AI ROI in finance: How leading CFOs measure success

The question CFOs are asking isn't "Should we use AI?" but "How do we prove it's actually working?"

While 56% of finance leaders now use AI in their work – a figure that's doubled year on year – considerable scepticism persists. As Phil Sharp, Interim CEO & CMO at Subscript, puts it: "There's widespread scepticism about AI. I rarely hear, 'AI is changing my life.' It's more like, 'This seems useful, but is it snake oil?'"

The root of this uncertainty? Most organisations are measuring AI success the wrong way.

The problem with time saved as your only metric

Ask most finance teams how they measure AI ROI, and they'll point to time savings: "This task used to take 30 minutes, now it takes 30 seconds."

It's compelling. It's quantifiable. And it's inadequate.

"If we only measure AI by time saved, we miss its real value – capabilities and quality," warns Dan Zhang, CFO at ClickUp.

Time savings capture only the most obvious benefit. But they ignore AI's broader value: better decision-making, new business capabilities and improved quality of work.

The three-bucket framework for AI ROI

ClickUp's CFO Dan Zhang has developed a framework that categorises AI investments into three distinct value types:

1. One-to-ten automation: Time savings

This is the traditional efficiency play. It's about automating repetitive tasks that used to require manual effort.

Examples include:

• Contract analysis that took hours now completing in minutes

• Reconciliations that previously required days

• Automated expense categorisation

How to measure: Hours saved per week, reduction in close cycle time (days after month-end: from D+5 to D+2), and decreased time-to-insight for stakeholders.

"Before, closing books in D+1 or D+2 was reserved for large corporations," says Spendesk CFO Pauline Babel. "AI is now enabling so much automation that it becomes possible for companies that cannot afford heavy and expensive tooling."

2. Zero-to-one capability: New possibilities

This is where AI's true value becomes clear. It enables work that was previously impossible due to resource constraints.

Examples include:

• AI agents engaging small customers where unit economics previously didn't make sense

• Knowledge retrieval systems answering complex questions from internal datasets

• Real-time spend analysis instead of waiting for month-end

"One use case is the AI BDR, engaging the long tail of small customers we've never touched before because the unit economics doesn't make sense," explains Zhang. "That's a zero-to-one capability – something entirely new."

How to measure: New business opportunities created, expansion into previously unviable customer segments, and business questions answered that couldn't be addressed before.

3. C-to-A quality boost: Better insights and decisions

AI doesn't just make work faster. It makes it better. This bucket captures improvements in accuracy, quality and business value. Think of it as elevating work from C-grade to A-grade quality.

Examples include:

• More accurate forecasts with less variance

• Improved stakeholder satisfaction

• Reduced error rates

• Enhanced employee satisfaction as teams move from manual to analytical work

How to measure: Forecast accuracy improvement, error rates in financial reporting, stakeholder satisfaction surveys, and employee engagement metrics.

Zapier's approach: Objective plus subjective metrics

At Zapier, where 98% of employees use AI tools, CFO Ryan Roccon balances hard metrics with qualitative feedback.

Objective metrics: Cycle times, pull requests merged, issues resolved, and transaction processing speed.

Subjective metrics: Employee satisfaction surveys, stakeholder feedback, and quality assessments.

"It begins with clean data collection and segmentation," Ryan explains. "We pinpoint where AI adds the most value: in code writing, refinement or QA."

The key insight: you can't rely solely on quantitative measures. Employee sentiment is a leading indicator of whether AI adoption will stick.

Real-world results: What's working

Finance leaders across industries are already seeing measurable returns from AI implementation.

OpenAI automated technical accounting with a ChatGPT agent that extracts contract terms, drafts memos and generates journal entries. "It's a fraction of the work of hiring and spinning up an entire team," says Yubo She, Head of Technical Accounting.

Adyen built a knowledge retrieval system for sales teams. The result: instant answers globally without waiting for finance, freeing the team for higher-value work.

Vitalize Health had an accountant who automated himself out of reconciliation and moved into FP&A. "He feels more fulfilled, and the company benefits from his new focus," says Rebeca Bichachi.

Common pitfalls to avoid

Even with the right metrics in place, most finance teams stumble in predictable ways. Here's what to watch out for:

Waiting for the perfect ROI model. "There's definitely no playbook," says OpenAI's Jessica Pillow. "We're treating it like a product: constantly testing, iterating and making small movements." Don't let analysis paralysis prevent progress.

Measuring only what's easy to quantify. Time savings are easy. Quality improvements are harder to measure but often more valuable. Use a mix of methods.

Ignoring employee sentiment. "Employee satisfaction is a great leading indicator for if this AI tool is going to be well adopted," says Dan Zhang. If your team hates the tool, it won't deliver ROI.

How to start measuring AI impact: 30-day action plan

The best measurement framework is useless without a starting point. Here's how to build yours in four weeks:

Week 1: Establish your baseline

Document the current state before implementing AI:

• Time spent on key workflows

• Current error rates

• Close cycle duration

• Employee satisfaction scores

• Forecast accuracy

Week 2: Define your target metrics

Choose one metric from each category:

• Operational (e.g. close cycle time)

• Business impact (e.g. forecast accuracy)

• People (e.g. employee satisfaction)

Set specific targets, such as "Reduce close from D+7 to D+3" or "Improve forecast accuracy from 15% to 8%."

Week 3: Pilot and track

Select one high-friction workflow for testing. Choose something with clear inputs and outputs. Implement the solution and track daily or weekly impact.

Week 4: Build your ROI dashboard

Create a simple tracking system showing:

• Time saved

• Quality improvements

• New capabilities created

• Employee sentiment

The path forward

AI ROI in finance isn't just about doing things faster. It's about doing them better, doing things you couldn't do before, and positioning finance as a business partner.

As Gabriel Hubert, CEO at Dust, frames it: "AI increases productivity in three ways: doing things faster, doing them better, and doing things you couldn't before."

The finance leaders seeing the greatest AI success aren't those with the most sophisticated frameworks. They're the ones who started measuring something, learnt from it and iterated quickly.

68% of CFOs say they've been slow to adopt AI because they don't know where to start. Don't let the pursuit of perfect ROI measurement become another reason to delay.

Start with one workflow. Track three metrics. Learn what works. Scale what delivers value.

The competitive advantage goes to those who begin, not those who wait for certainty.