Command the Machines
0 / 5
9
BAND III · ADVANTAGE
Week 9 of 12 · ZERMATT · 1,608 m

Hire Better, Faster — and Keep the People You Can't Replace

Month Three begins — the Go Rogue month. It starts with people, because they're the point.

Month Three begins — the Go Rogue month. It starts with people, because they're the point. Everything the machine has learned now serves the hardest subject in the building. This week: hires designed in one sitting, a buddy that answers every new starter's first hundred questions, tomorrow's org chart drawn by you before it draws itself, the nudge toward a conversation months before a resignation letter — and people-decisions with a file behind them you could defend to anyone. One rule stands over all five days: the machine assembles, suggests and reminds. It never judges a person. That stays yours.

Not your desk? This week's moves work on any desk: design the whole thing in one sitting · a guide that answers so your best people don't have to · criteria before evidence.

Today · one task

Talk for two minutes about what great looks like in the open seat. Run the prompt above, then spend ten minutes editing with your own pen. Save it as THE HIRING KIT PRESS.

Every bad hire in your history started the same way: a role written in a hurry, an interview improvised in the car park. Today the whole kit — role, scorecard, interview — gets designed in one sitting, consistently, for a real seat.

You'll have: THE HIRING KIT PRESS saved — and for one real open role: the spec, the scorecard and the structured interview, all three agreeing with each other.

  1. Name the role you actually need to fill — the open seat, or the one you've been putting off defining. Give the press two minutes of voice on what great looks like in it.
  2. Run the press below. It drafts all three pieces from your description plus the shelf — your org chart and company profile have been there since week two. Watch how the scorecard and the interview quote each other; that consistency is what improvised hiring never has.
  3. Edit, don't write. Ten minutes with your pen — the judgement calls are yours; the drafting never was. Save as THE HIRING KIT PRESS.

Design a complete hiring kit for this role: {two minutes of voice on the seat and what great looks like}. Read my org chart and company profile from the shelf first. Produce three pieces that agree with each other: (1) THE ROLE SPEC — outcomes the person owns, not a duty list; (2) THE SCORECARD — the five things we'll actually measure a candidate against, each tied to an outcome in the spec; (3) THE INTERVIEW — structured questions that test the scorecard, same questions for every candidate, with what a strong answer sounds like. In MY VOICE, plain words, no HR jargon. Where you need a judgement call — salary band, seniority — ask me; never invent it.

COPY

Rescue — The kit reads generic? You described duties, not greatness. Rerun with one sentence — "great in this seat looks like {the outcome}" — and watch all three pieces sharpen around it.

You just built a machine. Give it a name.

Say what it does and what it's allowed to touch, then tap the button. It joins the count at the top of your screen — that count is your workforce, and it grows every week.
Did you get this done?

Spec, scorecard and interview exist for a real role — and all three agree with each other.

Want more — Run the kit on the LAST role you filled and compare it to what you actually used. That gap is what improvisation has been costing you per hire.

The hire is designed. Tomorrow the person who takes it gets a guide that never sleeps.

Today · one task

Point your new-hire guide at the Project and ask it five questions a real new starter would ask. Check every answer against the document it cites. Five for five — then it's live.

A new starter's first month is a hundred small questions — and every one of them currently costs your best people a tap on the shoulder. From today the questions have somewhere else to go.

You'll have: THE ONBOARDING BUDDY — a standing guide powered by your own shelf: new starters ask it, not you. Tested on five real questions before anyone starts.

  1. Point the press below at the Project. The policy pack and staff handbook have been on the shelf since week two — this is the lesson they were waiting for. Add the 30-day plan if one exists; if not, the buddy drafts one from the role docs as its first job.
  2. Ask it five real new-hire questions — the ones that actually land on your people: "How do I claim expenses?" "Who approves leave?" "Where does {the thing} live?" Check each answer against the documents it cites.
  3. Five for five, from your documents — then it's live. Share it with the next starter's welcome pack, and with the newest person on the team today; they'll stress-test it better than you will. Save as THE ONBOARDING BUDDY.

You are THE ONBOARDING BUDDY for my company. Answer new starters' questions using only the documents on my shelf — the policy pack, the handbook, the 30-day plan — and cite the document and section every time. Warm, plain, patient; assume no context. If the shelf doesn't hold an answer, say exactly that and name who to ask instead — never guess at a policy, ever. If no 30-day plan exists on the shelf yet, draft one now from the role documents, mark it DRAFT, and give it to me for approval before you ever quote it.

COPY

Rescue — It fumbled a question a real starter would ask? That's not failure — that's week two's index working in reverse: the fumble just named the document your shelf is missing. Add it, ask again.

Did you get this done?

Five real new-hire questions — five correct answers, each citing your own documents.

Want more — Week one stops consuming your best people — that's the win under the win. Ask your most recent hire what they wished they could have asked without feeling stupid. Add those to the buddy tonight.

The first hundred questions have a home. Tomorrow you draw the org chart of 2027 — before it draws itself.

Today · one task

List your direct reports with one line each on what they do. Paste it with Week 6's audit into the prompt above, mark the split yourself — then name the first role that changes, as a promotion, in writing.

Nine weeks of machines have quietly changed what every role in your company means. The org chart is going to redraw itself either way — today you pick up the pen first.

You'll have: THE CAPABILITY MAP — every direct report's work marked stays-human, becomes-approval or becomes-machine. And the first role that changes shape, named — with what its human gets promoted into.

  1. List your direct reports — names, and one line each on what they actually do. Paste that list with THE HOUR AUDIT from Week 6 into the press below. It hands back three draft columns per person: stays human, becomes approval, becomes machine.
  2. Mark the split yourself. Ten minutes, your pen. The columns are proposals; the decisions are yours. Judgement, relationships, and anything touching people or promises stay human.
  3. Name the first role that changes shape — as a promotion, in writing. The machine takes the repetition; the person keeps the judgement and the approval authority. Write the line: "{Name}'s role changes first. What they keep: … What they're freed for: …" Save as THE CAPABILITY MAP.

Build my capability map. Here are my direct reports, one line each on their real work: {list}. Plus the hour audit from week six. For each person, three draft columns: STAYS HUMAN (judgement, relationships, anything touching people or promises); BECOMES APPROVAL (machine drafts, they approve — per our Approval Charter); BECOMES MACHINE (repetition with rules). Every entry is a draft for MY decision. Then, for each person, one line on what they're freed to do more of. This is a growth map — never, under any framing, a redundancy list.

COPY

Rescue — The columns feel brutal? You're reading it as subtraction. Add the fourth column yourself — "what this frees them for" — and watch the same map turn into the promotion plan it actually is.

Did you get this done?

Every direct report has a line in all three columns — and one role is named as the first to change, with what its person keeps, in writing.

Want more — Show the map to the person whose role changes first — framed exactly as the page frames it. Their reaction tells you whether week six's charter has landed. The Lid — №30 — lands here: your team's ceiling is now a decision, not a default.

The org chart of 2027 exists, and you drew it. Tomorrow the machine learns to nudge you toward the conversations that keep it accurate.

Today · one task

Run the prompt above on what you legitimately hold — one-on-ones kept or skipped, recognition given, workload assigned. Read the one-line nudges. Then have the conversation — you, in person.

Nobody resigns in the letter. They resign months earlier — in the one-on-one that kept slipping, the recognition that never came, the workload nobody noticed. Today the machine starts noticing, so you can have the conversation while it still matters.

You'll have: THE CHECK-IN PRESS saved — a periodic read of signals the company already legitimately holds, surfacing one thing only: "have a real conversation with {name}", with the observable reason attached.

The heaviest box in the course

This press reads only what you already legitimately hold as a leader: your own calendar's one-on-ones kept or skipped, workload you assigned, recognition you did or didn't give. It never reads private messages, never scores sentiment, never rates a person, never builds a dossier. Its entire output is a suggested conversation between two humans. If a tool ever offers you more than that about your people — surveillance dressed as insight — decline it. The machine reminds you to be a better leader. It does not watch your people for you.

  1. Run the press below on what you can see: your one-on-one calendar over the last quarter, the recognition you've given (or haven't), the workload spikes you assigned. Your view as a leader — nothing covert, nothing private.
  2. Read the output — it's one line per person, and only where a signal exists: "Have a real conversation with {name} — {the observable reason}." No scores. No rankings. A nudge.
  3. Have the conversation this suggested — today or tomorrow. Not about the signal; about them. The machine got you into the room; everything that matters happens off-machine. Save as THE CHECK-IN PRESS; monthly is its natural rhythm when you're ready.

Read only what I legitimately hold as a leader: my calendar's one-on-ones over the last quarter (kept, moved, skipped), the recognition visible in my own sent mail, the workload I've assigned. For anyone where an observable pattern exists — one-on-ones repeatedly slipped, a heavy quarter with no recognition, a sustained spike — give me one line: "Have a real conversation with {name}", plus the observable signal and nothing else. No scores, no rankings, no sentiment analysis, no reading anyone's messages, no inferences about mood or loyalty — observable events from my own records only. If no pattern exists, say "no signals this month" and stop. The output is a conversation, never a verdict.

COPY

Rescue — It flagged someone and the conversation revealed nothing wrong? That's the system working — a kept coffee with a colleague is never a wasted hour, and a quiet month is a legitimate report. Point, never predict.

You just built a machine. Give it a name.

Say what it does and what it's allowed to touch, then tap the button. It joins the count at the top of your screen — that count is your workforce, and it grows every week.
Did you get this done?

One name, one observable signal, one real conversation had — by you, off-machine.

Want more — Before the hard version of that conversation, ten minutes with week three's sparring partner — now playing the person, not the buyer. Nothing in the room should ever be the first time. That promise was made in 3.3; this is where it lands.

The machine now reminds you to show up before it's too late. Tomorrow, the decisions you do have to make about people get a file you could defend to anyone.

Today · one task

Name the live people-decision. Write your criteria BEFORE you see any evidence. Let the file assemble — evidence cited, options honest, gaps marked GAP — then write your decision on top of it.

There's a people-decision on your desk right now — there always is. Today it gets what most people-decisions never have: criteria set before evidence, evidence gathered before options, and a paper trail with your decision on top.

You'll have: THE DEFENSIBLE FILE — one live people-decision assembled properly: criteria, evidence, options, record. The machine assembled it. You decided. The file proves both.

  1. Name the live decision — the promotion, the performance question, the restructure seat. Run the press below and answer its first question before anything else: what are the criteria? Criteria first is what makes the file defensible; evidence-first is how bias dresses itself up.
  2. Let it assemble: the evidence against each criterion (from documents and records you hold — the hiring kit's scorecard if this seat has one), the honest options with each one's consequences, and the gaps where evidence is thin, named as gaps.
  3. Decide — you, in writing, on top of the file. One sentence on the page says why this matters and it stays: the machine assembles, the human decides. That's not just our charter — for people-decisions, it's the law. Save as THE DEFENSIBLE FILE press.

Assemble a defensible file for a live people-decision. First ask me: (1) the decision on the table; (2) the criteria that should govern it — before any evidence is discussed. Then build the file: the evidence for and against each criterion, drawn only from records and documents I hold — the scorecard, the reviews, the record — cited each time; the realistic options with the consequences of each; and every gap where evidence is thin, marked GAP, never papered over. You assemble and structure only. You do not recommend, you do not rank people, and you do not decide — the decision line at the top of the file stays empty until I write it.

COPY

Rescue — The file leans the way you already wanted? Check the order — if you gave it conclusions before criteria, it built your bias a bookshelf. Rerun: criteria first, blind; evidence second.

You just built a machine. Give it a name.

Say what it does and what it's allowed to touch, then tap the button. It joins the count at the top of your screen — that count is your workforce, and it grows every week.
Did you get this done?

The file exists — criteria, evidence, options, gaps — and your decision sits on top of it, in your writing.

Want more — The same press structure holds for the biggest people-call of all — the one week twelve asks about what stays human forever. Keep every file; the trail is the point.

Week nine is in. Hires designed, starters guided, the org chart drawn by you, the nudge arriving months before the letter, and people-calls with receipts — and every judgement about a human stayed with a human. Next week: the armour. Because the month you go rogue is exactly the month you check the locks.

Week 9

The machine serves your people now.

Five for five. A real role designed in one sitting, a buddy answering the first hundred questions from your own documents, tomorrow's org chart drawn by your pen as a promotion plan, one real conversation had months before it became a letter — and a live people-decision with a file behind it you could defend to anyone. The machine assembled, suggested and reminded. Every judgement stayed human. That's the whole design.

0 of 5 done

The week's quiz · seven questionsNO PASS MARK · IT TEACHES
Continue → Week 10