Eric Tribe

Food for thought

The Operating Rhythm, AI-Native

The same rhythm, on AI-native rails.


The original three parts of this guide were written tool-agnostic on purpose. Pen and paper, docs, calendar, Slack. The operating rhythm runs on whatever your team already uses. In recent years the tools have been shifting under all of us: every block of the day now has an AI version waiting to be installed. The principles still hold. What changes is where the friction lives.

This part is the newest. The other three cover practices refined over decades. AI-native execution is months old; expect it to keep shifting as the tools do.

Replicate: the five principles, AI-native

Every principle below maps to a small AI workflow. Call it a skill, a prompt template, a custom GPT, a Gem, whatever your tool calls “reusable instructions you save and call by name.” The goal isn’t one giant agent that does your job; it’s five small tools you reach for at predictable moments. The examples below are illustrative, not canonical. Start there, edit for your context, write your own once you see the pattern.

1 · Day 1 Answer, as a skill

The principle: write the answer you’d give if forced to deliver today, in one paragraph, before doing the work. The AI move is a small skill that takes your draft and stress-tests it: three alternative hypotheses you didn’t pick, the strongest counter-argument to the one you did. You write the paragraph; the skill does the breaking. Don’t let the AI write your draft, or you skip the part that makes you trusted.

Example: Day 1 Answer stress test
You are stress-testing my Day 1 Answer. Topic: [the work] Day 1 Answer: [my one-paragraph draft] Return: 1. Three alternative hypotheses I didn't pick. 2. The strongest counter-argument to my answer. 3. The single piece of evidence that would force me to change my mind. Be direct. Don't hedge. Don't compliment the draft.

2 · Ghost deck, as a prompt

The principle: draft the final artifact first, titles only, in delivery order. Then plan the work to fill it. The AI move is a one-shot prompt: you supply the project and a one-line success criterion; the model returns eight to twelve slide titles, in order. You edit the result, not a blank page. The slide your audience will fight over is the one you write yourself, by hand, before any other.

Example: Ghost deck
Draft a ghost deck for [project]. Audience: [who decides] Success criterion: [one line: what they must decide or believe by the end] Return: - 8 to 12 slide titles in delivery order. - For each: one line on what evidence or argument fills it. - Flag any slide where you can't yet say what fills it. (Those are gaps in my logic, not yours.)

3 · “So what?” as a check

The principle: an observation isn’t an insight. Ask “so what?” of every sentence until the answer either elevates or you cut. The AI move is a check on any draft before sending: paste, run, get a line-by-line readout. The friction that drops is the editing pass. The friction that stays is deciding which so-what matters most to this reader, this decision. The model gives you a plausible implication for every line; your job is choosing.

Example: So-what check
So-what check. For each sentence below: - If it's an observation without a clear implication for the reader, tell me what the so-what is, or recommend I cut it. - If it's already insight (observation + implication + recommendation), mark it [keep]. - Don't soften. Be ruthless. Audience: [who is reading this] Draft: [paste]

4 · MECE, as a check

The principle: when you break a problem into pieces, the pieces shouldn’t overlap, and together they should cover the space. The AI move is a short check against any structure you’ve drafted, looking for the overlap and the missing fourth. This one is underrated for one reason: you wrote three; you’ve stared at three; you can’t see the fourth. The model hasn’t bonded to your structure yet. Run the check before the structure leaves your machine.

Example: MECE check
MECE check on my structure. Problem I'm solving: [one line] Current breakdown: [the buckets, options, or workstreams] Tell me: 1. Where buckets overlap (not mutually exclusive). 2. What I might be missing (not collectively exhaustive). 3. Whether the level of MECE matters for this audience and decision. (Sometimes 80% is fine. Tell me when.)

5 · BLUF, as a rewrite

The principle: lead with the answer. Then the reasons. Then the supporting detail. The AI move is a rewrite skill: paste any draft and ask for the three-sentence BLUF version. If the rewrite is shorter than yours and still says what you meant, the draft was buried; ship the rewrite. If the rewrite reads weaker, your answer is weaker than you think. You can compress an answer you have; you can’t compress one you’re still avoiding.

Example: BLUF rewrite
BLUF rewrite. Return the three-sentence version: - Sentence 1: the answer. - Sentence 2: the two strongest reasons. - Sentence 3: what's underneath if the reader wants more. If you can't compress to three sentences without losing substance, tell me what's structurally wrong with the draft. Draft: [paste]

Mutate: what’s actually shifting

If “Replicate” answered can the rhythm survive the new tooling, “Mutate” answers what changes in the underlying disciplines when it does. Five shifts, one per principle.

Two common takes both miss. AI changes everything; throw out the playbook. AI changes nothing; the tools are toys. The principles were always about judgment, not typing speed. AI raised the floor on output volume by the same amount it raised the ceiling on judgment. The rhythm is what closes the gap.

The inversion

AI didn’t kill the McKinsey playbook. It made it the only thing separating signal from slop.

Which of the five would you build first, and what’s the smallest version that would still be useful?

About the author

Eric Tribe is a strategic advisor and operator. Works with Pioneer Square Labs. Former CRO at Flashfood, with over a decade at BCG before and McKinsey after.

Book 30 minutes · eric@erictribe.com

An advisor’s field guide.