Finance leaders need AI skills as governance tightens

Finance recruitment specialists and technology executives say artificial intelligence and data literacy are now central to senior finance and fintech roles.

The comments come as organisations reassess the skills and governance needed to use AI in core financial processes.

Phil Scott, Managing Director of FD Recruit, said demand for finance leaders with practical AI experience has risen sharply over the past year. The firm places finance directors and chief financial officers across the UK and interviews more than 5,000 senior finance candidates each year.

"Technical accounting and commercial judgment used to be enough to shortlist a finance director. Now employers want a finance leader who understands the systems the business runs on and can challenge the numbers they produce. We interview more than 5,000 senior finance professionals each year, and 80% of our current mandates ask for practical AI experience. When we surveyed more than 200 finance professionals, around one in ten had used AI beyond drafting commentary or summarising reports. Employers want someone who has led an AI implementation and taken responsibility for the outcome, so they compete for the same small group of candidates.

These decisions land with the finance director. ERP replacement and cloud finance platform selection sit in finance. The budget and the audit trail do too, so the finance director owns the outcome. One who cannot interrogate a data model is leaving that call to someone else, but is still the one answering for it in the boardroom.

The skills employers need are more focused than many job ads suggest, and they depend more on judgment than technical expertise. A finance director needs to understand a data model well enough to challenge the numbers it produces. They also need to know where a tool is appropriate and where it is not. Configuration and dashboard building belong in the nice-to-have column, because a finance director can hire people with that expertise. The candidates who meet that standard are not the ones who list tools on a CV. They are the ones who ran the project and put their name to the number that reached the board. That is the difference our clients pay for, and it is the gap that shows up on most shortlists," Scott said.

Scott's comments highlight a gap between employer expectations and the supply of senior finance professionals who have led AI projects end to end. Clients are already signing off finance headcount budgets for 2026, while the number of candidates with project ownership and data-model fluency remains limited.

Technology vendors are also focusing on governance and trust as AI moves into finance and audit workflows. Executives at MindBridge, Lucanet and saas.group described a shift from experimentation to managed deployment at scale.

Mike Maziarz, Chief Product & Marketing Officer at MindBridge, said the conversation is moving beyond model performance alone.

"Everyone is talking about better models. I think the bigger shift is what happens after organisations put AI into production. Getting AI deployed is becoming the easy part. Running it responsibly across critical business processes is much harder. The organisations that succeed will be the ones that build continuous oversight into the way AI operates from day one, instead of trying to add it after the fact. Governance used to be considered a brake. In the age of AI adoption, the companies that learn to govern and oversee agentic workflows will outpace those that do not," Maziarz said.

He said trust, explainability and oversight now define success in enterprise AI projects more than raw accuracy metrics.

"Everyone is focused on building smarter AI. I think the bigger opportunity is building AI that organisations can trust. The companies that pull ahead will not necessarily have the best models. They will be the ones confident enough to put AI into their most important business processes because they can explain what it is doing, monitor how it behaves, and hold it accountable," Maziarz said.

He warned that existing control frameworks are often not suited to AI-driven processes.

"One mistake many leaders will make is assuming their existing governance and control frameworks are enough for AI. They are not. AI changes how work gets done, how decisions are made, and how risk appears inside an organisation. Companies that treat AI governance as a technical implementation instead of a core business capability will find it much harder to move beyond pilots and scale AI with confidence," Maziarz said.

Maziarz added that AI is reshaping roles inside finance and audit teams.

"What we are seeing at MindBridge and across our customers is that AI is changing the role people play. Less time is spent doing the work and more time is spent reviewing it, questioning it, and making decisions about it. The value shifts from execution to judgment. The people who stand out will be the ones who can apply domain expertise, validate outcomes, and build trust in increasingly autonomous systems," he said.

"The signal I'll be watching is how many enterprise AI initiatives successfully make the jump from pilot to production. By the end of 2026, I think leaders will spend less time asking whether a model performs well and more time asking whether they trust it to run critical business processes. That will become the real measure of success," Maziarz said.

In corporate performance management and tax, Lucanet Chief Technology Officer Kevin Smith drew a similar line between automation and assurance in the office of the chief financial officer.

"The real breakthrough for AI in finance and tax is not automation itself. It is the ability to automate complex workflows while preserving the controls and accountability that finance depends on. At the end of the day, intelligence must be trustworthy, traceable, and defensible in front of an auditor," Smith said.

Smith said finance teams expect secure, isolated and certified environments around AI agents handling planning, closing and reporting workflows. Vendors in this segment now emphasise auditability and regulatory alignment alongside efficiency gains.

At saas.group, Co-Founder and Chief Executive Officer Tim Schumacher said cost, resilience and model diversity are becoming central themes for fintech and software-as-a-service providers that integrate AI.

"One of the biggest AI trends over the next few years will not be companies asking, 'Which is the smartest model?' but, 'Which is the best model for this task at this price?' The first wave of enterprise AI adoption was dominated by whichever frontier model performed best, but as businesses have started deploying AI across thousands or even millions of workflows every day, economics has become impossible to ignore. Inference costs add up quickly, and many companies are realising they do not always need the most powerful model to get the job done," Schumacher said.

He pointed to recent restrictions on some US AI models as a factor in enterprise strategy.

"The recent restrictions affecting access to some US AI models have also been a wake-up call. If access to a critical AI capability can change overnight because of geopolitical decisions or export controls, businesses need to start thinking about resilience as well as performance. It is opening the door to a much more diverse AI ecosystem. Open-source models, including increasingly capable Chinese models, are becoming attractive not because organisations necessarily want to replace US AI providers, but because they are looking for lower costs and greater flexibility," Schumacher said.

Schumacher said SaaS providers that avoid dependence on a single AI stack may benefit as enterprises seek flexibility.

"Many SaaS companies that embraced AI to defend themselves against disruption may now be among the biggest beneficiaries of this shift. Rather than building around a single AI provider, they are increasingly becoming model-agnostic, selecting whichever model offers the best balance of cost, performance and reliability for each use case," he said.

"I think this is the beginning of a long-term change, with enterprise AI likely to become a multi-model world. The winners will not necessarily be the companies with exclusive access to one frontier model, but those that can effectively integrate multiple models as the economics and technology continue to evolve," Schumacher said.