A Chick-fil-A Operator Uses AI to Prepare for Staff Coaching

Jon Harmeling described using AI to prepare for employee coaching in an interview published October 26. Marketing managers can borrow the approach: use the software to organize preparation, then take responsibility for the conversation and what happens afterward.

The Chick-fil-A operator in Miami Shores, Florida, uses a database of employees’ goals outside the restaurant to prepare support for them. Harmeling also proposes rehearsing difficult conversations with AI, using it to challenge an initial interpretation before speaking to the person involved. In the 42:58 transcript block, Harmeling says:

“do not let AI replace the human moments.”

Harmeling wants AI preparation to lead to time with employees. A promised conversation still needs a place in the manager’s day.

Harmeling’s official biography identifies them as an Army veteran and former West Point systems engineering instructor. It also describes their current practice: AI helps prepare a question or keep a commitment in view, and the leader still has to listen and follow through. The account concerns this operator’s restaurant. It doesn’t establish a Chick-fil-A policy or a company-wide rollout.

The same employee, a different question

The practice predates this week’s interview. In an October 5 Restaurant Technology Guys conversation, Harmeling described asking Claude to find an overlooked contribution in the team’s Slack activity. An employee who had appeared in an overtime warning surfaced again because they had volunteered to cover other people’s shifts.

Both questions concerned the same work. One helped manage labor costs. The other gave the operator a reason to acknowledge the employee’s contribution. The transcript records Harmeling’s account; it doesn’t independently establish how the employee felt afterward.

That earlier conversation also covered staff biographies and a goal-tracking program. Harmeling described a development structure that runs through the restaurant’s directors and shift leaders, rather than making every employee depend on the operator for individual attention. These are existing experiments discussed again in a new interview, rather than a system launched on October 26.

For a marketing team, the useful question is which work the manager routinely fails to notice. A campaign report can record the launch and omit the colleague who helped a new teammate finish it. An employee’s stated interest in learning customer research can disappear behind the next production deadline. Those are examples of what a manager should look for when reviewing approved team records, not findings about Harmeling’s staff.

An AI suggestion still needs checking. The person who appears least often in a team channel isn’t necessarily being overlooked, and a frequent poster isn’t necessarily contributing the most. The manager needs to understand the work before turning a summary into praise or performance feedback.

A human still makes the schedule

Harmeling’s scheduling example puts a useful limit on the account. In the transcript block beginning at 40:44, they say automated scheduling remains difficult. A person still does most of it; AI checks for problems such as a late closing shift followed by an early opening shift, or a mistaken end time.

A marketing manager can review campaign assignments the same way. A second check can flag a conflict. Deciding whose workload is sustainable requires a conversation with the team. An apparently complete schedule can conceal an unrealistic assumption about somebody’s capacity.

We saw a related separation of preparation from responsibility in Anthropic’s marketing operations account, which kept human validation in its recurring workflows. Here, the decision affects an employee directly. A manager needs to verify the underlying facts and let that person correct missing context before acting on the AI’s interpretation.

Our coverage of Tenex’s employee survey raised another question about managers’ judgments: whether feedback gives a colleague something specific to respond to. That standard also helps when preparing a coaching conversation. A manager should arrive ready to describe the work and hear the employee’s account.

The evidence remains limited. Neither of Harmeling’s interviews supplies a measured retention comparison or an independently audited time-savings result. Their book page schedules Frontline AI Leadership for October 27 and sells a method built around these ideas. Harmeling also markets speaking services. The interview is an operator’s account during a book promotion, not an independent test of the approach.

The book’s definition of recovered capacity does supply a useful accounting rule: subtract the time spent prompting and checking, including corrections and exceptions. A manager who spends an hour maintaining a coaching dashboard hasn’t automatically gained an hour for coaching. The dashboard has to help them carry out a commitment.

For a small marketing team, start with one existing development conversation. Prepare from information the employee has agreed to share, review the AI’s suggestions, and meet with the person. Agree on the next step together and record when the manager will follow up. Judge the workflow by whether that commitment happens, before claiming that the software has improved leadership.