT accountper payable. The more payables you process, the more the model learns about your payable structure and the better it becomes at assigning the right accounts.
So unlike classic automation - which is already very useful in finance - every time you correct the machine learning algorithm or tool, it makes a better choice next time around.
Finance teams are too valuable to spend time on repetitive tasks
Finance is at the heart of every good business. You have the central vantage point, analytical skills, and raw data to help everyone succeed.
And every minute spent fixing basic errors and categorizing cost centers takes you away from adding value to the company.
And despite your impeccable attention to detail and financial savvy, you’re not designed to perform endless repeated tasks over and over. In fact, that’s exactly what machines are built for. So it makes perfect sense to leave these to them.
How Spendesk uses machine learning
For some even more tangible examples, I’ll take you through how we’ve added machine learning to our spend management software. Spendesk lets companies centralize their “non-payroll spend” - predominantly card payments, invoices, employee expense claims. We put all of these in one place, so employees and finance teams don’t have to hunt all over for receipts and payment details.
We don’t only want to change how people spend money at work, but also how efficient the month-end closing process is for finance and accounting teams. This process is slow precisely because of how repetitive it is, and also because of errors and missing documents. But in theory, it’s a relatively straightforward exercise.
Classic automation gets us a lot of the way to a faster close. Since the early days, Spendesk has automated the following:
Fraud checks (ensuring expenses claimed match the receipt)
VAT detection (identifying the tax portion of a receipt or invoice via OCR technology)
Email and Slack reminders to contributors to upload receipts (a huge timesaver for finance teams)
Adding machine learning to the mix makes these tools even more effective. Particularly when allocating VAT and expense accounts.
It’s relatively easy to set standard rules for expense accounts. For example, every payment made to Easyjet or Eurostar can safely fall under “T&E Expenses.” But what about the edge cases, where more nuance is needed?
ML models learn based on historical data and predict values when they are very confident for a specific expense or VAT account. Early on, they can make mistakes which accountants will catch during reconciliations. The models learn from these corrections so you don’t need to fix them a second time.
In a nutshell this means that if you do your monthly closing and pre-bookkeeping with Spendesk, the process gets faster and more automated month over month!
The tangible benefits for finance teams
As with virtually all automated processes, the initial benefits are fairly obvious:
You save time since you don’t need to manually assign expense accounts to every payment. Reconciling card payments, expense claims, and invoices is up to 10 times faster in most cases.
You see fewer errors. Humans are a little more flawed than we like to think, and machines almost always do repetitive tasks better than us.
You stay up to date, with expense accounts and VAT updated in real time. This is a major benefit - most finance teams process payments once a month at most. Which means they don’t know what’s been spent until the end of the month. But the machines can do it 24/7!
And there are soft benefits that we tend to overlook. But these are no less important to finance professionals:
Less aggravation since you don’t have to deal with errors and missing receipts. Life is a lot more enjoyable without those little nuisances.
Better collaboration with other teams, because you’re not always bugging them for little things. Instead, you offer insights and contribute to their goals.
More value added to the company as a whole. Finance teams really come into their own when they can focus on what really matters.
Give your finance team the freedom to succeed
It’s really a no-brainer. Adding machine learning to your operations is basically the shortcut to becoming a highly impactful finance function.
As soon as your skilled team moves away from manual processes, they can become the analytical, insightful powerhouse every growing company needs.
The more you can automate - and the smarter and more nuanced that automation becomes - the better your finance team will be. We work with thousands of finance teams every day, and not a single one wishes they were back doing data entry.
There’s really nothing stopping you from taking this step. Get started today.
)
)
)
)
)