Chasing ghost assets: Can AI become insolvency practitioners' new investigative assistant?

Depending on who you ask, artificial intelligence is either the future of professional services or a highly enthusiastic intern who never sleeps but occasionally gets things spectacularly wrong. The reality probably lies somewhere in between.

As African businesses expand across borders, insolvency practitioners are increasingly confronted with investigations that extend far beyond a single jurisdiction. A distressed company may have operations in South Africa, bank accounts in Mauritius, assets in Kenya, suppliers in Botswana, and directors residing elsewhere.

The question is no longer whether AI will find its way into insolvency practice, but how to use it well. A September 2026 survey by R3, the UK's insolvency trade body, found that 52% of firms already use general AI tools such as ChatGPT or Copilot, yet only 11% believe current regulation and professional standards adequately address their use.

The increasing challenge of cross-border insolvency

Investigating an insolvency that spans more than one country is complex and resource intensive. Practitioners may have to trace assets, examine large volumes of financial records, identify related party dealings, and establish whether assets were improperly transferred before insolvency proceedings commenced. In South Africa, the problem is compounded because the Cross-Border Insolvency Act of 2000, which adopts the UNCITRAL Model Law, has never become operational because no foreign states have been designated under it. Recognition of foreign proceedings still turns on the common law, case by case.

Often, the volume of information itself becomes a significant obstacle. A modern insolvency investigation may involve thousands of emails, hundreds of contracts, years of bank statements, management reports, loan agreements and accounting records. Reviewing these manually is expensive, time consuming and vulnerable to human error. This is where AI could fundamentally change how investigations are conducted.

Asset tracing and recovery

Many insolvency investigations feel a bit like a treasure hunt, except nobody left a map and some of the treasure may have disappeared years ago. AI may not find every missing piece, but it can often help identify where practitioners should start looking.

AI tools can surface unusual payment flows and previously unnoticed links between companies, such as recurring transfers to entities that share directors, addresses, email domains or banking details with the insolvent business. Seemingly unrelated payments can reveal a trail to recoverable assets.

Identifying suspicious transactions

AI can also spot unusual activity in financial records, such as large transactions that differ from a company's normal spending patterns, unexplained transfers shortly before insolvency, or sudden changes in trading activity.

Importantly, AI does not determine whether a transaction is unlawful. It helps practitioners identify areas warranting closer scrutiny: an investigative aid, not a replacement for professional judgment.

Reviewing large volumes of documents

Let us be honest, most practitioners would probably agree that reviewing 50,000 emails is not why they entered the insolvency profession. AI does not eliminate the need for human oversight, but it does mean less time searching for the proverbial needle in the haystack and more time deciding what to do once it is found.

AI review systems can categorise documents, identify key topics and prioritise material for human review. For smaller estates, this may make deeper investigations economically viable.

The risks of over-reliance on AI

AI can be a powerful ally, but it should never be relied on without proper scrutiny.

AI systems do not investigate facts. They identify patterns, generate predictions and draw correlations, and their conclusions depend on the quality and completeness of the information provided to them. If the underlying data is inaccurate or incomplete, the resulting analysis may be equally flawed. AI may identify legitimate transactions as suspicious or overlook misconduct that does not fit recognised patterns.

Confidence and accuracy are also not always the same thing. AI can present conclusions with remarkable certainty, even when those conclusions are entirely incorrect. Anyone who has used AI long enough will know that it occasionally possesses the confidence of senior counsel and the accuracy of a candidate attorney who forgot to read the bundle.

Insolvency practitioners often hold highly sensitive information relating to debtors, creditors, employees and third parties. In cross-border investigations, information may be transferred across jurisdictions with differing privacy and data protection requirements. Practitioners must carefully consider where information is stored, who has access to it, and whether data uploaded to external systems could become accessible to unauthorised parties.

The human element remains essential

AI may be able to identify that money moved from Point A to Point B. It cannot explain why it moved, whether it should have moved, or whether the explanation offered by management is believable. At the heart of every corporate collapse is a very human story, shaped by decisions, judgment calls and sometimes costly mistakes.

An experienced practitioner applies judgment, context and healthy scepticism to the evidence. AI cannot determine the reliability of a witness, negotiate a settlement, exercise legal discretion or decide on the best strategy on a matter. Creditors can therefore rest easy knowing that robots will not be chairing creditors' meetings in the foreseeable future.

What AI is likely to do, however, is change the way investigations are conducted. As practitioners trace assets across jurisdictions, work through vast volumes of financial information and confront increasingly sophisticated attempts to shield assets from creditors, AI offers powerful new capabilities. Used properly, it can help uncover connections that may otherwise remain hidden. Used carelessly, it can lead investigators confidently down the wrong path.

AI may well be the most diligent assistant an insolvency practitioner could ask for. It never sleeps, never takes leave and can review a mountain of documents before most of us have finished our first cup of coffee. What it cannot do is apply common sense, commercial judgment or the instinctive doubt that arises when a director insists that a substantial sum of money simply "cannot be accounted for" shortly before liquidation.

Ultimately, AI is unlikely to replace insolvency practitioners, but practitioners who use AI effectively may well replace those who do not. The future may not belong to AI, but to sensible practitioners who know when to trust the machine, when to question it, and when to override it entirely. Because in insolvency, as in life, having all the data is one thing; knowing when something smells wrong is quite another.

Bridget Letsholo is a Director and Head of Business Restructuring & Insolvency at CMS South Africa. Nomcebo Ntuli is an Associate in the Dispute Resolution and Business Restructuring & Insolvency Department.