🚩Tactical Memo 083: Owning the Calls AI Can't Make

Welcome to Tactical Memo: Playbooks for high-performing operators. The moves that make a career compound in the AI era. Read by 18,000+ leaders at Amazon, Deloitte, and Google. If someone forwarded this to you, join our free community here.

Two deadlines hit this week. Here's what's up.

1. The AI-Powered Project Management course: 10% off dies July 31.

That's Friday. The fast action discount goes away and the price goes back up. Same course, same 4 weeks, same agents and automations you walk out with. Just more expensive on Saturday.

Grab it before Friday: https://maven.com/justinbateh/aiproj?promoCode=fastaction10

2. Project Management That Works begins August 3. It's almost full.

12 people registered last week alone. We're capping the cohort at 20 and closing enrollment when we hit it. That leaves 8 seats, and at last week's pace they won't survive the weekend.

Claim a seat: https://maven.com/justinbateh/proj

Last year I shipped a recommendation I couldn't defend, and I knew it when I hit send.

The deadline was tight, the AI draft sounded right, and I was tired. The recommendation read like something I would say. So I let it stand.

A week later the client asked one question: why option B over option C? I had nothing. The reasoning was not mine, because I never did the reasoning. I shipped a conclusion I never made.

At the time I called it efficiency. It was hiding. I let a machine make a call so that if the call went wrong, it would not fully be mine.

AI still drafts almost everything I touch, and my output keeps climbing. But that week left me with one rule that now runs all of it: judgment is the job.

Here is how I keep the calls that matter mine.

1/ Write my call before I read the machine's

Before I open an AI draft, I write one sentence: here is what I think we should do. Then I read the output against that sentence. When the two disagree, I sit with the disagreement instead of folding, because that gap is where my actual thinking lives. If I read the draft first, it becomes the anchor and my judgment becomes the edit. The sentence comes first or it never comes at all.

2/ Rebuild the reasoning with the document closed

Before anything ships, I close the file and explain the recommendation out loud from memory: the call, the why, the strongest objection. If I cannot rebuild the chain without the draft, I do not understand it. I just agree with it, and agreement is not judgment. The client question that exposed me took thirty seconds. Every deliverable now survives those thirty seconds before it leaves my desk.

3/ Make the model argue against me

Once I have a position, I give the AI one job: prove me wrong. I ask for the strongest case against my own recommendation, then I go through it line by line and mark every point I cannot answer. Those marks are my homework before the meeting, because a client will find them if I do not. The draft makes me fast. The counter-case makes me safe.

4/ Keep the kill decisions

A model will build a beautiful plan for a project that should not exist. It has no skin, so it never asks whether the work deserves to happen. What we ship, what we kill, and what waits stays my call, made before any drafting starts. On one engagement, that single question ended a six-week workstream in a morning. The machine would have planned it perfectly to the end.

5/ Sign the output like I wrote it

Everything that leaves my desk goes out saying I recommend, never the analysis suggests. If the call blows up in six months, the person answering for it is me, so the person making it has to be me too. That sentence structure is not cosmetic. It forces me to check whether I actually believe the thing before my name rides on it.

6/ Slow down when the draft reads well

The better the output sounds, the more suspicious I get. My instincts were trained in a world where fluent writing meant someone did the thinking, and that world is gone. The deliverable that burned me was the best-written thing I shipped that month. So polish now buys extra scrutiny from me instead of less. Polish is not proof.

The work got cheap. The calls did not. Watch who gets pulled into bigger rooms this year: the operators who can defend their recommendation after it goes sideways, without a document in front of them.

Take one thing you shipped this month, close it, and explain the recommendation from memory. Whatever you cannot rebuild is your work this week.

Free Library of Cheat Sheets👇

I put together a bundle of my top playbooks and cheat sheets for people who want to think sharper, lead better, and navigate the real career rules most people only learn the hard way → Grab it here for free.

Until next time,

Justin Bateh, PhD
Founder and Editor
Tactical Memo