AI tools for accountants are already embedded in workflows across public accounting and corporate finance. The question in 2026 isn’t whether to engage with it but how to do so without creating professional liability exposure.
For CPAs, the stakes around AI adoption go beyond productivity. The same tools that accelerate research, document review, and client communication also create new risk surfaces: confidential client data submitted to third-party platforms, AI-generated work product presented without adequate human review, and agentic systems executing multi-step tasks with limited oversight. Understanding where the professional standards apply, and what competent use actually looks like, is the starting point.
AI Tools for Accountants: The Use Cases Seeing the Most Adoption
Across public accounting and corporate finance, AI adoption is clustering around a consistent set of tasks. Tax research is one of the highest-adoption areas: large language models can query tax code, regulations, and revenue rulings quickly and synthesize results in plain language. Document summarization, engagement letter drafting, and meeting notes are also common entry points. In audit and assurance, AI tools are being used for data analytics, anomaly detection, and preliminary workpaper documentation.
The AICPA has developed an AI-enabled research tool called Josi that provides access to the AICPA’s professional standards library, allowing CPAs to query more than 40,000 accounting and auditing materials through a generative AI interface. That development signals the direction of AI integration within the profession’s own infrastructure, not just in practice management software.
On the corporate finance side, AI is seeing adoption in financial planning and analysis, variance analysis, and forecasting. Controllers and finance teams are using AI to accelerate close processes and generate first-draft management reporting commentary.
What Is the Difference Between Generative and Agentic AI?
Generative AI systems produce outputs in response to prompts: text, summaries, analysis, code. The practitioner provides input and reviews what the system returns. The human remains in control of each step.
Agentic AI systems operate differently. An AI agent can be given a goal and carry out a series of tasks to achieve it autonomously: gathering data from external sources, processing files, executing workflows, and generating outputs without requiring human input at each intermediate step. As described in reporting from the Journal of Accountancy (February 2026), agentic AI can perform tasks like real-time reconciliation across general ledger and subledger data, continuous monitoring for anomalies, and multi-step compliance workflows, all with limited manual intervention.
That autonomy is what changes the risk profile.
With generative AI, the practitioner reviews each output before acting on it. With agentic AI, multiple actions may be completed before a human sees the results. For accounting and audit work, where professional judgment and skepticism are required at each significant decision point, the human-in-the-loop principle becomes a governance requirement, not just a best practice.
The Confidential Client Information Problem
Before any AI tool is used on client work, the AICPA’s Confidential Client Information Rule (Section 1.700.001 of the AICPA Code of Professional Conduct) applies. The rule requires consent before a CPA discloses client confidential information. Submitting client data to a third-party AI platform, whether a general-purpose large language model or a specialized accounting tool, is a form of disclosure.
The practical question is what a firm’s AI tool vendor does with the data submitted to it. Does the data remain in the firm’s environment or is it transmitted to external servers? Is it used to train the model? Is it retained? General-purpose consumer AI products frequently use submitted data for model improvement unless users actively opt out or use enterprise-tier products with different data handling terms.
Not reading the data handling terms is not a defense.
Firms using AI tools in client-facing work are building a disclosure and consent process into their engagement terms, along with evaluating whether the AI vendor’s data practices are consistent with client confidentiality obligations. The AICPA published non-authoritative guidelines for the use of AI in forensic and valuation services engagements in 2025, which reference the Code of Professional Conduct’s Compliance With Standards Rule as the applicable framework. The underlying principle, due care in tool selection and documentation of AI use in engagement files, applies across practice areas.
What Does Competent Use Look Like Under Professional Standards?
The AICPA Code’s Due Care principle requires members to perform professional services with competence. Competent use of AI doesn’t mean avoiding it. It means understanding what a tool does, evaluating its outputs critically, and maintaining professional responsibility for the work product.
Three principles govern that in practice. First, AI-generated work product requires human review before it’s relied upon or delivered to a client. A summary generated by an AI tool may contain hallucinated citations, misapplied rules, or plausible-sounding but incorrect conclusions. The practitioner’s signature on any work product is an attestation that it meets professional standards, regardless of how it was produced.
Second, documentation of AI use is becoming standard practice. Recording which tools were used, what data was submitted, and how outputs were reviewed creates an audit trail that supports both quality management and professional liability defense. The AICPA’s forensic and valuation AI guidelines specifically reference documenting AI prompts and outputs as part of engagement files.
Third, firms evaluating AI tools for client-facing work are applying structured evaluation criteria: data privacy and security, accuracy and reliability in their specific domain, integration with existing workflows, and alignment with applicable professional standards.
Questions to Ask Before Adopting Any AI Tool
The evaluation framework firms use for AI tools for accountants starts with a few threshold questions. Where does client data go when it’s submitted to this tool, and under what terms? Can the vendor demonstrate that its outputs in this specific domain are accurate and reliable? What is the firm’s process for reviewing AI-generated work product before it reaches a client? How is AI use documented in engagement files?
Beyond the tool itself, firms are asking whether their engagement letters address AI use, whether client consent has been obtained where the tool involves data disclosure, and whether the individuals using the tool have sufficient understanding of its limitations to apply appropriate skepticism to its outputs.
The AICPA’s position, reflected across its published guidance and professional development initiatives, frames AI adoption as a competency question. Using AI without understanding its limitations isn’t more efficient. It’s a liability.
The AI-augmented Professional: Practical Tools, Real Risks, and Immediate Use Cases (AUG2) from Surgent CPE covers the full spectrum of AI adoption for accounting professionals, including generative vs. agentic AI, governance frameworks, cybersecurity risks, professional liability, and immediate use cases.
Sources: AICPA Code of Professional Conduct, Section 1.700.001 (Confidential Client Information Rule), pub.aicpa.org; AICPA, Guidelines for Responsible Use of Artificial Intelligence (AI) in Forensic and Valuation Services Engagements (2025), aicpa-cima.com; Journal of Accountancy, How AI Is Transforming the Audit (February 2026), journalofaccountancy.com.