What impact will artificial intelligence really have on B2B payments?

Visit any social media news feed and countless posts will tell you that AI means “nothing will ever be the same again” or even that “you are doing AI wrong.” The sheer volume of exaggerated opinions being spread makes it almost impossible for companies to distinguish between hype and reality.

This issue is intended to be addressed by the European Union’s “AI Law” (the “Law”), which came into force on August 1, 2024. The law is the world’s first regulation on artificial intelligence and specifies how the deployment and use of AI systems should be regulated. The law recognizes the transformative potential that AI can have for financial services, but also recognizes its limitations and risks.

As part of the ongoing debate about AI in financial services, B2B payments processes have been identified as an area where AI has enormous potential to accelerate digital innovation. Today I will do my best to go beyond the hype and provide a real perspective on what AI is Really Funds specifically for B2B payments.

Understand what artificial intelligence is and is not

In short, AI is a system or systems that can perform tasks that normally require human intelligence. It involves machine learning (ML), which has been used by developers for years to give computers the ability to learn without being explicitly programmed. In other words, the system can view and analyze data to refine functions and results.

A newer part of this is “deep learning,” which uses multi-layer neural networks to simulate the complex decision-making power of our brains. The benefits of deep learning described later in this article rely on Large Language Models (LLM) that are pre-trained on representative data (e.g. payment/transaction/tender data). Deep learning AI not only looks at and learns behavioral patterns from the data, but is also able to make informed decisions based on that data.

Before I explore what this could mean for B2B payments, I want to clarify one caveat: human oversight is still required to ensure operations run smoothly. AI is a supporting tool, not a single answer to every question. Because the technology is still mature, you can’t yet completely hand over the keys to your B2B payment process. Manual processes will still have their place in B2B payments today, but AI tools will help you learn, adapt and improve faster and at scale.

The AI ​​law – what you need to know

The law attempts to categorize different AI systems based on potential impacts and risks. The two main risk categories include:

1. Unacceptable risk – AI systems are considered a threat to people and are banned. These include systems for cognitive behavioral manipulation, social assessment and real-time biometric identification.
2. High risk – AI systems that have a negative impact on security or fundamental rights. High-risk AI systems are subject to rigorous scrutiny and must meet strict regulatory standards before being brought to market. These high-risk systems are divided into two further categories:

• • AI systems used in products that fall under it the EU product safety legislationincluding toys, aviation, automobiles, medical devices and elevators.
  • AI systems fall into specific areas that must be registered in an EU database.

The currently most widely used form of AI, “generative AI” (think ChatGPT, Copilot and Gemini), is not considered high-risk but must comply with transparency requirements and EU copyright law.

General-purpose, high-impact AI models that could pose a systemic risk, such as GPT-4o, would need to undergo a thorough evaluation and any serious incidents would need to be reported to the European Commission.

The law is expected to come into full force by May 2026 following consultations, amendments and the creation of “supervisory authorities” in each EU member state. However, the EU will start banning AI systems with “unacceptable risk” as early as November and the “codes of conduct” will be applied by February 2025.

Given the law, how can AI be used risk-free to optimize B2B payments?

AI will transform how payment data is analyzed

Today’s B2B payment platforms are not one-size-fits-all; Instead, they provide businesses with a toolkit to customize their payment interactions.

AI-based language models and machine learning can be used by payment providers to quickly understand and interpret the rich data they have access to (e.g. invoices or receipts). This gives us insights into trends, buyer behavior, risk analysis and anomaly detection. Without AI, this is a manual, time-consuming task.

A concrete advantage of this data analysis for companies comes from combining the extensive payment data available with knowledge of the capabilities, products and/or services of a wide range of providers. AI could then, for example, detect when an existing supplier is able to deliver something that is currently sourced elsewhere. By using one supplier for both products/services, the company saves through economies of scale.

Another benefit of data analysis comes from payment technology experts. We trained a service to extract data from an order or invoice to pass Level 3 data, which is tax recognizable in some territories. This automatically provides the buyer with further details of the transaction, including relevant tax information, invoice number, cost center and a breakdown of the products or services delivered. This makes managing tax reporting and remittances, purchasing control and reconciliation simple and straightforward.

However, AI-supported data analysis not only saves time and money. It also creates new value by allowing providers to use the data to create hyper-personalized payment experiences for each buyer or supplier. For example, AI and ML tools could look for buying and selling opportunities and run a “matchmaking supplier activation service” that recommends the best payment methods – and the best rates – for different accounts or transactions. The more personalized a payment experience is, the happier the buyer is and the more likely they are to buy (again).

Efficient data flows mean stronger cash flows

Another practical application of AI is to optimize cash management for shoppers. This is done by using the data to determine who is strategically important and when they need to be paid. It could even be recommended to group certain invoices for the same supplier, combining them into one payment per supplier, reducing interchange fees and reducing card acceptance costs.

AI can also perform predictive analytics for cash flow management and quickly analyze historical payment data to predict cash flow trends, allowing businesses to anticipate and proactively address potential challenges. This is particularly valuable in the current economic climate where cash flow is critical.

By extracting value-added, tax-relevant data from an order or invoice, AI can quickly analyze invoices and receipts to enable efficient and accurate automation of the VAT recovery process. Imagine: It’s time for your finance team to reclaim VAT on current invoices and receipts, but they don’t have to manually go through each receipt or invoice and categorize them into a reclaimable or non-reclaimable pile. It sounds like a dream, but it’s becoming a reality for companies everywhere: AI does the heavy lifting and humans check it, saving a lot of time and resources.

Faster and more accurate invoice reconciliation

The third key advantage of AI is automated invoice reconciliation. By identifying key information from an invoice and recognizing regular payees, AI can streamline and automate the verification process. This has the potential to significantly speed up transactions and enable more efficient payment processing.

Bringing together all supporting formalities such as shipping, customs, routes and JIT (just-in-time) requirements can also be handled by AI and is likely to be less prone to human error.

This presents a great opportunity to make B2B payments faster, reduce costs and increase efficiency. Companies know this:
44% of them Midsize businesses expect cost savings and improved cash flow as a direct result of implementing further automation within the next three years. According to American Express 48%of medium-sized companies expect faster payment processes, more reliable payments and a wider range of payment options.

When. Not if.

There are significant opportunities to leverage AI in B2B payment processes to do the heavy lifting. However, it is important to view these opportunities with a balanced understanding of the limitations of AI.

Although all of the opportunities for AI in B2B payments described today are based on relatively low-risk AI systems, human oversight of these systems is still essential. However, this problem can be avoided as implementing AI frees up time and resources.

AI in B2B payments is not an if, but a when. The question is when will you take the plunge hand in hand with technology instead of either fearing it or letting it take full control.

In order to grow, it is important that users see the tangible benefits. For example, by increasing efficiency in accounts payable, companies can reallocate time and resources previously spent in accounts payable to other areas of the business. Early adopters are starting to test the waters, but only time will tell what impact AI will have.

Most companies will likely wait for early adopters to fail, learn, and progress. As we know, when something goes wrong with B2B payments, it can have a huge impact on individuals, businesses and the economy. Only when the risk is clearly defined and manageable will AI really become the game changer in B2B payments that all advertising claims.