How to start an AI project

aat comment

Adopting artificial intelligence (AI) systems within your business can be a great way to boost productivity and improve the accuracy of your accounting. Here’s everything you need to know about getting started…

The accounting industry has been around for thousands of years. During that time, it’s embraced new technologies such as papyrus scrolls, Casio calculators and Excel spreadsheets. The latest disruption-causing tech facing the industry is, of course, the introduction of artificial intelligence (AI).  

The benefits of AI are evident

As we’ve explored in our previous two articles on AI seeWhat AI can already do for your company and Adopting AI in your business, accountants might be approaching AI with some trepidation, but the benefits to their businesses are all-too-evident.

Instead of your staff spending hours inputting tiresome tax data into spreadsheets, AI can do the task within seconds. For clients, AI analytics can use forecasts to predict where savings could be made within their firms. Customer enquiries can also be dealt with 24/7 thanks to the use of AI chatbots.

Unfortunately, having an AI-powered office isn’t a matter of getting some IT wizards to sprinkle algorithmical magic over your computer systems. You need to identify what problems AI will solve, as well as rigorously testing the software with a pilot project too.

Here’s what you need to know about getting the show on the road…

Consider how AI will be used in your workplace

If you’re dipping your toes into AI waters for the first time, it’s generally not a good idea to sink all your resources into the first snazzy software you latch eyes upon. Leave that for the big four, who can afford it.

Instead, figure out why you actually need AI in the first place:

  • is the company displaying any signs of inefficiency or inaccuracy? If so, what departments? Would staff workload be eased by speeding up administrative tasks such as processing invoices or preparing tax returns? 
  • would it help client relations if you could use data analytics to predict financial trends and forecast revenues/savings?
  • does your customer service need to improve? If so, would a natural-language bot that allow them to pay for services and/or offer customer support be beneficial? 

Things to consider when getting set up

  • how easy/difficult would it be to implement your proposed AI solution?
  • how well-equipped is your organisation to build, manage and support AI systems?
  • what would the cost be?
  • can you add AI systems to existing software, such as Sage or Xero?
  • how would AI affect the division of labour between humans and computers? In some cognitive AI projects, 80 per cent of decisions are made by machines. What impact would this have upon your workflow?
  • would adopting AI require a restructure of your organisation with staff working in different roles?
  • what staff will you allocate to the project? Or will you bring in new employees? Who will you get to lead the project?
  • does your company have cloud and data infrastructure large enough to deal with the changes?
  • do you know of any firms who have adopted similar AI systems? What has the impact upon their business been like? Has it delivered a return on investment?

Confused? Most accountancy firms start with some form of robotic process automation (RPA) see here:  What AI can already do for your company. This can relieve staff of repetitive tasks that are often prone to human error.

Having considered these factors and selected the AI systems you want to use, it’s time to move onto the next stage: launching a pilot project to see how well it works.

Launching a pilot project

A pilot project – where your company road-tests the AI software before potentially rolling it out to the rest of the firm – provides a great opportunity for you and your staff to gain first-hand knowledge of how the systems will work.

Providing you’ve got the resources, it makes good sense to trial-run several smaller pilots at once, which will increase your odds of success. If you hurl all your eggs into one big AI-shaped basket by choosing a single standalone project, you’ll be back at square one if it fails. 

Ideally, the first project(s) that you launch should be something that can be completed quickly (six-12 months) and have a high chance of racking up some small ‘wins’. It’s also worth selecting software that generates a ‘buzz’ among your staff and/or customers. Try fostering a culture whereby staff can regularly engage with the new technology and provide feedback on it.

Establishing timelines and budget

Next up? Well, because you’re evidently a sensible business-owner, you’d need to decide upon a timeline and earmark a budget for the project(s). Even though some RPA can accelerate tasks such as invoicing, it could also slow down staff productivity as employees become adjusted to the new tech. Therefore, scheduling your pilot project(s) for December and January, when accountants are at their busiest, would be foolhardy. But you knew that, right?

Recruitment is key too. Your organisation may want to hire external AI specialist engineers and/or consultants. Try to allocate eager members of your staff to the project; you’d probably want people who could act as AI champions within your workplace, spreading the word to staff about how effective it is.

Getting cynical employees on board is one of the biggest challenges businesses face when adopting AI. Try hosting presentations on its benefits and broadcast any breakthroughs or landmark successes you experience during the project(s).

Running a pilot project

After a while, you should soon able to gauge whether the pilot(s) have created any value.

  • Has your company increased revenue as a result of the AI? If you’re using data analytics, has it resulted in any financial gains for your clients?
  • Has it resulted in any new business opportunities?
  • How has the AI changed workflows? With more tasks carried out by machines, are your staff now feeling redundant with nothing to do? Or are they pursuing new assignments that drive business?
  • Have you detected any changes among your clients or customers? Do they seem satisfied now that work is (hopefully) being completed quicker?


Implementing an AI project is a daunting experience for any company. But launching a portfolio of projects (rather than just one) is a fantastic way to gain some insight into the revolutionary new tech.

The key to making AI work is taking some time out to identify <why> you need it and whether your company has the infrastructure to support it. Achieve that and you should hopefully see some results within months of starting the projects.

Further reading:

Christian Koch is an award-winning journalist/editor who has written for the Evening Standard, Sunday Times, Guardian, Telegraph, The Independent, Q, The Face and Metro. He's also written about business for Accounting Technician, 20 and Director, where he is contributing editor.

Related articles