AI in Source to Pay: seize the opportunities without losing the human touch

“Artificial intelligence is a powerful engine behind digitalisation. Within Source to Pay too, you can see AI gaining more and more ground”, says Paul van Rietschoten, Managing Director of Spendlinq. “Think of applications that analyse contract texts, predict supplier risks or automatically generate orders based on consumption data. But what’s often forgotten is that the strength of digital transformation lies in the combination of data, processes and people. And you can’t replace that last factor with any technology.”

What AI can and can’t do

AI has now developed to the point where it can noticeably strengthen specific process steps within Source to Pay, according to Paul. “In spend analyses, for example, AI can recognise patterns and suggest categories. In contract management, you see solutions that automatically flag deviating terms. And in supplier management, models can build risk profiles based on external data, such as news, ESG ratings or delivery reliability. That speeds up decision-making, sharpens insights and prevents errors.”

But AI can’t work miracles, he stresses. “We regularly see organisations expect AI to take over entire processes. That’s not realistic. AI is very good at recognising patterns, but not at understanding context. It can’t weigh up interests, sense nuance or read an organisation’s culture. And that’s exactly what makes Source to Pay complex.”

The human role remains crucial

Paul gives supplier selection as an example. “An AI model can make recommendations based on data: this supplier performs best, this one meets your criteria. But deciding whether that supplier fits your sourcing strategy, your sustainability policy or the sensitivities within your organisation remains human work. You need experience, instinct and alignment for that.”

Human control also remains indispensable when drafting and managing contracts. “AI can flag deviations, compare versions or suggest standard clauses. But interpreting what is strategically important, or conducting negotiations, will never be taken over by an algorithm. And rightly so.”

Data quality as a foundation

Paul also points to an important foundation: data quality. “AI is only as good as the data you put into it. And let’s be honest: in many organisations, supplier data, procurement categories and contract records aren’t optimally structured. You can deploy AI, but if the basics aren’t in order, you’ll get messy results. Then human correction is still needed.”

AI supports but doesn’t replace

In practice, Paul sees AI mainly helping to support processes, not replace them. “We help customers take a realistic look at where automation adds value,” Paul explains. “For example, in generating order suggestions based on consumption and contract agreements. Or in automatically generating reports on deviations or risks. That helps the buyer or contract manager, but it doesn’t take over the work.”

Stay in control

What matters here is that organisations stay in control. “We often hear: ‘the supplier offers AI as standard in the platform’. That’s positive in itself, but it also means that as an organisation you need to understand how the model works, what it’s based on and what it does with your decision-making information. Especially when it comes to compliance or audits,” says Paul.

New legislation underlines this. “The AI Act will set requirements for transparency, explainability and risk management. Within Source to Pay, that affects, for example, the use of AI in supplier selection, contract review or invoice approval. You’ll soon have to be able to demonstrate why a decision was made. And that’s only possible if you still understand the process.”

Authorisation, oversight and accountability

Governance, security and privacy also remain important themes. “AI models often process sensitive information: pricing agreements, supplier terms, contract texts, financial commitments. That calls for strong authorisation, oversight and accountability. Not just technically, but organisationally too,” Paul continues. “You need to know who’s responsible for monitoring outputs, training models and correcting course when results deviate.”

AI as part of a broader Source to Pay strategy

According to Paul, the biggest risk is that AI is seen as a self-evident solution. “But without the right set-up, processes and people, you won’t get results. What’s more, you’re taking risks. If a model suggests something that’s wrong – and nobody checks it – you make mistakes that can have a serious impact. And in the Source to Pay domain, that can lead to the wrong supplier choices, breaches of contract or unwanted commitments.”

That’s why he advises against looking at AI in isolation from the bigger picture. “AI has to be part of a broader Source to Pay strategy. It has to fit within your governance, your processes, your architecture. And above all: it should strengthen your people, not replace them.”

Real value lies in the balance

According to Paul, the strength lies in the balance. “Let AI comb through the data, raise signals and make suggestions. But use human expertise to make choices, weigh up interests and adapt processes. That’s where the real value lies.”

Spendlinq supports organisations in making exactly that trade-off. “Together with customers, we look at where AI does work, where it doesn’t, and how to set it up responsibly. We don’t just look at technology, but also at processes, people and data. Because that’s what it takes for a sustainable digital transformation of Source to Pay.”