Who Leads Finance AI Transformation?

Microsoft's Treasury organization works with more than 100 banking partners and collects about $300 billion from customers annually. In a September 24 account of its AI work, the company describes consolidating fragmented systems before adding intelligence to collections workflows. Kathy Brustad, a director in Global Treasury and Financial Services, identifies adoption as harder than building the technology. That's a useful observation from a company with considerable access to both. Someone still has to decide which work should change, negotiate the handoffs, and get people comfortable relying on the result. Those decisions need an accountable leader inside Finance.

The workload is expanding. In McKinsey's 2025 survey of 102 CFOs, 44% of respondents reported using generative AI for more than five use cases, compared with 7% in the previous year's survey. Those figures describe that survey population, and use-case counts say little about financial returns. They do suggest a coordination problem. Once several teams are building at the same time, they compete for data access, engineering capacity, and the attention of the same finance reviewers. A CFO needs a way to choose among those investments and resolve conflicts before they become expensive commitments.

A head of finance innovation, or a VP of finance AI transformation, should have a clear mandate to make that happen. The CFO sponsors the agenda and settles decisions that exceed the leader's authority. The transformation leader owns the portfolio of finance opportunities, agrees on outcomes with process owners, and carries finance requirements through delivery. Technology retains responsibility for engineering and production infrastructure. Controllers and other finance owners remain accountable for their processes and approvals. Internal Audit provides independent challenge. Writing down those boundaries early gives each team a way to identify decisions that are drifting between departments.

Consider a proposed agentic reconciliation workflow. The agent retrieves approved source records, identifies unmatched items, assembles supporting evidence, and routes exceptions for review. Before engineering begins, Finance has to define acceptable matching rules and who can approve an adjustment. Someone must decide what happens when a source feed is incomplete, how unresolved items escalate, and which evidence is retained. Now imagine rolling that workflow across business units with different calendars and systems. The leadership work includes negotiating a common standard, understanding legitimate local differences, and deciding whether the expected benefit justifies the integration effort. A functioning demonstration answers only part of that investment question.

The person leading this work needs enough technical depth to question an architecture and understand the consequences of a design choice. They should be able to discuss data lineage, access boundaries, model evaluation, and recovery with engineering leaders, then explain the financial implications to the CFO. They also need operating experience: knowing which approval protects a meaningful risk, why a team maintains a spreadsheet outside the ERP, and when a proposed change would disrupt the close. That combination helps them set credible requirements and recognize when a conventional integration would solve the problem more reliably. Technical fluency earns its place through better decisions.

A finance transformation program also needs a disciplined way to measure progress. Establish a baseline before implementation and assign someone to validate the result after deployment. For the reconciliation example, review effort and the age of unresolved exceptions might matter alongside completion time. Track rework and control failures as well. An estimate of hours saved becomes useful when the team can show where the capacity went: fewer overtime hours, broader review coverage, or time available for analysis. A finance innovation council can make those trade-offs visible, provided it has decision rights and a manageable agenda. Each review should conclude with a funding, sequencing, or operating decision.

Adoption deserves the same attention. Give the people who perform and review the work a role in shaping it, including the ability to challenge outputs and identify missing evidence. Training should use their actual workflows, with practice handling an exception and recovering when something fails. Managers need to make room for that learning during delivery. A team asked to maintain the old process indefinitely while testing the new one is carrying two workloads. The transformation leader has to negotiate the conditions for transition, including the criteria for retiring redundant steps. Otherwise, an apparently successful launch can leave the organization with more work than it started with.

For the next planning cycle, CFOs should put a named owner and an explicit mandate alongside the AI budget. That might mean expanding an existing finance transformation leader's remit or creating a dedicated finance innovation role as the portfolio grows. The appointment should come with access to executive decisions and a working agreement with Technology. Give that leader an initial portfolio small enough to govern closely, then require evidence before expanding it. Over the following quarters, the job will be to decide which capabilities earn broader use, which need redesign, and where the organization is ready to delegate more work. Someone has to remain responsible after the launch.