AI is taking over finance from the bottom up, not the top down

A peculiar conflict is unfolding in the finance departments of large companies: employees are adopting AI tools en masse to work faster, while leadership rushes to catch up with a policy for governing the technology. The result is paradoxical: one of the most regulated sectors of the economy is going through a spontaneous AI revolution.

How AI is entering finance

Finance staff already use AI for tasks that used to take hours. Analysts upload quarterly reports to ChatGPT and other LLM apps to extract key figures. Risk specialists run AI models to assess loan portfolios. Teams automate document processing: scanning contracts, KYC checks, data extraction. All of this happens quickly and often without the explicit consent of compliance departments.

Employees see the efficiency gains: processing a document takes minutes instead of an hour. Portfolio calculations run several times faster. It works, and people are doing it.

Why leadership is lagging

CFOs and chief risk officers recognize the problem too late. By the time they start writing an AI usage policy and audit requirements, half the department already depends on ChatGPT for daily work.

Challenges facing leadership:

• Lack of visibility: it is hard to track which tools each employee uses
• Conflict with regulators: financial regulators require an explanation of the sources and reliability of the algorithms in use
• Risk of data leaks: employees may upload confidential reports to public services
• Accountability for errors: if an AI model miscalculated a risk, who is responsible?

Add to this a staffing constraint: there are no specialists who understand AI deeply enough to assess it properly.

Regulatory risks

Regulators are already paying attention to the problem. Financial supervisors are beginning to demand that banks and investment funds explain which AI systems they use, how those systems are tested, and who is responsible for errors.

"Financial institutions must have full visibility into AI use, from

tool selection to validation of results," regulators say, but there are no standards or rules yet.

This creates a wave of retroactive regulation: companies are forced to document and audit after the fact what has already happened. Some financial institutions have already received reprimands from regulators for using unvetted AI solutions in critical processes.

What this means

The financial sector shows that regulation always lags behind technology. By the time a governance framework starts to take shape, dozens of solutions that nobody approved are already running in practice. Ahead lies long work to bring order and develop standards that balance innovation with risk management.