AI in Supply Chain Finance – Seven Years On

Originally published in 2019. Updated and republished in May 2026.

In 2019, at the SCF Forum Europe, CODIX COO Laurent Tabouelle was the dissenting voice on a panel about AI in supply chain finance. While other panellists framed AI as the enabler that would change everything, Laurent argued the opposite. AI was not a magical solution for everything. Most of what was being called AI in SCF at the time was better characterised as augmented decision-making, not pure AI. And the path to genuinely AI-enabled SCF solutions would be slow, with real reliability only emerging once humans had spent years feeding the systems the right data.

Seven years on, the market has proved him right.

What the room got wrong in 2019

The 2019 SCF Forum panel was unable to agree on whether AI would live up to the hype. The optimists framed it as an enabler that would unlock new efficiencies, accelerate onboarding, and rewrite credit decisioning. The audience pushed back, raising real concerns about accuracy, data quality, and the time it would take to develop fully AI-enabled platforms. The vendors caught in the middle leaned into the buzz, because that is what vendors do when a new technology is rising and the buyer is uncertain.

The idea was clear at the time: a machine learning tool can help a lender decide whether a buyer is a good or bad payer, but only after a human has spent considerable time labelling what good and bad actually mean. The intelligence sits in the training data, and that data takes years to build properly. Most of what was being marketed as AI in SCF was, in his words, augmented decision-making with the right packaging.

What has moved since 2019

Three things have changed in commercial finance AI between 2019 and 2026.

First, the training data caught up. Lenders running converged platforms across factoring, ABL, receivables and SCF for ten years or more now hold enough labelled history to make machine learning genuinely useful. Payment allocation, debt collection prioritisation, anomaly flagging, document classification: these are all reliable now because the data underneath finally has the depth required. Institutions with long-standing unified data environments are generally seeing stronger AI outcomes because they possess deeper and cleaner operational histories.

Second, generative AI changed the surface area. Drafting, summarising, document handling, contract analysis: all transformed by large language models that did not exist in 2019. This is the layer where the productivity gains are real and immediate, and where most lenders are concentrating their AI spend in 2026.

Third, the discipline question got harder, not easier. New entrants are now positioning entire platforms as AI-native, with AI baked into every layer of the lending stack. The marketing is louder than ever. But the underlying question Laurent raised in 2019 is still the question that matters: how much of this is genuine intelligence, and how much is augmented decision-making in a heavier wrapper?

What hasn’t changed

The debate has not moved on as much as the technology has.

Industry forums in 2026 are still divided on the same questions: where AI belongs, where it should not, what counts as augmented versus autonomous, and how much trust to place in models trained on data the lender cannot fully inspect. The audience pushback at SCF Forum Europe in 2019 sounds remarkably similar to the credit committee pushback at any commercial finance conference in 2026.

The reason is structural. AI in commercial finance is not simply a technology problem. It is a data problem, and an orchestration problem. Lenders running fragmented systems cannot give AI the unified data it needs to be reliable, no matter how loud the marketing. Lenders running converged platforms have the data, but still need the judgement to decide where AI belongs and where the human still wins.

A broader industry lesson

The evolution of AI in supply chain finance over the last seven years suggests that the industry may be entering a more pragmatic phase of adoption. Early expectations centred on disruption and replacement. More recent experience points toward augmentation, operational efficiency, and incremental improvement.

Rather than eliminating human decision-making, AI in commercial finance is increasingly being deployed to support analysts, underwriters, operations teams, and relationship managers by reducing manual workload and improving access to information. In this sense, the most sustainable AI strategies may prove to be those focused less on technological novelty and more on data quality, process integration, and institutional discipline.

The wider lesson for the market is that AI adoption in finance is unlikely to be defined by a single breakthrough moment. Instead, competitive advantage may increasingly depend on long-term investments in data architecture, governance, and operational consistency – foundations that are significantly harder to build than AI interfaces themselves.

Looking ahead

As discussions continue across the commercial finance industry in 2026, the debate is becoming less about whether AI matters and more about how it should be applied responsibly and effectively.

The past seven years have demonstrated both the potential and the limitations of AI in supply chain finance. The next phase is likely to be shaped not by the loudest claims, but by the organisations able to combine technological capability with high-quality data, sound governance, and realistic expectations around where AI can genuinely deliver value.

Authorship:

This article revisits and builds upon the original analysis published in 2019 by Laurent Tabouelle, COO of CODIX Group. The present version includes updated industry observations and contextual analysis by Daniel Bielsa, Head of CODIX Spain and LATAM.