TL;DR

R3, the trade body for restructuring and insolvency practitioners, has published what it believes is the first thorough audit of how the profession actually uses AI. The picture is wide but shallow adoption, held back less by scepticism than by nobody knowing who carries the risk.

Adoption stops at the assistant layer

Document intake has been digitised at 81% of respondents’ firms, and 52% reach for general assistants like Copilot and ChatGPT. Past that point the numbers collapse: under one in ten have deployed machine learning or agents of any kind.

Enthusiasm is not the constraint. Over 90% expect time savings, nearly three-quarters anticipate lighter administrative loads, and two-thirds think the efficiency gain will be a step change rather than an increment. More than three-quarters also expect the job itself to change, putting a premium on digital literacy, ethical judgement, data analysis and client communication.

What stops them is a cluster of unresolved liabilities: confidentiality and data security, no clear regulatory guidance, and — the item worth noting — uncertainty over whether professional indemnity cover responds when AI is in the workflow. Traditional partnership economics and thin training budgets do the rest.

The indemnity question is spreading

That insurance gap is not peculiar to insolvency. GPs are discovering the same hole from a different angle, with NHS partners personally exposed where indemnity covers a rogue individual but not a structural failure. Two professions, unconnected regulators, identical unanswered question.

The pattern across UK professional services this year has been consistent: adoption races ahead, assurance lags. Lawyers have reached 94% adoption while their worry about AI errors keeps climbing. Sonia Jordan, who presides over R3 and practises as a partner at Knights, argues the firms that gain most will be those building technical literacy and governance together, rather than one before the other.

Looking forward

Report author Robert Guidi, chief executive of Alph4, makes the reasonable case that AI will not displace professional judgement so much as clear time for the investigative work that requires it. The obstacle is structural, though, not technical. Until an insurer or a regulator says plainly what happens when an AI-assisted file goes wrong, cautious firms will keep their usage parked safely at drafting and search — which is roughly where this survey found them.