The AI-first finance function: from automation to advantage
AI in finance has moved from demo to product, yet most teams cannot show the impact. BCG argues the binding constraint is now talent, with data and process close behind.
TL;DR
- BCG reports that AI spending is set to roughly double as a share of revenue, with 94 percent of organisations planning to keep investing.
- Nearly 90 percent of CEOs expect AI agents to deliver measurable ROI, yet few organisations can point to consistent results today.
- The binding constraint has shifted from model capability to data, process, and talent, with finance leaders naming talent as the most pressing challenge.
In The CFO's AI Agenda, BCG captures a moment familiar across functions. Every CFO has seen the demos: an agent that drafts variance commentary in minutes, builds scenario analyses from a prompt, or reconciles thousands of transactions overnight. The capabilities are moving from prototype to product. The impact, for most teams, is not yet showing up in the results.
The expectation gap
BCG's AI Radar 2026 finds that spending on AI is set to double as a share of revenue, and that 94 percent of organisations plan to keep investing despite uncertain short-term returns. The expectation gap is wide: nearly 90 percent of CEOs expect AI agents to deliver measurable ROI, while relatively few organisations can show consistent results so far. Most strikingly, an early-2026 Gartner survey of CFOs named building AI talent within finance, ahead of technology or budget, as the most pressing near-term challenge.
| The 2026 finance picture | Signal |
|---|---|
| Investment | AI spend set to roughly double as a share of revenue |
| Commitment | 94 percent plan to keep investing |
| Expectation | Nearly 90 percent of CEOs expect measurable agent ROI |
| Reality | Few can show consistent results today |
| Constraint | Talent is the top near-term CFO challenge |
The reason talent is the constraint is structural. As agents take on transactional and analytical work, the human role shifts from executing tasks to navigating outcomes. The controller who once compiled variance reports now validates AI-generated analysis, challenges its assumptions, and decides which findings warrant action. That is a different, more demanding job.
Where the value actually sits
The headline tension, soaring investment against thin proven returns, is the same gen AI paradox visible across the enterprise, and the resolution is the same. Value comes from redesigning a workflow, fixing the data that feeds it, and reskilling the people who supervise it.
For finance specifically, two points stand out. First, the highest-confidence value still sits in well-understood, rules-based work: reconciliations, anomaly detection, and first-draft reporting, where outcomes are quantifiable and oversight is straightforward. Second, governance is not optional here. In a function answerable to auditors, an AI result that cannot be traced to source data becomes a liability. Human-in-the-loop is a requirement here, not a preference.
A CFO's starting point
- Start with quantifiable, rules-based workflows, and measure the result against a baseline rather than a feeling.
- Invest in talent deliberately. Build AI fluency across the team, because the scarce capability is judgement over AI output.
- Make traceability a design requirement, so every agent action can be explained to an auditor.
The finance function that wins will be the one that treats AI as an operating change, grounded in clean data, clear process, and people equipped to navigate outcomes.
Related reading
- 5 AI trends in finance for 2026 every CFO must know (LucaNet)
- Closing the gen AI paradox: a leadership agenda
Source: The CFO's AI Agenda: From Automation to Advantage, Boston Consulting Group.