The Accountability Edition

AI is becoming more capable, more accessible, and more deeply embedded in professional work.

This week, the AICPA released new guidance addressing AI in financial reporting and financial statement audits. IRS expectations for responsible AI use in federal tax practice received renewed attention. Google also expanded access to a frontier model designed for complex work across finance, tax, legal, cybersecurity, and other professional domains.

Together, theses developments point to an important shift.

The next state of AI adoption will not be defined only by who has the most capable tools. It will be defined by who can use those tools within reliable, reviewable, and professionally responsible workflows.

As AI moves closer to consequential work, governance cannot remain separate from implementation.

- Brianne

What Happened This Week

AICPA Releases Guidance for AI in Financial Statement Audits

• Why it matters: On October 13, the AICPA issued a practice aid addressing how an organization's use of AI in financial reporting may affect an audit, as well as how auditors should approach their own use of AI-related technology. The guidance covers matters including risk assessment, controls, data reliability, quality management, supervision, review, professional judgment, and professional skepticism. For accounting firms, the message is clear: AI does not sit outside the existing assurance framework. 
• What to do: Map where AI may touch the work, including client systems and third-party platforms. For each point, identify the relevant data, internal control, potential risk, evidence requirement, and human reviewer.
• Source/Further Reading: Journal of Accountancy

IRS Expectation Put AI Governance Into Practice

• Why it matters: The IRS Office of Professional Responsibility's guidance applies existing Circular 230 responsibilities to AI-assisted federal tax practice. Those responsibilities include competence, due diligence, written advice, firm procedures, appropriate fees, confidentiality, and verficiation of AI-generate facts, calculations, and citations. The practicel issue for firm leaders is no longer simply whether employees may use AI. It is whether the firm has documented which tools are approved, what information may be entered, how work must be reviewed, and who is accountable for the final result.
• What to do: Take one common tax workflow and document:
  • The approved AI platform
  • The information the platform may access
  • The information that must never be entered
  • The required verification steps
  • The documentation that must be retained
  • The person responsible for final review
• Source/Further Reading: Thomson Reuters

Google Expands Access to Gemini 4 Argon

• Why it matters: Google made Gemini 4 Argon more broadly available on October 13, positioning it for complex, multistep work in areas including finance, tax, legal, cybersecurity, science, and software development. This is another sign that frontier-level capabilities are moving closer to everyday business workflows. But increased capability does not automatically create a reliable process. Firms still need approved sources, defined permissions, clear review standards, and a record of how consequential outputs were produced.
• What to do: Do not begin with, "what can this model do?" Instead, begin with:
  • What specific task are we improving?
  • What information should the model use?
  • What evidence must support the result?
  • What can it recommend but not decide?
  • At what point must a professional take over?
• Source/Further Reading: Google

One Action This Week