Three questions that help you find the routine worth automating first.
Many CFOs are interested in AI, but few are sure where to start.
The starting point is often a tool. You buy ChatGPT licences or switch on a new AI feature in your accounting software and wait for time savings to appear. But a tool doesn't tell you what it is worth using for.
A better starting point is the work itself. Once you know where your team's time goes, the choice is usually easy. Below are three questions that help you find the first routine AI can take over.
1. Where does your team's time really go?
If time isn't measured, everything feels urgent. Then the automation easily goes into the work that annoys people most, not the work that takes the most hours.
List the recurring work in your finance team and estimate the hours each takes per month. A rough estimate is enough. The list usually includes at least the monthly close, purchase invoice processing, reconciliations, management reporting and VAT returns.
Often one or two work steps take clearly more time than the rest. They are the first candidates for automation.
2. Is the work done the same way every time?
AI automation fits best with work that repeats weekly or monthly and follows the same rules every time. A good sign is that you can explain the work to a new employee with a few instructions.
At the Pesonen accounting firm, for example, the tax account of every client company was checked manually in OmaVero, the Finnish Tax Administration's online service, every month. For a single company the work wasn't much, but it repeated for every company every month in almost the same way. Now the automation does the check, and the person in charge only sees the companies that need action. Read the full customer story.
Not everything needs to be fully automated, though. If the work has a step that requires judgement, the automation can prepare the material and a person makes the decision. Posting proposals and exception detection, for example, can be handed to AI while approval stays with an expert.
3. How much would the automation save, and what would it cost?
Once you have a candidate, work out its value. Monthly hours multiplied by the hourly cost give you the current cost. From that, subtract the work that remains after automation and the monthly maintenance fee.
An example calculation might look like this:
- Cost now: 725 € / month
- Time saved: 90 %
- Savings: 653 € / month
- Maintenance: 20 € / month
- Added value: about 633 € / month
The calculation also shows how quickly the automation's build fee pays for itself. If the payback period gets long, choose another process.
Start with one process
Once you have gone through the questions, choose one process that takes a lot of time and repeats the same way. In many companies it is purchase invoice processing or the reconciliations in the monthly close.
You don't need a big systems project. The automation can be built on top of the software you already use, including older systems that have no API.
Once the first automation works, the next one is easier to build on the same foundation. That is exactly how Pesonen went about it. First came the OmaVero reconciliation, and then a digest of the Tax Administration's information requests on the same foundation.
”We have moved forward one step at a time, which has given us an understanding of how AI automation works as well”, says Ismo Jukola, CEO of the accounting firm (translated from Finnish).
If you want to find out what your finance team in particular should automate first, we go through it in a free AI roadmap. After a one hour interview you see which processes are worth automating and how much value AI brings you. You can also see what can be automated in finance teams.