CFO Advisory #2: Lead finance through the uncomfortable middle

AI's impact on the finance profession remains contested because the technology shifts quickly, the vendor landscape is unsettled, and the use cases are still maturing. Uncertainty, in the absence of direction, produces predictable behaviour: people experiment independently, pilots scatter across the function, and the operating model is quietly restructured without anyone designing it.

The challenge is partly technical; the conversation is not. The questions are about where the function creates value today, what it is willing to stop doing to make room, and how it intends to behave under sustained uncertainty. This is the discipline at the heart of any sound strategy: translating a complex and ambiguous challenge into simple, deliberate choices.

Treat learning as a deliverable

There is no such thing as a finance function that has opted out of AI. If leadership does not set the direction, people will, without coordination, without governance, and often without leadership’s awareness.

The workload-creep trap. Recent research found that AI adoption, in practice, did not reduce workloads. It intensified them. People worked faster, took on broader scope, and extended their hours. Not because anyone asked, but because AI made additional work feel attainable. The trap operates at two levels. Individuals drift into self-directed activity where extra effort does not translate into outcomes that advance the organisation's priorities. And the initial productivity surge eventually gives way to cognitive overload and burnout (Ranganathan & Ye, 2026). Output to the business is unchanged. The cost to the people is not. This is precisely why leadership must set and hold the direction.

There is a broader risk worth naming. The most consequential question in any AI transformation programme is not how much productivity can be captured, but what human work should become once AI absorbs the repetitive, administrative, and coordination-heavy tasks that currently consume so much organisational capacity. Many professionals are already caught in a cycle of execution with insufficient time to ask the more fundamental questions: Are we optimising for the right outcomes? Where is the real value in what we do? How should our operating model evolve as AI capabilities mature? If the capacity that AI frees is simply reinvested in throughput, the function ends up faster but not meaningfully better. The winning finance organisations will be those that redirect freed capacity towards higher-order contribution: sense-making, judgement, design, governance, and continuous transformation.

Most pilots will not scale, and that is by design. The discipline is to ensure that every pilot produces something useful, even when the decision is not to scale it: knowledge about what fits your data, your processes, and your people. Document the learnings, share them, and allow them to compound. In a market where tools and vendors continue to evolve, the organisational ability to absorb and apply is the durable asset.

Design the approach

Start with the why. Before any pilot, the function needs a reason that pulls rather than pushes, because AI is not the goal; it is an enabler of the strategy the business is already committed to. A finance leader should be able to answer two questions with clarity: Where do we have to win? And how does AI get us there faster, better, or at lower cost? Answering both questions directs scarce capacity towards the work that matters and frames this as a leadership conversation rather than a delegated task. The stakes are real: standing still is itself a decision, and organisations that do not build the capability fall behind quietly.

Construct the right team. The instinct is to hire an AI specialist and point them towards the data. It rarely works in isolation. AI experts without business context produce answers to the wrong questions. Finance professionals without technical grounding produce wish lists. The effective team is small and genuinely cross-functional: a business owner who knows the process, an AI or data engineer who can build, a data steward who holds the definitions, and a change owner who ensures the result actually lands. Without all four roles covered, pilots reliably stall.

Establish governance that enables speed. Governance has a reputation in finance as a constraint. In the AI context, the opposite is true. Useful AI inside organisations rarely emerges as a single large solution. It emerges as many bounded agents and automation steps working in combination. That makes governing it fundamentally different from past technology programmes. Organisations need to be explicit about which agent owns which decision, where the human review gate sits, and what the audit trail looks like when no person has touched the output. Clear data ownership, explicit decision rights, and a defined threshold for human review reduce friction in every pilot and make it both safer and faster to move.

This is not a theoretical concern. Governance failures in AI programmes tend not to look like dramatic breakdowns. They look like interpretation drifts: a metric calculated differently across teams, a control that has not been valid since last quarter, an exception that was escalated in one country and auto-resolved in another. The organisation that avoids these failures is the one that treats governance as design infrastructure, not an approval layer.

Free and redirect the capacity to act

None of this works if the people responsible for running pilots are already at capacity. Leadership must create genuine space for AI work and then protect it. The amount will vary by function; the act of creating it forces a deliberate choice about what is being prioritised and what is not.

A note on change communication

Employees frequently experience change communication as insufficient. Not because leadership has communicated too little, but because the communication has not been translated into locally meaningful forms. Organisations tend to communicate from the level of strategy, ambition, and business case. Individuals need to understand what the change means for their specific role, their daily priorities, their skills, and their sense of contribution.

The goal is not more communication; it is more relevant communication: role-specific, channel-appropriate, and responsive to genuine questions, while preserving a consistent core message across the organisation.

What you do on Monday

Call the meeting. Set the agenda. Start the conversation.

The single most consequential action is often also the simplest. Schedule a meeting with your finance leadership team and, where relevant, peer leaders from across the business. Bring the following questions:

• Where does finance create the most value today, and where is it creating the least?
• Where could AI plausibly amplify that value in the short term? Not in five years, not in theory.
• Which of the three modes of engagement are we currently leaning into (personal productivity, process improvement, process transformation)? Is the portfolio deliberately balanced?
• What current activity are we willing to stop, redirect, or buy ourselves out of, to free the capacity to act?
• What two or three pilots do we commit to? Each should carry a clear KPI, a named owner, and a defined point at which we will either scale or stop.

The CFO's role has never been solely about the numbers. It has always been about positioning the organisation to make better decisions when the path is unclear. The leaders who shape what comes next will be those who started a different kind of conversation in the function: one in which uncertainty is named openly, the approach is designed deliberately, and learning is treated as a deliverable in its own right.

If this is a conversation you would like to continue this conversation, we would welcome it.