AccountsIQ warns UK finance leaders on AI governance

AccountsIQ has warned that UK finance leaders are making a series of mistakes in how they adopt artificial intelligence, with weak governance among the main risks, according to Chief Technology Officer Gavin McGahey.

McGahey said finance teams are moving quickly to use AI in daily work but are not always putting the right controls in place first. He identified five main errors: skipping governance, trusting outputs without checking them against source data, removing human oversight, deploying tools without proper training, and favouring speed over transparency and control.

The warning comes as AI use in finance becomes widespread. AccountsIQ cited industry figures showing 97% of the sector has embedded AI solutions into teams, while other research points to low confidence and limited training among finance staff.

McGahey argued that many existing frameworks are too narrow because they focus on data protection, security and regulatory compliance rather than the quality and accountability of AI-supported decisions.

"Finance teams are now expected to trust AI-generated insights despite not knowing how they were created. Teams are being pushed to act on AI recommendations and defend decisions to stakeholders, which is a dangerous game. This is just one of many examples of the mistakes that finance leaders are making with AI, as without knowing how AI-generated insights were sourced, finance teams can take responsibility as they did not produce the initial recommendation," said Gavin McGahey, Chief Technology Officer at AccountsIQ.

Governance gaps

One of his central criticisms is that organisations are adopting AI before setting standards for explainability, auditability and fairness. In his view, compliance alone does not give finance teams enough assurance when using AI in reporting, analysis and decision support.

"Frameworks are failing to address explainability (if AI-generated insights can be clearly understood), Auditability, (if AI-driven processes can be reliably verified) and perceived fairness (if stakeholders can trust AI-driven outcomes)," McGahey said.

"With so many teams utilising AI to streamline their operations, simply following regulations is not enough to drive consistent confidence in financial decision-making. Financial leaders cannot rely on standalone tools and should be prioritising in-workflow AI solutions that support finance processes while maintaining full control and audibility," he added.

Checking outputs

He also pointed to the risk of accepting AI-generated material at face value. The warning follows several high-profile cases in which published reports were withdrawn or criticised after investigations found fabricated citations, hallucinated case studies or unsupported claims.

For finance teams, where outputs can feed dashboards, planning models and board-level decisions, unchecked AI responses can spread errors quickly across an organisation, he said.

"Too many leaders are allowing AI-generated insights and numbers to proceed without proper verification against source data," McGahey said. "AI tools are known for adding in multiple layers of interpretations on top of processing data, with these inferred outputs offering limited visibility into how they were produced."

"To combat this, finance teams can use screening tools to ensure compliant finance systems designed for AI regulation are implemented, allowing leaders to gain visibility and trust in newly adopted AI solutions," he added.

Human oversight

Another concern is the removal of human judgement from AI-assisted processes. McGahey said finance departments should use AI to reduce repetitive manual work, but should not hand over accountability for decisions that affect clients, stakeholders or regulatory reporting.

"Finance leaders cannot risk relying on AI without control and transparency. Their teams need regular human oversight of their AI use so they can take accountability for decision-making supported by AI data. AI should only be used to complete manual work, freeing up teams for strategic tasks," McGahey said.

"A mistake from an AI model affects every dashboard, portfolio and decision output, rather than being a single mistake that can be contained. Reputational damage will outweigh savings on cost and clients, and regulators will not be very forgiving when learning of an AI-driven incident," he said.

Skills shortage

Training is another weak point, according to AccountsIQ. McGahey said the UK finance sector faces an AI skills gap, with relatively few finance leaders saying their teams are confident in using the technology and many reporting that they do not receive enough training.

He argued that AI adoption in finance cannot be treated as a technical deployment alone. Teams that use AI for analysis, forecasting or operational work need guidance on when to use it, how to test it and how to question its results.

"The UK is facing an ever-growing AI skills gap, with 15% of CFO's claiming that their teams were confident in their AI use. Investment in AI training is significantly lacking across finance teams, with 34% of finance leaders claiming that they are not getting enough. If finance leaders are poised to integrate AI solutions into their teams, then they need to ensure that they have an appropriate training plan in place. There is no point in leading investing in AI if they're not committed to investing in their teams to support its use. AI should not be treated as an IT deployment.

"Too many leaders are underinvesting in their people while overinvesting in AI solutions. Nevertheless, teams should be encouraged to use AI, but only with guided collective intent and encouragement to use it both responsibly and cautiously.

"Finance teams are continuing to rely on AI-driven insights and tools to support daily activities. The real challenge facing finance leaders is not ensuring that data is protected, but that decisions made on the data are transparent, explainable and trustworthy," McGahey said.