AI in Accounting Firms: Governance for Audit, Tax, and Advisory Practices

AI in audit: independence and quality

The introduction of AI tools into audit practice creates specific independence considerations that do not arise with conventional audit tools. Independence standards (APES 110 Section 290 and Section 291 in Australia, and the IESBA Code internationally) prohibit auditors from having interests or relationships that could compromise their independence from audit clients. AI tools that are connected to client systems — for data extraction, analytics, or continuous monitoring — must be assessed for whether that connection creates an independence threat. AI tools that are trained on client data create a different type of independence concern: if the AI's outputs for one client are influenced by data from another, or if the AI develops expertise about a client's business through training data, this may affect the independence analysis.

Quality management under ISQM 1 requires firms to establish a system of quality management that includes policies and procedures for the use of technology in audit engagements. AI tools used in audit procedures must be within this system — they must be assessed for reliability, have defined scope of appropriate use, be subject to partner oversight, and be covered by the firm's quality monitoring programme. Deploying AI in audit engagements without bringing it within the ISQM 1 quality management framework creates quality management non-compliance.

AI in tax practice

Tax AI is being used for: tax research (identifying relevant legislation, rulings, and case law), tax calculation (modelling the tax implications of transactions), tax return preparation (automating data extraction and return completion), and tax controversy (researching applicable arguments in disputes). Each use case has different accuracy requirements and different liability implications. Tax research AI that returns an incorrect analysis of a provision may lead to incorrect advice — the adviser remains liable for the advice regardless of the tool used. Tax calculation AI that makes computational errors creates the same liability. The professional standard is clear: AI assists, the professional is responsible.