Your Team is Confused about AI - Here's A Simple Fix

In some cases, if you don't use AI, I'm going to fire you. In other cases, if you use AI, I'm going to fire you.

Both true. And if that sounds contradictory, that's a LEADERSHIP problem, not a team problem.

Right now your team hears "use more AI" and "why does this read like a robot wrote it" in the same week.

So they either use it for everything and produce bland work, or avoid it entirely and waste hours on tasks a machine should handle.

Here's the fix: an AI Stoplight.

I invented this.

So when some dweeb on twitter knocks it off - you know who they got it from.

One page. Role-specific.

Green light (low risk): AI away. Internal meeting summaries. Interview questions. Candidate descriptions. Anything low-stakes and not customer-facing. If someone's hand-transcribing meeting notes for two hours, we need to talk.

Yellow light (medium risk): Use it, but you own stamp the output. Customer emails. Content drafts. Anything customer-facing that needs context. This is the Professional Engineer zone: AI drafts, you stamp it. The workflow is simple: you give direction up front, AI generates the middle, you review and season. Paste-in, paste-out, send? That's a fail. You'll have a job until I automate that process.

Red light (DANGER): Humans only. Emotional conversations. Financial decisions. Relationship-critical moments. Can you use AI to prepare? Pull transcripts, spot patterns, brainstorm approaches. But the conversation itself is yours.

Why this matters beyond Sagan.

Without a stoplight, you get two failure modes: your team plays it safe and leaves productivity on the table, or they use AI for everything and the output reads like boneless chicken.

Technically edible. Zero flavor.

AI is great at getting D-level work to B-minus. The leap from B-minus to A? Still requires someone who knows their craft, knows their audience, and gives a shit.

The people who thrive will use AI to move out of the low-level grind so they can double down on the human stuff: relationships, judgment, taste, accountability. The people who just shuttle text in and out of ChatGPT are training their own replacement.

Build your stoplight in an hour. List every recurring task for a specific role. Sort into red, yellow, green based on one question: what happens if the AI gets this wrong? For yellow items, define the review process. Who checks it? What are they looking for? Share it with the team. Make it editable. No committee. No six-month rollout.

Your team is waiting for permission to be smart about this.

Yallah Habibi,

Jon