Finance leaders push disciplined AI pilots: humans in the loop, clean data, and clear use cases
Artificial intelligence is already reshaping finance workflows. The message from senior leaders at Alphabet, HP, IBM, Meta, and ServiceNow during a recent Financial Executives International conference was consistent: move with intent, build guardrails early, and keep humans in the loop.
Success isn't about speed. It's about control, clarity, and measurable impact.
Financial Executives International
Governance is the differentiator
"Embedding responsible technology principles and governance from design to deployment to monitoring will really be a differentiator in the long term for companies," said Denise Lucas, VP and assistant controller at IBM.
That means establishing model ownership, documented data flows, documented prompts and changes, and continuous monitoring before scale. Frameworks like the NIST AI Risk Management Framework can help structure risk controls.
Humans stay in the loop
ServiceNow is using a third-party NLP and generative AI tool to speed contract review within order-to-cash. "We're being thoughtful; we're being careful," said assistant controller Danielle Fontaine, noting quarter-end pressure and the need to focus on the "weird and wonderful" deals while standard contracts flow through.
The tool automates data extraction and prepares revenue contract review checklists so the team can "know quickly where to put their energy." Given revenue recognition risk, 100% of AI-prepared checklists are still reviewed by a human.
- Use AI to extract, summarize, and route work.
- Require human review where misstatements carry material risk.
Start with the problem, then pick the tech
HP begins every transformation by asking, "What problem are we trying to solve?" said Meg Dholabhai, global finance transformation leader. Technology comes second.
Peleton's Saqib Baig warned that digitizing a bad process just creates long-term tech debt. Get the sequence right before automating.
Clean data wins
Alphabet's chief accounting officer and corporate controller, Amie Thuener, stressed the obvious (and often ignored): you need very good, clean data - and you need to know where it lives - before you push a transformation forward. No clean data, no reliable AI.
Real finance use cases gaining traction
- ServiceNow: AI-assisted revenue contract review with human sign-off.
- IBM: AI applied to expense forecasting.
- Meta: AI supporting prepaid accounting workflows.
- HP: Agentic AI to resolve customer payment disputes faster within credit and collections.
What high-performing finance teams do next
- Map high-friction workflows (order-to-cash, close, disputes) and define risk thresholds.
- Stand up governance: model inventory, data lineage, monitoring, exceptions, and escalation paths.
- Pilot in lower- to medium-risk areas first; scale only on evidence (accuracy, cycle time, control adherence).
- Keep human review for revenue recognition, close, and external reporting until error rates are proven low.
- Clean and consolidate data sources; centralize contracts and standardize fields before automating.
- Measure outcomes: forecast accuracy, dispute resolution time, rework rates, and quarter-end overtime.
- Upskill the team on prompt reviews, validation techniques, and updated controls.
Bottom line
AI is already useful in finance, but only for teams that lead with governance, data discipline, and focused pilots. The companies above aren't chasing novelty - they're removing friction with controls intact.
If you're evaluating practical tools for finance, here's a curated list: AI tools for Finance.
Your membership also unlocks:
