Gargi Ray, VP Finance, Synopsys Inc.
Enterprise spending on artificial intelligence is soaring, yet many CFOs find themselves trapped in an uncomfortable paradox: massive capital allocation paired with small financial returns.
According to research by Boston Consulting Group (BCG) in their report “How to Get ROI from AI in the Finance Function,” the median reported ROI for AI in finance sits at a meager 10%—far below the 20%-plus target many executives set. Nearly a third of finance leaders report limited or no financial gains at all.
BCG’s study reveals that high-ROI finance teams can better capture real value by making four distinct strategic choices:
1. Focusing Relentlessly On Value: They target hard business impact from day one rather than testing tools for learning’s sake.
2. Broad Transformation View: They embed AI directly into their overarching finance strategy rather than building isolated, single use point solutions.
3. Active Collaboration: They partner closely with IT and existing vendors instead of trying to build everything internally from scratch.
4. Well-Sequenced Execution: They roll out changes in clear, targeted steps to capture real value continuously as they scale.
Real ROI can happen when finance leaders step up to own the AI strategy—turning AI-driven workflows into the single, mandatory way of working across the enterprise to speed up execution, scale operations and incur minimal expense.
The Intelligence Layer: Translating Judgment Into Code
Generic AI tools can fail in core financial operations because off-the-shelf software doesn’t usually understand complex business rules. To capture value across any financial workflow—whether in order-to-cash, financial planning and analysis (FP&A), record-to-report or procurement—finance teams must define the logic. Internal finance experts must design the decision frameworks and translate practical institutional knowledge into clear, repeatable skills.
Real Business Judgment And Edge Cases: Decision logic must handle complex cases, regional risk nuances and edge cases in a deterministic way.
Tool-Agnostic Setup: When finance owns the core logic and decision rules, the system isn’t locked into a single vendor. Underlying AI models can be updated or swapped over time without breaking core operations.
Smart Capital Allocation
A major reason some AI projects fail to show ROI is the heavy burden of new software licenses, expensive implementation consultants and unmonitored token usage.
Finance leaders must enforce strict capital discipline. Rather than buying stand-alone point solutions, finance teams should work directly with IT to deploy specialized AI skill agents on top of the enterprise AI capability already embedded in their existing tech stack.
Case Study: Transforming High-Volume Credit Risk Analysis
To illustrate how these principles can be put into action, consider how our team transformed a high-friction operational workflow: credit risk evaluation for high-volume sales quotes for a $10-billion-plus organization.
The Traditional Challenge: Fragmented Risk Evaluation
Credit risk assessment has historically been bottlenecked by manual data retrieval and fragmented analysis across disparate sources.
Internal Enterprise Signals
For my organization, implementing AI successfully required navigating disconnected ERP and CRM interfaces to reconstruct payment histories, DSO metrics, total account exposure and open sales pipelines. Even when dashboards simplified retrieval, synthesizing these metrics and weighing their trade-offs still relied entirely on subjective human judgment.
External Market Intelligence
Manually parsing unstructured third-party data—including credit agency reports, sovereign risk profiles, audited financial statements, bank statements and regulatory filings to analyze the credit risk of a client—can take two to three hours rather than 10 to 15 minutes, dragging down sales velocity.
Redesigning The Enterprise Workflow With Skill Agents
To eliminate the manual bottleneck, we engineered three specialized skill agents: an internal credit data agent to pull ERP/CRM information and apply parameter scoring to compute a master score, an external risk monitoring agent to contextualize agency reports and a financial deep-dive agent to analyze statements and output structured risk scores.
By deploying native agents on top of pre-funded data platforms, core databases and enterprise security frameworks, the true investment isn’t vendor software—it is the deliberate focus, time and intellectual property of our internal finance leadership.
Deterministic Rules And Governance In Practice
Consistent Scoring Standards: Every input metric (e.g., DSO, liquidity ratios or sovereign risk) is checked against exact numerical scales designed by finance experts. The agents apply strict rules rather than generating open-ended text.
Predictable Master Outcomes: Subscores from all agents feed into a clear formula. Identical financial inputs will always produce the exact same score and recommendation, guaranteeing complete consistency.
Safety Stops And Human Control: If key data or required documents are missing, the workflow halts and alerts a human analyst. The multi-agent setup generates a baseline recommendation (Can Approve or Risky/Escalate), but a human expert retains final decision authority.
Complete Audit Trail: To satisfy audit and compliance rules (including SOX), every transaction generates a unique ID, storing all source data, reasoning steps and analyst approvals in an immutable log.
Scalable Impact Across The Enterprise
Whether applied to credit reviews, contract variance analysis, capital allocation modeling, scenario-based forecasting or vendor invoice matching, applying this framework can yield immediate, scalable impact:
Accelerated Cycle Times: Reduction in review cycle time reviews may drop from hours to minutes, driving faster execution across the sales cycle.
Consistent Risk Control: Codified rules can help enforce corporate policy uniformly across global regions, reducing subjective bias and financial exposure.
Nonlinear Operational Scale: Transaction volume can double or triple without requiring a linear increase in finance headcount.
Security And Governance Native To The Stack: Building native AI workflows directly within our existing, enterprise-governed architecture preserves strict data boundaries and satisfies demanding compliance and SOX audit requirements out of the box.
Minimal Infrastructure Cost: Because these native agents leverage preexisting cloud and data infrastructure, external software costs, both one time and ongoing, remain minimal. The main investment is our team’s domain expertise in building the intelligence layer, allowing transaction volume to scale exponentially without inflating our software budget or adding headcount.
When finance leaders focus on true value from day one, collaborate with IT and embed clear rules into existing infrastructure, AI can become a dependable engine for enterprise growth.
The information provided here is not investment, tax or financial advice. You should consult with a licensed professional for advice concerning your specific situation.
Forbes Finance Council is an invitation-only organization for executives in successful accounting, financial planning and wealth management firms. Do I qualify?