AI In Finance: What Finance Leaders Are Telling Us

Last week we hosted AI, Automation and Finance Transformation: From Hype to Impact in Dallas with RSM. This article shares what finance leaders told us, what the latest research says, how to connect AI to finance systems safely, and how to build an AI business case that gets signed off.

Before the event we surveyed attendees. 85% said they want practical AI adoption strategies. Yet 70% rated their own AI adoption at just 1 or 2 out of 5. Accounts payable and month-end reconciliations topped the list of most manual finance processes.

That gap between ambition and adoption shaped the narrative throughout our event. Here is what we heard, how it lines up with the latest research, and what we think finance leaders should do next.

Key Takeaways

• Most finance teams now use AI yet realized value is lagging CFO expectations.
• The biggest blockers are governance, no internal champion, and collecting AI ideas instead of building a plan.
• Conversational intelligence lets finance teams ask their systems questions in plain English, but it delivers most when built into specific daily workflows.
• Finance leaders find a redeployment business case, with clear baselines and measurable targets, the easiest to get signed off.

How Widely Is AI Used in Finance in 2026?

Our Dallas room is not unusual. Across the market, almost every finance team is using AI in some form, but far fewer can point to real business value.

• Adoption is now the norm. Consero's 2026 CFO Survey of 102 investor-backed finance leaders found 42% have AI broadly or fully embedded in finance, up from 22% a year earlier (Consero).
• Value isn't keeping pace. Gartner's 2026 Finance AI Report Card, based on a March 2026 survey of 204 finance leaders, found adoption advancing while realized value trails CFO expectations. 63% said AI implementation was slower than expected in 2025 (Gartner).
• Few are satisfied. Bain's 2026 CFO survey found just 31% of CFOs are satisfied with their AI outcomes, rising above 60% among the most AI-mature companies (Bain via Consultancy-me).

Gartner's message to CFOs was blunt: counting pilots and tools rolled out shows activity, not impact. That matches exactly what we heard in Dallas.

Why Is AI Adoption in Finance Stalling?

When we asked why adoption scores were so low, the answers were rarely about the technology itself. Three themes came up.

Governance. Many leaders want to move but don't yet have clear rules on which tools are approved, what data can be used and who signs off. Without that, AI stays stuck at the experiment stage.

No champion. In most teams, nobody owns AI adoption. It sits between finance and IT, and without a named person driving it, initiatives stall.

"We need AI ideas" instead of a plan. Several teams are collecting AI ideas rather than starting from a problem. A list of possible use cases is not a strategy.

What Is Conversational Intelligence in Finance?

Conversational intelligence is the ability to ask your finance systems a question in plain English and get back the relevant data right away, instead of waiting for a pre-built report. It was the topic that sparked the most discussion in Dallas.

Traditional reporting gives you the reports someone built in advance. They rarely answer the question you have today. A new question usually means exporting from the ERP, exporting from AP, matching in Excel and then investigating the gaps.

Conversational intelligence turns that around. You prompt your systems in plain English and get back the data that matters right now:

• Which goods were received but not invoiced at month-end?
• Which suppliers invoiced above agreed PO prices?
• Which invoices may have incorrect tax treatment?

The AI draws on connected data such as the vendor master, PO pricing, goods receipts and AP invoices, with the ERP at the core, and returns the answer in minutes instead of days.

What we heard: plenty of leaders are already doing a version of this, often with a general-purpose assistant such as Claude bolted onto their systems. It's useful for ad hoc questions. But it isn't yet part of the everyday process for specific functions. The value comes when conversational intelligence sits inside the workflow, so the AP team asks and acts in the same place, rather than in a separate tool on the side.

Is It Safe to Connect AI to Your Finance Systems?

One of the biggest surprises of the day was how many finance leaders have already connected AI assistants directly to their finance systems. That raised an obvious question in the room: is that safe?

The honest answer is that it depends on how it's done. Connecting an AI tool to your ERP or AP data is not risky in itself. Doing it without controls is. The questions to ask:

1. Which account is it? A company-approved business or enterprise plan, with clear data-handling and retention terms, is very different from an individual's personal or free account.
2. What can it see and do? Give the minimum access needed. Read-only access to reporting data is far lower risk than access that can create, approve or post transactions.
3. Who built the connection? Connectors and integrations should come from trusted sources and be reviewed by IT, not set up ad hoc by individuals.
4. Is it logged and owned? Every connection should be on an approved list, with a named owner and an audit trail of what was accessed.

Kurt Meier from RSM shared how RSM approaches this internally within a closed, tightly secured environment. That's the model to aim for. Give people powerful tools inside approved guardrails, so they don't go around them.

How to Build an AI Business Case for Finance

We asked the room which AI business case is easiest to get signed off: scale (more volume, same team), redeploy (less processing, more value) or reduce (lower operating cost). The answer was clear. Redeploy won by a distance, with real interest in the cost savings that come with it.

Interestingly, despite recent headlines about AI-linked layoffs at large tech companies, headcount anxiety barely came up. Finance leaders want to show value. What they're less sure about is where to start.

Our view: the business case comes from the measurable value you create with the time that you save rather than just saving time itself. An example we shared from one of our clients on the day: 78% of 5,000 monthly stock invoices are now processed touch-free, returning around 325 hours a month. Purchase ledger clerks have moved into analyst roles, identifying savings and negotiating early-payment discounts.

Most AI business cases in finance are written around hours saved. The ones that get approved, and deliver, name what that time will be used for: capturing early-payment discounts, tighter credit control, or supporting acquisitions without new hires. Time saved is the input. Measurable value is the outcome.

A simple framework for an AI business case

"We need to use AI" won't get signed off. Instead:

1. Start with a high-friction process, not an AI idea. Month-end accruals and AP invoice processing are common starting points.
2. Quantify the baseline. Cost, time, error rates and how many people are involved today.
3. Assess data readiness. Is the data standardized, or chaotic? Be honest, but don't wait for perfect.
4. Define success precisely. "Automate AP" is vague. "Automate AP invoice data entry to save 8 hours a week and redeploy that time to finance business partnering" is a business case.
5. Agree it with your AI vendor up front. Documented as-is and to-be processes, mutually agreed success criteria and known risks, before the commercial proposal.
6. Measure it. Track FTE hours saved, dollar savings, the % of the process fully automated, and progress against the original objective.

Where Should Finance Teams Start With AI?

If the conversations in Dallas had one lesson, it's that the tools are ready before most finance teams are. Three steps close the gap:

1. Name a champion. One person who owns AI adoption in finance and works hand in hand with IT.
2. Put governance in place early. Approved tools, clear data rules and secure connections, so people use AI inside the guardrails, not around them.
3. Pick one process and prove it. Start with a high-friction process like AP, set a measurable target, and build from there.

Let's talk

We'd love to hear how you're applying AI in your finance function, whether you're already piloting or still working out where to start. Book a 30-minute call with our team.