Sage announced new research on July 6 showing 71% of finance leaders would reject an AI tool that was 99% accurate if it could not explain its answers. The findings come from The Emerging Economics of AI in Finance, an IDC whitepaper sponsored by Sage and based on a global survey of 2,275 senior finance decision-makers and influencers across North America and EMEA.
That cuts against the usual AI sales pitch. Vendors often lead with performance, speed, automation, or model intelligence. Finance leaders are saying the output still has to be explainable before it can be trusted in workflows tied to reporting, compliance, audit, and financial decision-making.
Sage also found more than half of organizations would pay more for AI that provides greater visibility into how outputs are generated. That pushes the explainability factor from a product preference into a procurement issue.
This is the finance version of the broader AI cost reset now unfolding across the enterprise. Companies are asking whether AI usage produces measurable value. Finance teams are asking a more specific question: Can the organization defend the answer after AI produces it?
Analysis
What this means: Finance AI has a trust tax. If teams have to spend hours reconstructing how an answer was produced, the automation gain starts leaking away through review, validation, and audit preparation. The question is not only whether AI can generate a correct result, but whether the finance team can explain that result without becoming a detective agency.
The Verification Tax
Sage’s research puts a number on the work finance teams do after AI gives an answer.
Finance professionals spend nearly 13 hours each week reconstructing, validating, and defending AI outputs, according to the study. In the US, 49% spend 15 or more hours a week on verification, while 19% spend 30 or more hours.
That is the hidden cost behind black-box AI in finance. The model may produce an answer quickly, but the organization still pays if a person has to reverse-engineer the reasoning, check the source data, validate the calculation, document the decision, and prepare for audit questions.
In high-stakes finance workflows, “almost right” can still create material risk. An unexplained variance, forecast, cash-flow recommendation, revenue-recognition output, or payment decision may not be usable if the team cannot trace how the system reached it.
That makes explainability an operating requirement, not just a user-experience feature.
Analysis
What this means: Explainability needs to be part of the finance business case. A cheaper or faster AI tool may still lose if it creates more review work than it removes. Finance teams will increasingly compare vendors by the cost of confidence—how much human effort it takes to trust, document, and defend the output.
Black Box, Bigger Bill
Sage is positioning the research around the shift from black-box AI to glass-box AI.
The company describes glass-box approaches as systems that provide visibility into the reasoning, sources, and logic behind AI-generated recommendations. The research found 71% of finance leaders say a vendor’s shift to glass-box design principles would strongly or critically elevate its status as a preferred partner.
Finance AI is moving closer to actual work. AI is no longer being used only to summarize information or produce draft analysis. It is being considered for complex workflows where outputs may influence reporting, planning, forecasting, procurement, cash management, and compliance.
The closer AI gets to the financial record, the less tolerance finance teams will have for unexplained outputs. A fast answer that cannot be audited may simply create a new bottleneck.
Analysis
What this means: Finance will be one of the hardest places for black-box AI to survive. A 99% accurate tool can still fail if the remaining 1% creates audit exposure, compliance risk, or hours of manual verification. The winners in finance AI will be the systems that make trust cheaper, not the ones that only make answers faster.
Get Our Free Weekly Newsletter
Finance Wants Final Say
The study also shows how AI is changing what finance leadership skills are worth.
When asked what skills mattered most for a finance leader hired today, US-based respondents ranked risk, governance, and decision judgment first at 32%. That was nearly twice the share that selected deep technical accounting, at 17%.
That does not mean accounting expertise matters less. It means AI is shifting more of finance’s value toward judgment, control, and accountability. As AI takes on more analytical and operational work, finance leaders have to decide when outputs can be trusted, when humans need to intervene, and how decisions will stand up to audit or board scrutiny.
That is a very different role from simply adopting AI tools. Finance teams want to decide whether AI-generated answers are safe enough to enter the business record.
For ERP vendors, this creates a product challenge. Finance AI cannot be sold only as an assistant that gives better answers. It has to show the evidence behind those answers, preserve controls, track decisions, and support the human judgment finance teams still own.
Procurement Asking Harder Questions
Sage’s research suggests transparency is becoming a vendor selection factor.
Fifty-four percent of organizations said they would pay a premium for AI that gives greater visibility into how outputs are generated. That is an important signal for ERP and finance software buyers because it changes the procurement conversation.
Finance teams will still care about accuracy, usability, automation, and productivity. But they will also want to know how the AI system handles source data, reasoning traces, confidence, exceptions, approvals, audit logs, and human oversight.
The vendor that cannot answer those questions may struggle, even if its model performs well enough.
Analysis
What this means: The AI buying checklist is getting more financial. Procurement teams will not only ask what the model can do; they will ask how the finance team proves it was right. That puts pressure on vendors to package transparency, auditability, and control into the core product rather than sell them as governance extras.
