What AI Can and Cannot Do in a Forensic Engagement

Put artificial intelligence (AI) to work in forensic engagements without losing control of the evidence.

AI tools can:

• Classify documents
• Extract invoice fields
• Flag possible duplicate payments
• Draft summaries

They can also produce confident, polished conclusions that the records do not actually support.

This webcast follows one fictional vendor-payment case from the first AI-assisted screen to a defensible, carefully limited finding.

Presented by a forensic accountant and an attorney, this webcast gives practitioners a practical way to use AI while keeping conclusions within what the evidence supports.

• Choose AI tasks you can check: Tammy Berry, CPA, CFF, CFE, CGMA, will show how to scope bounded, reviewable tasks, control and reconcile the source population, preserve original records, and test a small sample before scaling up.

• Verify the output and document the work: Starting from a deliberately incorrect AI-style output, see how an unsupported $30,000 "fraud" claim becomes a supported, dated finding through independent reconciliation of credits, settlement dates, and refunds, and how to record the AI-assisted step in working papers so another professional can follow it.

• Address the legal questions early: Robert Akouri covers confidentiality and data handling, attorney-client privilege and work product, expert discovery, preservation, admissibility of expert opinions, and disclosure, using federal rules, ABA Formal Opinion 512, and AICPA standards as reference points.

• Explain and defend your conclusion: An illustrative examination tests how the accounting judgments hold up under the questions opposing counsel, a reviewer, or a client might ask, followed by a live Q&A.

Key Topics

• Scoping reviewable AI tasks like document extraction, candidate matching, and summaries
• Verifying AI output like reconciliation, credits, settlement dates, and refunds
• Documenting AI use in work papers and reports
• Confidentiality, privilege, work product, and preservation
• Expert discovery, admissibility, and defending conclusions

Learning Outcomes

• Identify forensic tasks suited to AI assistance that produce output you can check against independent evidence
• Recognize common errors in AI-generated forensic analysis, such as unsupported conclusions, omitted credits, and mixed measurement dates
• Indicate the documentation needed to support an AI-assisted step in forensic working papers
• Recognize issues — in confidentiality, privilege, work product, preservation, and disclosure — raised by using AI in litigation matters
• Distinguish supported findings from conclusions the evidence does not support, such as intent, final loss, or expense effect

Who Will Benefit

Forensic accountants, litigation support professionals, CFF and ABV credential holders, FVS section members, and attorneys who work with financial experts.