THIS MONTH'S FIELD SIGNAL

 The Signal in the Pushback

How autonomous do you want your car to be?

For some people, the answer is: fully. Hands off the wheel, attention elsewhere, trust the system. For others, the idea of not being involved in the decisions the car is making is a dealbreaker. Neither answer is wrong. But you would want to know your answer before you buy the car.

I have been thinking about that question because a version of it keeps surfacing in my conversations with leaders this year. As AI moves deeper into real workflows, organizations are discovering they have strong feelings about how much autonomy AI should have, and where human judgment needs to stay behind the wheel. Those feelings rarely arrive as a clean policy discussion. They arrive as pushback: employees insisting on review steps, questioning outputs, moving slower than the rollout plan assumed.

The pattern I keep seeing is in what leaders do next. The instinct is to treat pushback as an obstacle: a change management problem, a training gap, resistance to be managed down. Sometimes it is. But more often, the pushback is the most useful information the organization has. It tells you where trust is thin. It tells you where the workflow design does not match how the work actually happens. It tells you where people believe judgment still matters, and why.

This is surfacing now because most organizations are past the pilot phase. AI is touching real work, real customers, and real risk, and the question of who goes first when a human and an AI system disagree is no longer hypothetical. Leaders who read pushback as signal get a roadmap: it shows them exactly where to invest narrative, enablement, and guardrail design. Leaders who read it as obstruction get compliance at best, and quiet rebellion at worst.

I go much deeper on this in my new Medium article, the fifth and final piece in my Compass series, which looks at what the fight over city streets a century ago can teach us about deciding who has the right of way today. 

FROM THE FIELD

A Real-World Example

I recently spent time running listening sessions across a client organization in a highly regulated industry, talking with people at every point on the AI enthusiasm spectrum, from early adopters automating everything they can to people who would prefer AI never touched their work.

One signal came through from nearly all of them, regardless of where they sat on that spectrum: the importance of human review of AI output. The enthusiasts said it. The skeptics said it. In a business where a wrong answer can carry regulatory consequences, and in functions like legal where precision is the job itself, it came through with particular force.

Leadership had picked up the same theme from a different angle. Their worry was the opposite failure: that over time, employees would become too trusting of AI outputs, and review would quietly become a rubber stamp.

It would have been easy to read the sheer volume of this feedback as resistance, a workforce telling leadership to pump the brakes. I read it differently: as the organization articulating what it values. So instead of managing the concern down, we are building it in. Human review of AI output is becoming an anchor value, woven into the leadership narrative, manager enablement, and employee training. Something that surfaced as a worry bead is becoming something the organization stands for.

The lesson: the loudest worry in your listening data just might be a cultural cornerstone waiting to be named.

QUESTIONS FROM THE TRAIL

Ask Yourself

1. When employees push back on an AI initiative, where does that information actually go? Who hears it, and has it ever changed a decision?

2. How much autonomy are you comfortable giving AI in decisions that carry your organization’s name? Could your managers state that answer without checking with you first?

3. What are your most reluctant adopters protecting? If you cannot name it, you have not listened closely enough.

TOOLS FROM THE TRAIL

Synthesizing What You Hear

If you are running listening sessions or collecting employee feedback on AI, the volume gets unwieldy fast. I have been using AI to synthesize anonymized session notes into recurring themes: what people are worried about, what they are excited about, what keeps coming up unprompted. It is how patterns like the one above become visible across dozens of conversations.

Two boundaries worth keeping. Anonymize before anything goes into the tool. And do not let the synthesis substitute for being in the room: summaries flatten emotion, and outliers get smoothed away, which is often where the real information lives. Useful for finding the pattern; not a replacement for hearing the person.

ONE SMALL STEP

Let's Chat

If pushback is showing up in your own AI rollout and you are wondering what it is trying to tell you, I would love to hear about it. Just hit reply.

A quick note: You’re receiving this email because we’ve connected at some point about AI, leadership, or communications/marketing, and I thought you might find these field notes useful. I’ll send this about once a month. If it’s not helpful, you can unsubscribe anytime using the link below—no hard feelings.