Gaurav Vashisht has 22+ yrs building and supporting Finance Systems. AAAI · IEEE · BA · ACM.
The most common failure mode I’ve seen in AI-driven finance transformation isn’t a technology failure. It’s a governance failure disguised as a technology success: the automation works, the close is faster and the metrics look good right up until the audit, when it becomes clear that no one thought to ask what happens when a system makes a decision that a human has to own.
Material deficiencies have a way of answering that question for you. The honest answer, in more cases than anyone likes to admit, is: the system did. But in a regulated environment, that answer isn’t acceptable.
The Pattern I Keep Seeing
Finance transformation has become one of the most overpromised initiatives in enterprise technology. This is not because the underlying tools are bad—many of them are genuinely impressive—but because organizations keep deploying AI on top of architectures that were never designed to support it.
The pattern is everywhere: an intelligent AP automation tool running on top of a chart of accounts that hasn’t been cleaned in eight years, a predictive close solution pulling from three different source systems that reconcile manually every month, a generative AI assistant for FP&A that can’t answer basic questions because the data model it queries is inconsistent across entities—I could go on and on.
The tools are smart, but the foundation underneath them isn’t, and you can’t transform a finance function by making the surface layer faster while leaving the structural problems intact.
You Can’t AI Your Way Out Of A Data Problem
The first things I ask when a company tells me they’re doing finance transformation is: What does your data model look like? How many ERPs are you consolidating from? How clean is your chart of accounts? How do you handle intercompany eliminations today? The answers are usually uncomfortable.
AI in finance is downstream of data quality. Always. A machine learning model trained on inconsistent general ledger data produces confident-sounding noise. The more sophisticated the AI, the harder that noise is to detect.
The organizations getting real, durable results from finance AI are the ones that invested in data unification before they invested in AI tooling. That’s unglamorous work—nobody gets a press release for cleaning a chart of accounts—but it is the work that makes everything else possible.
Governance Isn’t A Compliance Tax—It’s Architecture
Here’s a conversation I’ve had more times than I can count. A finance or IT leader wants to move fast on AI automation, but the controls team pushes back. The response is usually some version of, “We’ll handle compliance later; let’s prove the value first.”
This is exactly backwards.
For any company operating under SOX, preparing for an IPO or managing any meaningful audit exposure, the governance implications of AI-driven finance should not be a downstream concern. If your AI system is making autonomous journal entries, classifying transactions or generating financial narratives, the question of who is responsible for reviewing and approving those outputs needs to be answered before you go live, not during your next audit. Human in the loop is a must.
Governance architecture—role-based access, automated control testing, change management, audit trails—needs to be a first-class citizen in the design of any AI finance system. Not a feature to add later. Not a checkbox. Architecture.
Finance Doesn’t Operate Alone
One of the structural limitations of most finance AI initiatives is that they’re scoped as finance initiatives.
But revenue recognition depends on how contracts are structured in legal. Headcount expense depends on how positions are managed in HR. Procurement controls depend on approval workflows that cross IT and operations. Finance accuracy is a cross-functional problem, and solutions scoped only to the finance function will always hit a ceiling. It’s a combined cross organization effort.
The CIO’s value in finance transformation isn’t as a technology vendor to the CFO, but as the person who can see the connective tissue across the enterprise and who understands that a clean finance AI deployment requires aligned data models in HR, legal and procurement as well. That cross-functional view is hard to develop and easy to underestimate. It’s also where most of the actual transformation value lives.
Build For Where You’re Going, Not Where You Are
Here’s the last thing I’d say to any CIO or finance leader starting this journey: The architecture you build today will either enable or constrain every business event that comes after it, including acquisitions, a new product line, new geography and an IPO.
Digital asset finance sits at the most architecturally demanding intersection in enterprise technology right now, right between novel instrument types that no ERP was designed to handle, transaction volumes settling in milliseconds, regulatory frameworks that are still being written and audit requirements that don’t map cleanly to anything in a traditional chart of accounts. The pressure to move fast is constant.
What keeps us grounded is asking a simple question before every major systems decision: Will this still work at five times the scale, with 10 additional entities, under a public company control environment? If the honest answer is no, we redesign before we build.
What This Actually Requires
I’m not arguing against AI in finance. I’m arguing for doing it in a way that actually works.
That means treating data architecture as a prerequisite, not an assumption. It means designing governance into the system, not around it. It means accepting that finance transformation is a cross-functional program, not a departmental upgrade. And it means building for the organization you’re becoming, not the one you are today.
None of that is easy, but the companies I’ve seen do it well usually have a leader sitting at the intersection of technology, finance and enterprise architecture who is willing to get the foundation right.
That leader, more often than not, is the CIO. The question is whether they’re in the room early enough to matter.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?