If we look at how incredibly fast developments in the field of AI are currently changing, sometimes it feels like it is almost impossible to keep up. However, it is good to understand how much (or perhaps how little) influence AI can have on accountancy. That is why in this blog I would like to give you an introduction to how AI works as well as how it may influence accountancy.
With AI continuing to be on the rise, thanks to the likes of image generation tools and ChatGPT, we have been constantly hearing the term AI almost everywhere over the last year. As we become more familiar with AI, we are understanding the power of it but one thing that is not quite as clear is in the professional application.
What exactly is the difference between AI and machine learning and why is it important to understand? A beautiful quote I recently came across is:
“If it’s written in Python it’s Machine Learning. If it is written in PowerPoint, it’s probably AI.”
Essentially, something that is branded on the cover as being AI, when you look under the hood it is machine learning. Overall however, these are the same thing with the names used interchangeably generally based on who the audience is.
In relation to accounting, it is outcomes from machine learning, along with data analysis, that we as accountants can use. From structuring and standardising data and gaining insights through to automating processes and informing clients, these are all actions by accountants that can be aided by machine learning.
What are language models, or Large Language Models (LLM)
In principle, a language model with AI works very simply. You provide text as an input which is then processed by a magical Language Model, you then receive output in text from the language model. These language models are very useful in practice, for example for automatically generating a text or a translation. But language models can also be used to indicate the probability that something is or is not correct, to categorise information, or to give approval about the value of, for example, a review.
Once you make this model larger (with more parameters), they become so-called Large Language Models because of two discoveries that happen at a large scale. I call them discoveries, as opposed to inventions, because AI experts spend lots of effort and creativity in solving scaling issues, but what capabilities emerge are often truly unexpected. The first discovery is that simply by training these models to predict the next word, they start to learn complex capabilities such as reasoning, abstract understanding etc. So these models are able to help in tasks they were not trained for. Secondly, already with a few input examples (or sometimes even a single one!) these models are able to learn generalisations quickly. So size does matter!
Lessons from using AI in accountancy
While AI is taking its first steps in many places in accountancy, we at Silverfin are lucky that we’ve been integrating AI in our platform for many years, including language models and deep learning. Here are the key lessons we’ve learned so far from AI use in accounting.
Start with efficiency
Starting with the area that will provide the best efficiency means that you get an easier win at the start of the process of introducing AI. This also helps you to see the power and benefits of what you are implementing. We first started with account mapping when we began using AI for accounting purposes. That doesn’t always seem exciting, but it is actually something that yields a lot of efficiency. If we were to do the account mapping manually, it would take much more time and manpower than if AI did it for us. Additionally, it saves much time by automating where possible because the meaning of each account is known.
A change in the UX
But it’s not just about gaining efficiency. A lot of the software that accountants have been using looks and feels the same. It often tends to have the same mechanisms working in the background that software has had for years. These software tend to work on an excel-like principle of data ingestion, calculation and data presentation. However, as AI is becoming more prevalent, the user is presented with uncertainty in the outcomes. Where previously they were able to see an excel formula, for example, with AI they are not able to see the background workings as easily. This means that users need to be more comfortable in accepting the uncertainty of the result. Furthermore, when AI is being introduced, we must be careful as to how we display and interact with these probabilistic outcomes.
Overall, a good rule to follow is ‘focus on value, not tech’ when it comes to the efficient use of AI in accountancy. By this we meant to focus on what the benefits are that the AI application can bring to your firm and employees, rather than focusing on the technology itself.
Build trust first
Like anything, when you are learning something new there is always some hesitancy. When people don’t know exactly what they are doing when they are learning a new system they often have an extra critical attitude. Even if the system gives the desired result 20 times but it does not on the 21st time, the user’s confidence can immediately drop again. It is therefore important to guide people in the introduction of AI applications. Here the right UX helps as well, as mentioned above. At Silverfin, however, we see that our users often master everything in about five sessions and that accountants are generally not that suspicious of AI because they then experience how it works.
Assist instead of automate
Like any impactful probabilistic system, it is important that the user remains in control. Generally the aim is that AI can help the vast majority of the time. We never aim for AI to be able to assist in 100% of what is done because the user needs to remain in control, rather than handing that control over to AI. Therefore, when AI is introduced, the focus should be mainly on assisting, not on automating. This also allows the accountant to clearly explain the process and the result of the AI system to their customer, step by step. It also allows the accountant to get used to, and trust, the information that is being generated from the AI. For new graduate accountants in particular, this allows them to still do all the on-the-job learning that they would normally receive whilst also learning how AI can assist them.
What impact will AI have on accountancy?
Without us often realising it, AI is already increasingly present in the professional office environment. This is because new tools are added to existing and trusted applications such as Microsoft Word, Teams, or Zoom. A very basic example of this is if you are typing in a Google Document you may have noticed that common words and phrases may come up as suggested text that you might want to use. As we get used to this type of AI, it is up to the providers of these types of software and technology to make sure that they match what their competitors are doing. If they do not do this, users will quickly switch to a competitor that does offer the AI tooling that people expect.
No accountancy firm is big enough for its own AI applications
Adding tooling to existing and trusted software applications also now applies to accounting software. No accounting firm is large enough to build a complete AI application itself, therefore it is mainly up to the software suppliers of the accountancy sector to introduce AI tooling. An example of this is how the likes of Xero with JAX and Sage with Sage Ai are releasing AI tools which are becoming the norm in accounting software.
The question now is what AI is going to suit each firm, and to what degree will it be adopted. As AI starts to become more frequently used, it is good for a firm to have an idea about how they can make the most of it in their firm so that accountants are able to focus on the real value-added tasks that are not able to be outsourced to AI. These value-add areas tend to be the ones that involve the most collaboration and work directly with clients.
Destructive or transformative?
The fear of the destructive effect of AI on certain professions may be great, but is not always justified. Seeing what amazing images AI can generate, we might also have expected photographers and designers to disappear in early 2023. However, we see that the opposite is true. Companies like Adobe are building AI into their products to assist photographers, rather than replace them.
Whilst no one can see into the future, one thing for sure is that human contact is crucial for trust, and people need to have trust in their accountants. However, we cannot ignore the fact that the accounting profession will certainly change and the challenges and possibilities will also change with it. While the field has mainly had changes in the field of digitalisation in recent years, in the future it will be much more about the human side. In other words, how do we as accountants enter into more and (even) better conversations with clients based on the things that the AI assistant will offer us?
“The big challenge may be that some partners are not yet really open to the use of AI. This may put you at a disadvantage compared to competitors who already embrace AI. But ultimately, I suspect that the work of the accountant itself can only become more interesting and that there are certainly no reasons yet to embrace this special and promising development.”