Top 6 questions finance leaders are asking about AI (with practical answers)

Finance leaders aren’t in denial about AI. Most understand it will reshape their roles and change how their teams operate. At the same time, many are still searching for finance-specific use cases and bring legitimate questions about accuracy and security.

Those concerns came through in the candid questions hundreds of finance leaders asked throughout our recent workshop, “AI 2026 kickoff: 5 high-impact use cases for finance,” with AI Finance Club founder Nicolas Boucher. We’ve rounded up answers to the most common ones to help you get started.

1) Which AI use cases are actually working in finance right now (not theoretical)?

Start with tasks that involve repeatable work. Think cleaning data exports, building first drafts of reports or dashboards, or summarizing variance drivers

In the workshop, Boucher and Ramp Head of AI and Operations Ben Levick showed several practical examples:

• Dashboards: With a simple prompt and a spreadsheet of cost data, ChatGPT created a dynamic, filterable HTML dashboard in minutes.

• 3-statement models: Boucher used Excel’s Agent Mode to build a three-statement model and headcount plan across five years.
• Vendor spend analysis: Levick demonstrated how Ramp Sheets, an AI-powered spreadsheet built by Ramp Labs, can group messy vendor data by importance and recommend next steps.

Our advice: Pick one recurring spreadsheet-based task and see how AI can handle the first draft.

2) How do we make sure AI outputs are accurate enough for finance?

A useful mindset from the hosts: treat AI like a junior analyst. Give it clear context and instructions, then review its work carefully.

If AI can handle the first 80–90% of a task (drafting, structuring, summarizing, categorizing), you can focus on review and judgment. And save a lot of time.

Ways to keep accuracy high:

• Ask AI to explain its outputs (formulas, sources, reconciliation checks)
• Centralize and regularly update the knowledge sources it relies on
• Keep humans in the loop for anything that moves money or posts to the GL

3) Is it safe to use AI with confidential financial data?

It’s no surprise this was one of the most frequent questions among finance professionals.

In general:

• Free versions of GenAI tools can access and retain whatever data you share with them.
• Enterprise versions of LLM providers — the highest pricing tier, not to be confused with other paid plans — can usually be configured so your data isn’t used for model training or shared beyond your account. But options differ by platforms, so make sure you understand the nuances.

Ramp Sheets, while free to use, offers zero data retention. That means customer data is completely confidential and is never used to train any models.

4) Which tool should we choose: Copilot, Gemini, ChatGPT, Claude…?

There’s no single “best” AI tool for every company. Boucher’s recommendation: choose the one that best fits your existing environment. That means:

• If you’re a Microsoft shop (Excel, Outlook, SharePoint), Copilot is a natural fit.
• If your team runs on Google Workspace (Drive, Gmail, Calendar), Gemini is a strong option.
• If you’re tool-agnostic, expect the “best” model to keep changing, so optimize for usability and controls instead.

Adoption matters most, so bias toward tools that feel familiar and can safely access essential data.

5) Does Ramp Sheets integrate directly with Ramp or QuickBooks? Can multiple people collaborate in it?

Today, Ramp Sheets does not yet connect directly to Ramp’s spend management platform or accounting systems like QuickBooks. But this is on the near-term product roadmap and will roll out in the coming weeks.

Multi-user, real-time collaboration (like Google Sheets) is also on the roadmap. For now, teams can download their work, share it with colleagues, and those colleagues can re-upload files to maintain progress.

More than 10,000 finance teams are already using Ramp Sheets, proof it’s delivering value. To be the first to hear about new capabilities, follow Ramp Labs on X and Substack.

6) How do we turn AI workflows into something repeatable?

One of the most important lessons from the session: one-shot prompts won’t always produce great results.

Boucher demonstrated this with journal entry creation. The initial results were disappointing, so he asked the AI to try a new approach, added examples, and tried again. After several iterations, the model finally produced journal entries ready for ERP upload.

So was it worth the effort? Yes, because it was repeatable. Once the workflow worked, Boucher asked AI to generate instructions for an assistant that could repeat the process. He saved those instructions in a custom GPT that can painlessly automate this work every week or month.

The goal isn’t just getting AI to do the work once. It’s capturing the process so you don’t repeat the same effort next month.

Watch the full workshop

This only scratches the surface of what Boucher and Levick covered. For more real-world demos and battle-tested guidance on how AI can make a real impact for finance, watch “AI 2026 kickoff: 5 high-impact use cases for finance” on demand.