Is AI creating value in financial management and ERP applications or is it just hype?

The article reflects my own thoughts, maintaining a practical approach and looking beyond the marketing and hype. In the final chapter, I may take some liberties to share more future-oriented ideas, but only briefly.

Let's begin by considering what AI actually is.

Is it AI or just standard automation?

Challenge of defining artificial intelligence

The term AI seems to be used quite freely nowadays, even for simple automation logic in software solutions. I can imagine that true AI professionals (which I’m not) might be a bit irritated by this trend. However, the hype can also bring positive developments to the AI and automation industry as a whole. Automation, after all, has been a cornerstone of the financial management software industry for a long time.

Automation in the Nordics

The financial management software market in the Nordics is already quite mature, both in terms of solution usage among Nordic businesses and the level of automation achieved. For many Nordic companies, "automation" has been business as usual for over two decades. Even SMBs are using advanced solutions to manage their finances: e-invoice rates are high, most data is processed in a structured digital format, and large parts of the accounting process, from invoice data capture to posting and final bookings, are automated, requiring little manual work.

Interestingly, in some countries outside the Nordics, this level of automation is now referred to as “AI accounting.” However, it is simply human-configured rule-based automation.

Distinguishing AI from automation

Machines are far from being sentient, so where is the line between standard rule-based automation and AI? While I don't claim to have a definitive answer, I can share how I draw the distinction. In our context, I consider something to be "AI" or "machine learning-based" (as opposed to rule-based automation) when the solution can apply advanced automation logic to a specific scenario without prior human configuration.

Although the algorithm that ultimately automates a scenario is created by a human, an AI or machine learning system can apply a variety of different automation logics based on different inputs and scenarios. In my view, if a computer algorithm is advanced enough to handle multiple automation logics without human pre-configuration, it qualifies as "AI" or "machine learning" rather than standard rule-based automation.

Example: AI vs. automation

Consider this example: "If a purchase invoice includes data value 'X' or comes from 'supplier Y,' then apply posting values A and B; otherwise, leave the posting fields empty." If this rule is pre-configured, it is standard automation in my view.

However, if no one configured this rule, yet it can still be applied to this and a variety of similar but different scenarios, it can be considered "machine learning."

AI in financial management solutions today

Separating reality from hype

So, is there really AI and machine learning in financial management or ERP applications today? The answer is yes. However, be cautious. Just because something is labeled as AI doesn't mean it truly is. Hopefully, this article will help you perform some basic due diligence.

One of the widely used AI applications is similar to the example I used in the previous chapter. It's often called AI postings or something similar, depending on the vendor. It works by analyzing how human users post and book invoices, picking up even complex logics from these actions, and then applying them to future invoices. In short, the machine learns from the human user and starts to assist, without the user needing to set up any rules or configurations.

Some of these applications are very advanced, using state-of-the-art AI technologies such as neural network-based learning, even from very complex user logics. Others, however, are not as advanced. Similar logics are already used in capturing invoice data from various formats or invoice images. AI is also widely used in fraud detection, financial forecasting and reporting, risk management, and many other areas of financial management and enterprise resource planning.

The benefits of AI vs. rule-based automation

In many financial management and ERP applications, AI already performs a significant amount of work. While it is not fully autonomous, it can "configure the automation logics" on behalf of the human user in many scenarios and help analyze vast amounts of data. This results in substantial time savings, higher levels of automation, and the ability to get more value from the data.

Practical examples from real life

Example A: The high cost of manual invoice processing

A medium-sized company may receive thousands of invoices per month, many of which arrive as PDFs via email or even on paper. Typically, the company's accounting staff manually enter the invoice details into the financial management system and add necessary posting information (such as tax codes, cost centers, and other accounting dimensions) before the invoices can be processed. Even for a senior professional, this task can take several minutes per invoice. When you consider the high volume of invoices, along with potential errors and delays in visibility, the total cost of this manual process becomes substantial.

Example B: Enhancing efficiency with rule-based automation

A more advanced company employs rule-based automation for their invoice processing. They may have created supplier-specific templates that use OCR to automatically capture key invoice data from PDFs. Additionally, they might have set up rules to post invoices from specific suppliers in a particular manner, as long as the scenarios are straightforward enough for standard rule-based automation.

Compared to the traditional method, this company’s accounts payable process is significantly more efficient, with approximately 50% less time spent, 50% fewer errors, and 50% better on-time payments and control. However, whenever the company adds a new supplier, experiences changes, or encounters more complex scenarios, manual intervention is still required.

Example C: Maximizing process efficiency with AI-powered automation

Finally, consider a company using AI in addition to rule-based automation. In this setup, there is no need to configure supplier-specific invoice capture templates, as the AI adapts automatically to capture most invoice data. A larger portion of invoices can be posted and booked automatically, eliminating the need for users to pre-configure rules.

AI-enabled automation starts working immediately and becomes more efficient as more invoices are processed. Companies using AI can achieve up to twice the process efficiency compared to those relying solely on rule-based automation.

What’s next for AI in financial management

Current adoption and future trends

Although some AI applications are already in use, they are primarily employed by a small number of pioneering businesses globally. Most companies still rely on paper invoices and manual processes, especially outside regions like the Nordics and Latin America, where e-invoicing is mandatory. Over time, more businesses are expected to gradually adopt basic AI applications to enhance financial management. Meanwhile, pioneers will continue to explore new AI applications, testing various innovations over the next five years. The extent to which these applications become mainstream will depend on their success and benefits, as well as regulatory factors in financial management.

Long-term vision for financial management

In the long run, financial management and ERP applications are likely to undergo significant changes due to AI advancements. Financial management may see incremental improvements with increased automation, while accountants will focus more on handling anomalies and value-added tasks. The core accounting functions will become 90 % automated, changing the role of accountants but not eliminating the need for their expertise.

Transforming the ERP landscape

AI is poised to revolutionize not just ERP applications but also the underlying business processes. This transformation could be as profound as the introduction of limited liability companies, altering how business is conducted. AI might enable a significant portion of business activities to occur online within data networks. In these networks, AI could identify optimal trading partners, facilitate transactions and contracts efficiently, support delivery processes, and provide secure access to raw accounting data for traditional accounting and invoice automation solutions via APIs. While this vision is speculative, it underscores the importance of keeping an open mind about future AI applications.

Practical tips on how to AI future-proof your financial management

1. Set clear goals: Define your company’s targets for utilizing automation and AI, and understand why.
2. Stay informed: Ensure someone in your company follows developments in AI, even if your goals are not big.
3. Test and learn: Experiment with interesting use cases and seek vendor support for AI solutions. Learn and test with simple tools and scenarios.
4. Organize your data: Prepare your data for better future utilization with AI.

These steps won’t make you a leader in AI, but they will help you stay abreast of key developments and build readiness to react quickly. It’s also pretty low risk, as you’re not required to invest a ton of money and time into uncertain projects. I think most of us will be happily in this same group of “be prepared to see what comes next”.