CIMA hosted a discussion in which finance and academic leaders said AI in finance should be used to raise productivity through workflow redesign and skills development. The debate focused on a gap between strong belief in AI’s importance and weak confidence in organisations’ readiness.
Research cited in the discussion found that 88% of 1,500 senior finance leaders see AI as a game-changer, while only 8% feel very well prepared to adopt high-impact technologies. That gap reinforced a broader argument: finance teams stand to lose more from poor implementation than from any immediate effect on employment.
Bart van Ark of the Productivity Institute said the main risk was not widespread job cuts but failing to embed AI in everyday working practices. Organisations often buy tools before changing processes, he argued, leaving staff with new software but old ways of working.
That view was echoed throughout the session, which presented finance functions as a key test of AI’s ability to deliver measurable gains. Speakers said the technology can remove repetitive work, improve forecasting and support decision-making, but only if organisations redesign workflows and define where human oversight is still required.
Alexander Ilkhan, a treasury practitioner and consultant, warned against presenting automation primarily as a means of reducing headcount. He said that can discourage staff from engaging with change programmes and undermine adoption.
Instead, Ilkhan said teams respond better when AI is framed as a way to remove routine tasks and free up time for judgment-based work. In finance, that could mean less manual processing and more focus on analysis, planning and business partnering.
Tara Alas of McKinsey UK and the Productivity Institute said productivity gains often come when specific tasks are fully automated while people remain responsible for design and supervision. She also stressed the need for deliberate process redesign and clear operating guardrails, rather than isolated experiments.
Readiness gap
The discussion suggested that organisational readiness depends less on access to the latest tools than on leadership, training and governance. Skills shortages and weak motivation were described as bigger barriers than the technology itself, along with incompatible systems and poor coordination during implementation.
Panellists said many employees are already using AI tools at work, sometimes without formal approval. That, they argued, makes it more important for leaders to set rules on data use, acceptable applications and review processes, while still allowing room for testing and learning.
The group distinguished between bottom-up experimentation and top-down change. Letting teams test practical uses can help identify where AI adds value, but larger gains depend on leadership deciding which processes to redesign and where automation is appropriate.
Finance leaders were presented as central to that shift because their teams sit at the intersection of operational data, internal controls, and management decisions. The panel argued that chief financial officers and senior finance executives can either remove barriers to adoption or entrench them.
Data discipline
Fred Fowler of Coty said data quality is the starting point for any serious AI project in finance. Inconsistent information can block progress long before organisations reach more advanced forms of automation, he said.
Panellists described data management as an ongoing discipline rather than a one-off clean-up exercise. Maintaining master data, standards, and common definitions was presented as essential if finance teams want AI systems to reliably support planning, reporting, and analysis.
The debate also highlighted limits in the technology itself. Large language models were described as useful for language-heavy work, such as drafting process documents or standard operating procedures, but not as tools that should be trusted for numeric accuracy without controls.
Finance teams, therefore, need to choose tools based on the task and maintain validation processes. Speakers said that enthusiasm for AI should not lead organisations to apply a single model to every problem.
Skills and trust
A recurring theme was that successful adoption requires continuous learning rather than one-off pilots. Participants said organisations need leadership support, internal champions, access to tools, and baseline training so that staff can understand simple agents and properly check outputs.
They also argued that badly designed systems can damage trust and slow change. Involving end users early was presented as an important step in avoiding technology that adds friction and leaves staff disengaged.
Progress, the panel suggested, should be measured across several dimensions rather than through labour savings alone. These include quantity, such as time saved or efficiency gains; quality, such as fewer errors or smoother processes; and broader strategic benefits that may be harder to quantify at first.
The discussion extended beyond the private sector. Public and not-for-profit organisations were described as strong candidates for AI-driven productivity gains because of the volume of routine cognitive tasks they handle, although the same issues around skills, culture and implementation were said to apply.
In summary, the panel argued that finance departments will get the best results from AI when they start with the outcomes they want to improve, set controls around data and usage, and train staff to work with the technology rather than fear it. As one panellist put it, “Get your data right” is step one.