Finance is one of the most promising areas for AI adoption, yet also one of the most sensitive. Sergey Smirnov, Group Financial Controller at DataArt, shares the four-step approach that helped the company’s finance function adopt AI and achieve measurable business results. Based on hands-on experience, he explains where AI creates real value, where it introduces risks, and why human judgment still remains essential.

AI in Finance: Between Hype, Real Value, and Responsibility

In boardrooms and financial departments, AI is increasingly positioned as a solution to efficiency challenges, talent shortages, and growing compliance pressure. Finance, however, is one of the few areas where AI can simultaneously deliver exceptional value and introduce significant risk.

AI is often presented as a silver bullet: faster closes, flawless forecasts, and automated compliance at the push of a button. In reality, finance and accounting operate under unique conditions. They are highly data‑intensive, heavily regulated, and unforgiving of errors. Unlike other business functions, finance cannot afford confident outputs that are only “almost correct.”

This creates a paradox. Finance is one of the most suitable domains for AI adoption, yet also one of the most sensitive. The real challenge is not technological capability, but governance.

Used correctly, AI removes friction from finance work. Used incorrectly, it introduces silent risk.

What AI Can Actually Solve in Finance and Accounting

When stripped of marketing claims, AI delivers the most value in finance where processes are repetitive, data‑heavy, and rule‑driven. Its strongest impact is in automating preparation work, supporting analysis, and accelerating research, while still relying on professional judgment for final decisions.

At DataArt, AI has demonstrated practical value across a broad range of finance and accounting activities, including payroll and HR reporting, billing and expense verification, transfer pricing documentation support, tax research and drafting, audit preparation, and internal finance knowledge management.

Effectiveness has been measured by business outcomes rather than experimentation. As a result of structured implementation, the finance function has achieved the following improvements:

Reduced manual effort in recurring reporting and payroll processes: Automating initial preparation and consistency checks allows accountants to focus on exceptions and validation instead of manual compilation.

Faster document generation in payroll and HR workflows: Drafting standard payroll and HR documents, confirmations, and internal explanations significantly shortens preparation cycles while retaining mandatory human review.

More efficient billing and expense verification: AI assists with separating, reviewing, and flagging transactions so reviewers can focus on anomalies rather than routine items.

Streamlined transfer pricing documentation support: Structuring reports, preparing narratives, and organizing supporting data reduces reliance on purely manual work and external support.

Accelerated tax research and drafting: AI supports initial reviews of legislation, guidance, and case law, enabling tax professionals to concentrate on interpretation and judgment rather than information gathering.

Improved internal finance knowledge sharing: Easier access to internal policies, prior solutions, and documented practices reduces reliance on individual expertise and increases consistency across teams.

These improvements were achieved without compromising control, auditability, or quality. When implemented responsibly, AI strengthened process discipline rather than weakened it, allowing the finance function to scale efficiency while maintaining rigor in regulated and audit‑sensitive environments.

The Hidden Risks Finance Leaders Must Actively Manage

Despite its benefits, AI introduces risks that are particularly dangerous in finance if left unmanaged.

Key concerns include:

  • Hallucinations: Outputs may sound confident while being factually incorrect.
  • Regulatory inaccuracies: Jurisdiction‑specific tax or accounting rules may be outdated or misapplied.
  • False precision: Numbers can appear internally consistent while being entirely wrong.
  • Data confidentiality risks: Improper tool usage may expose sensitive financial information.
  • Over‑reliance by junior staff: Early‑career professionals may trust outputs without sufficient critical review.
  • Auditability challenges: AI outputs must remain explainable and traceable.

In finance, perception matters as much as accuracy. A polished but incorrect output can be more dangerous than an obvious error.

Human Supervision: A Non‑Negotiable Principle

A fundamental principle governs responsible AI use in finance: AI supports professional judgment, but never replaces it.

Every AI‑generated finance output must:

  • Be reviewed by a qualified finance professional
  • Be verified against source data
  • Be assessed for regulatory and jurisdictional correctness
  • Receive human approval before operational or external use

AI is treated as an advanced assistant, not an authority. This approach is critical for financial statements, tax positions, transfer pricing documentation, audit responses, and management reporting.

For finance leaders, human supervision is becoming a core element of modern financial governance.

How AI Has Been Implemented in Finance

Instead of deploying tools first and addressing governance later, we adopted a structured, people‑first approach.

Education Before Automation

AI adoption began with internal presentations and workshops for accountants, focusing on capabilities, limitations, and responsible usage in financial contexts. The goal was understanding, not blind adoption.

One‑to‑One Enablement

Individual feedback sessions and working meetings helped tailor AI use to real processes. These sessions built confidence while reinforcing the importance of professional judgment and verification.

Knowledge‑Sharing Culture

Accountants actively shared prompts, use cases, and lessons learned across the department. This ensured consistency, transparency, and avoided fragmented AI practices.

Internal AI Support

A helpdesk chatbot based on the internal knowledge base reduced reliance on uncontrolled external sources and supported data confidentiality requirements.

Why AI Adoption Worked

Successful AI adoption in finance is driven not by technology, but by disciplined processes and governance.

AI delivered value because:

  • Accountants were trained, not replaced
  • Governance was defined before scaling
  • Use cases were approved and reviewed
  • Human verification was embedded in every process
  • AI literacy became part of the finance culture

Technology accelerates processes, but process design determines outcomes.

The Road Ahead for AI in Finance

AI will become a standard finance tool, much like spreadsheets or ERP systems did in the past. Competitive advantage will not come from adopting AI faster, but from adopting it responsibly.

For modern finance leaders, the role is evolving into that of an AI supervisor – ensuring that technology enhances accuracy, improves efficiency, and strengthens governance rather than undermining it.

The future of finance is not a contest between humans and machines. It is a collaboration where AI removes friction and humans provide judgment. In finance, responsibility scales with automation – and so does accountability.