Last week, I spoke with an old friend and colleague who is a COO. She said, “Monica, I feel I am sinking in this sea of AI uncertainty, and I am late to the race. Everyone’s talking about agents and copilots. We’re still arguing about pilot scopes.”
I paused for a couple of minutes and then shared with her the story of the first time I learned to float. The instructor did not start with laps. She said, “Exhale. Trust the water. If you thrash, you sink. If you breathe, you float.” Five minutes later, I was not swimming Olympic lengths. I was simply not drowning. That was the win.
Leaders treat AI like a race when the first skill is floating. You don't need a perfect roadmap to get started. You need one safe lane, a clear breath, and the discipline not to thrash. From there, we practice strokes. Small tasks. Short feedback loops. Visible human judgment. Progress comes from technique, not heroics.
There is a Spanish dicho I love that my Dad always says:
A nadar se aprende nadando.
You learn to swim by swimming.
We have to start small. Breathe. Then add a stroke or two. That is enough to get started in your AI journey.
Your team is learning to float. These signals indicate the lane lines, allowing you to swim with confidence.
Let’s clear two common blockers so you can move forward.
Myth: The model matters more than management.
Reframe: Models don’t manage people. Managers do. Redesign roles, set clear rules, and measure the work.
Myth: We should wait until the rules are final.
Reframe: Waiting increases risk. Start with one low-risk use. Add a one-sentence disclosure. Keep a short log of human judgment. The basics are already set.
This week, try the following framework and Prompts to Steal to help you take action. These will help you move from floating to a few clean strokes.
Here is an original framework I created that you might find helpful.
S - Scope the job
Pick one repeatable task. Write success in one sentence. Example: “Send a clear client follow-up within 10 minutes that answers the question and proposes next steps.”
W - Write the guardrails
Make a small box: Allowed, Not Allowed, Must Disclose. Add a line for “W’ Add a line for “What the human decided.”
I - Instrument the test
Choose two metrics. Time saved and rework rate are a strong start. Name who signs off.
M - Move, then learn
Run 5 to 10 real cases. Debrief with Keep, Change, Why. Update prompts and the policy note.
ProTip: When work is high-stakes, pair your review with the SKYE framework I shared in last week’s newsletter.
Show • Keep/Change/Why • Your Defense • Elevate.
Steal these prompts and use them in ChatGPT, Copilot, Gemini, Claude, or your company-approved platform. Doctor them up with your context.
1) Policy in a Box
“Act as an AI adoption coach. I will describe one workflow. Return an ‘Allowed | Not Allowed | Must Disclose’ box, one sentence disclosure for users to complete, three acceptable prompts, three not acceptable prompts with reasons, and a five-minute oral check outline for managers.”
2) Fast Metrics
“You are an ops analyst. Design a simple scorecard for this workflow with two outcome metrics and one risk metric. Include how we will collect each metric and a weekly review ritual in five bullets.”
3) Keep, Change, Why Debrief
“Summarize this draft with a Keep, Change, Why table. End with three edits that elevate voice, clarity, or originality.”
4) Agent Guardrails
“You are a compliance partner. For the task below, list sensitive data to avoid, red flag outputs to reject, and a short escalation path that non-lawyers can follow.”
Float first.
Choose one lane.
Technique beats thrashing.
Small strokes, steady breath, shared wins.
Remember...you learn to swim by swimming.
¡Hasta la próxima, un abrazo fuerte!
(Until next week, a big hug!)
