Many SME finance teams now have access to AI tools such as Microsoft Copilot and ChatGPT, but knowing where to start is harder. In this article, Matt Swift, Finance AI & Automation Consultant at Yorkshire-based finance recruitment and consultancy business Headstar, shares what he has learned from introducing automation in Headstar’s own finance team and working with other SMEs. He explains how to find the repetitive work worth automating first, when straightforward automation is enough and when AI is useful, and why human judgement should stay at the centre of finance.

AI has become much easier for SMEs to access. Many businesses already have tools such as Microsoft Copilot available, while ChatGPT and other platforms can be set up in minutes.

But access to the technology is not really the problem. The question heard much more often from finance teams is: what should we actually use it for?

Having spent the last year looking at this in our own business and with other finance teams, I’ve found it’s much more useful to start by looking at the work your finance team is already doing and where people are spending time on repetitive processes that don’t really need a person involved at every stage.

Start by finding where the time goes

Most finance teams have processes that have built up over the years. A spreadsheet gets updated every morning. Someone copies information from one system into another. Reports are pulled together manually every month because that’s simply how they’ve always been done.

Individually, those tasks might only take 15 or 30 minutes. Across a week, month or whole team, they quickly add up.

The processes to automate first tend to be regular, repetitive and predictable. They often involve moving, checking or reformatting information, without needing much judgement at each step.

A useful question is: Which regular tasks are taking more time than they need to?

That’s usually a better place to start than asking: “What can we use AI for?”

Look for the repetitive work first

We started this process in our own business by looking at the tasks people were doing repeatedly and asking whether every step really needed to be manual.

That included a report we rebuilt by hand each week, duplicate data entry across several spreadsheets, timesheet hours being keyed in manually, credit control chasing and process documentation.

Across the finance team, the changes we’ve introduced are now saving around 20 to 30 hours a month.

We found each of those opportunities by looking at a process that was taking too much time and working out how we could improve it.

If someone is spending time every morning pulling together information for a cashflow report, there might be an opportunity to automate how that information is collected and presented.

If someone is sending the same chaser emails by hand every week, that’s worth looking at.

We now pull the debtor list from Xero every day, send staged reminders automatically and copy in the right people when something escalates. That saves around 6.5 hours a month.

If someone spends time every week rebuilding the same report, there may be an opportunity to build it once and automate the updates. Ours took around 90 minutes a week to put together. It now updates automatically rather than being rebuilt each week.

If the same information is being typed into several spreadsheets, look at whether it can be entered once and used elsewhere. We did this with our forecast, so the underlying data now feeds through automatically instead of being keyed in and reconciled by hand.

Other areas worth looking at include supplier statement checks, reporting packs and drafting monthly commentary for someone in finance to review.

Start by looking at the process itself, then work out which technology can improve it.

Not every automation needs AI

Automation and AI are not the same thing.

Sometimes the best solution is a straightforward workflow. It might be an Excel formula, Power Query, Power Automate or simply a better way of moving information between systems.

You don’t need AI involved in every process.

AI becomes particularly useful when you’re dealing with less structured information, such as written documents, or where information needs summarising or interpreting.

As a simple rule, if a process follows the same steps every time, traditional automation may be enough. If it involves language or interpretation, AI may have a role.

Often, the best solution is a combination of the two.

Do not automate the judgement out of finance

Just because something can be automated doesn’t mean it should be.

Finance teams still need people making decisions, questioning numbers and understanding the wider commercial picture.

Automation can take away a lot of the manual work around those decisions, while leaving the judgement with the person who understands the business.

Take credit control. Our chasers now go out on a schedule, and escalations get copied to the right people. But the automation doesn’t decide what to do when a key client pushes back or a relationship is under strain. That still needs an experienced person.

The same applies to reporting. We’ve found AI can produce a useful first draft of commentary from cashflow data. But someone in finance still needs to check it’s accurate and reflects what’s actually happening in the business.

Human review matters most where mistakes could affect payments, reporting, customers or key decisions. That’s why every automated credit control chase we send is logged, so someone can see exactly what went out and when.

Pick one process

One of the biggest mistakes businesses can make is trying to tackle too much at once.

You don’t need an AI strategy covering the whole finance function. Start by spending a week looking at where the team’s time actually goes.

Ask each person to note down the repetitive tasks they carry out, how often they do them and roughly how long they take.

Then ask three questions:

• Which tasks take the most time?
• Which are the most repetitive?
• Which carry relatively little risk if part of the process is automated?

That should give you a shortlist.

Then pick one.

You’ll learn far more from improving a real process than from spending months discussing what AI might eventually do for the business.

The technology is the second question

There’s a lot of pressure on businesses to start using AI, but that can lead to the conversation starting in the wrong place.

Do not rush into choosing Copilot, ChatGPT or any other platform.

Start with your finance team.

Spend some time looking at what your finance team is actually doing. Where are people still manually copying information between systems, rebuilding the same reports or carrying out repetitive tasks that take hours every month?

Identify those processes and start there. Once you understand the problem you’re trying to solve, it becomes much easier to decide whether you need AI, straightforward automation or simply a better process.

You don’t need to transform everything at once. Find one thing that’s wasting time and see if you can make it better.

Because often the biggest opportunity is not to replace what your finance team does.

It’s to give them back the time to do the work that actually needs them.

About the author

Matt Swift is a Finance AI & Automation Consultant at Headstar, a Yorkshire-based finance recruitment and consultancy business. He works with SME finance teams to identify opportunities to reduce manual work through AI and automation. Having helped introduce automation across Headstar’s own operations, he focuses on practical ways businesses can improve processes, save time and make better use of their existing systems.

Website: https://www.headstar.co.uk