Dana spent the summer planning. Her eight-person firm would take on 60 new individual clients this season. It would roll out an AI-assisted review workflow in January and keep extensions to a minimum. The numbers worked. She announced the plan at the October staff meeting, and people nodded politely.
Three days later, a document showed up in the firm's shared drive with no author. Someone had asked an AI tool to critique the plan, and the answer was blunt. The staff was already stretched thin last April. A new workflow launched mid-season would slow reviewers down before it sped them up. And fewer extensions meant more returns finished by exhausted people in the last two weeks before the deadline.
Dana might be a fictitious character for the sake of this illustration, but the story itself is not. Your team is using AI for more than you think, and one of the ways they are using is to critique you and firm process.
The Concern Nobody Will Sign
In a recent Harvard Business Review article, Amy Edmondson and Jayshree Seth named this “safety by proxy.” It happens when someone relies on AI's reputation for neutrality to raise a concern they don't feel safe raising themselves. The AI takes the heat, and the person stays hidden. Great instances where you might need help addressing a challenging client’s email, not as great when your team isn’t being transparent with you.
Tax firms are especially prone to this. The hierarchy is steep, credentials carry weight, and busy season rewards people who keep their heads down. A second-year staffer won't tell the partner who signs the returns that the plan is unrealistic. But she can type the question into a chatbot in thirty seconds.
Why It's a Tax Season Problem
In a tax firm, the things people don't say affect accuracy, not just morale. Picture a preparer in her 70th hour of the week who doesn't feel safe admitting she's too tired to catch things. That risk lands in your review queue, your E&O exposure, and your clients' returns. People don't complain about fatigue, and fear of looking weak in March keeps them even quieter.
An anonymous AI memo is uncomfortable, but it's information. It tells you what your team believes. It also tells you that saying it out loud feels too risky and both need to be addressed.
What to Do Before January
Stress-test your own plan first. Before you announce capacity targets, new software, or deadline policies, ask AI for the strongest case against them. Where will the workload spike? Which assumptions only work if nobody gets sick? When you raise the hard questions yourself, your team sees that those questions are welcome, or at the very least that you considered the holes in your own plan first.
Talk to people before you decide. AI can list the usual risks. Only your team can tell you which ones worry them most. Hold short one-on-ones in October and November, while there's still time to adjust or host a team “issues” session allowing them as a group to raise concerns that they can then “think tank” through possible solutions for.
Watch how you react. If an anonymous memo shows up, don't try to find out who wrote it. Respond to what it says. A leader who goes looking for the author shows everyone why it was anonymous in the first place.
Close the loop. You don't have to change the plan. But explain your reasoning, and show that raising a concern led to a real conversation. That's what makes the next person willing to speak up directly.
The Window Is Now
Busy season puts the most strain on trust inside a firm and leaves the least time to repair it. By February, nobody has time to rebuild. The firms that get through April intact tend to be the ones that made it safe, back in October, to say “this isn't working.”
If your team is already running its concerns through AI, they're not being sneaky. They're trying to tell you something. What matters is whether you listen before the season starts or only after it ends.