MIT recently published a striking study: 95% of enterprise AI projects fail to deliver ROI.
At first glance, this makes AI look like a high-risk bet for finance leaders. But dig deeper and the issue is less about the technology itself, and more about how you define ROI.
Most organizations still measure AI the way they would an IT project. They use cost savings and headcount reduction. That’s a narrow lens.
Modern finance is about building resilience, speed, and better decision-making. If ROI is limited to “did we save FTE hours?”, the real value of AI in finance will always look invisible.
Why CFOs Need to Rethink ROI
This narrow view is creating two problems:
1. Missed strategic impact. AI can improve forecast accuracy, shorten close cycles, and reduce working-capital drag. These are balance sheet and P&L levers, not IT metrics.
2. Credibility gap. When AI pilots only track token savings, boards and executives lose trust in the technology’s ability to drive business outcomes.
The paradox, as MIT points out, is that while most AI projects fail, the small fraction that succeed generate outsized value. Because they treat ROI differently, less like a software pilot, more like a financial system upgrade.
Measuring ROI Through the CFO Lens
So, what should CFOs actually measure? The answer is to align ROI with the financial system’s performance. Think beyond tools and features, and look at the metrics that move enterprise value.
Operational impact:
– Days to close: Has AI reduced the reporting cycle by 25–30%?
– Forecast accuracy: Are MAPE/WAPE scores improving quarter by quarter?
– Variance explanation: Can the system surface drivers in hours instead of days?
Strategic impact:
– Scenario velocity: How quickly can the team build and test three alternative cases?
– Capital allocation: Are IRR assumptions more accurate against realized outcomes?
– Pricing and mix optimization: Has AI flagged margin opportunities earlier than traditional analysis?
Risk and control:
– Exception detection: Is AI catching anomalies faster than manual controls?
– Audit trail: Can outputs be explained and trusted by auditors?
– Model drift monitoring: Is governance in place to ensure reliability over time?
Adoption:
– User engagement: How many finance professionals actively rely on AI outputs?
– Assisted actions: How many insights are acted upon, not just generated?
These measures speak the language of the boardroom. They also provide a balanced scorecard that captures both efficiency and strategic value.
Time Horizons Matter
ROI from AI won’t appear overnight. CFOs should set expectations in phases:
– Near term (0–90 days): The first tangible benefits usually show up in process efficiency: faster close cycles, automated variance commentary, and better adoption of dashboards by business users. These early wins prove that AI can save time and reduce manual effort.
– Mid term (3–9 months): Once the models stabilize, you’ll see sharper improvements in forecast accuracy and anomaly detection. This stage is where finance starts to trust AI outputs enough to use them in decision-making.
– Long term (9–18 months): Over time, AI becomes embedded in core processes. The real financial impact shows up in better margins, improved ROIC, and stronger working-capital efficiency. These outcomes position AI not just as a tactical tool but as a strategic enabler.
Data Readiness Is Half the Battle
It’s worth repeating what both MIT and McKinsey stress: AI without data discipline is useless.
– Standardize core definitions (ARR vs. NRR, SKU, customer, plant): Without alignment on how metrics are defined, AI models will produce inconsistent outputs that undermine trust across business units. A shared dictionary ensures comparability and accuracy.
– Establish data quality SLAs: Clear standards for timeliness, completeness, and accuracy prevent downstream errors. SLAs keep both IT and finance accountable for the integrity of data feeding AI models.
– Break down silos between FP&A, accounting, treasury, and operations: AI only adds value if it can connect the dots across the enterprise. Eliminating silos creates a single version of the truth and allows scenario planning that reflects real business conditions.
Treat AI Like an Engineering System
The lesson from the 5% who succeed is simple: don’t treat AI like a shiny demo. Treat it like software engineering. Build specifications and set guardrails. Run tests and establish feedback loops. And most importantly, measure ROI in terms of financial outcomes, not IT savings.
When CFOs apply this discipline, AI moves from experiment to trusted infrastructure.
Final Thoughts
AI in finance is about building systems that compound value over time. The right ROI lens transforms AI from a risky bet into a competitive advantage.
For finance leaders who want to accelerate this journey, external partners can help design ROI scorecards, align baselines, and implement governance frameworks.
At Centida, we support companies in structuring AI initiatives with measurable outcomes, ensuring investments drive both short-term wins and long-term enterprise value.