The latest news out of the tax and AI world points in the same direction for firm leaders: pressure is rising to do more work, with fewer errors, and with better controls. ERC fraud cases show how easily bad filings can move through a process, while new survey data suggests many accounting firms are still using AI only in isolated ways instead of building it into the way work actually gets done.
ERC fraud is a workflow problem, not just a compliance story
The recent ERC fraud cases are a reminder that firms need more than knowledge of the rules. They need intake, documentation, and review processes that make it harder for false or incomplete claims to move forward.
For law and accounting leaders, that is where custom AI can help. The right workflow can flag missing support, route exceptions for human review, and keep a clear record of what was checked before a filing or recommendation is made.
The survey data shows why situational AI is not enough
The 2026 Accountant Technology Survey points to a profession under pressure from too many tools, too much cleanup, and a gap between the work firms want to do and the capacity they have. It also suggests that firms using AI only situationally may fall behind competitors that use it more strategically.
That is the practical case for building AI into repeatable workflows. Instead of asking staff to decide case by case when to use AI, firms can define where automation helps with intake, summarization, document routing, and first-pass checks, then keep humans in control of judgment calls.
What agentic workflows should do inside a firm
For professional-services firms, agentic workflows are most useful when they are narrow, testable, and tied to a real business process. Examples include an intake flow that gathers client facts, a review flow that compares those facts against firm rules, and a follow-up flow that sends exceptions to the right person.
The goal is not to replace professionals. It is to reduce manual cleanup, standardize steps, and make it easier for teams to move from one matter or engagement to the next without losing control over quality.
Where leaders should start
A good first step is to pick one workflow where errors, rework, or delays are already costing time. That might be client intake, ERC-related review, or another recurring process with clear handoffs.
From there, firms can test a custom AI workflow with human checkpoints, clear escalation rules, and simple success criteria. That approach is more useful than a general chatbot because it is built around how the firm actually works.
- Use AI to strengthen intake, review, and escalation steps, not just to generate drafts.
- Treat fraud prevention and quality control as workflow design problems.
- Start with one narrow process and add checkpoints before expanding automation.
Sources watched
- Indictment Charges Drug Addiction Treatment Center CEO with Trying to Resell Millions of Dollars Worth of ERCs (CPA Practice Advisor AI)
- The 2026 Accountant Technology Survey: Turning Data Revelations into a Firm of the Future (CPA Practice Advisor AI)
- Former Tax Preparer Ordered to Repay $1.4M in ERC Tax Fraud Case in Mississippi (CPA Practice Advisor AI)
