---
title: "The outside help will not show up in time"
subtitle: "Rank five live jobs this week and put someone you already pay on each one."
principle: "Staff the five"
date: "2026-09-23T14:12:36.162Z"
author: "Mikko Atienza"
canonical: "https://ciel.codes/showcase/blog/the-outside-help-will-not-show-up-in-time"
edition: "2026-09-23"
source: "https://ciel.codes/showcase/pulse/2026-09-23"
abstract: "Demand is about to hit every shop at once. The people who get paid to wire this into a real week will be booked. If your plan assumes a specialist shows up, the plan is already late. You pick five"
cite: "Atienza, M. (2026). Staff the five. Pulse."
citations:
  - handle: "@gabriel1"
    url: "https://x.com/gabriel1/status/2102527249643888830"
---

Demand is about to hit every shop at once. The people who get paid to wire this into a real week will be booked. If your plan assumes a specialist shows up, the plan is already late. You pick five live jobs this week and you put a name you already pay on each one. Everything else waits.

## Your request pile is already too long

Someone on your team will ask, this week, to add it to intake email. Someone else will want quote follow-up. Status updates. Onboarding. Inventory. Each ask is real. Each ask has a person who is tired of doing it by hand. Say yes to all of them and nothing ships. You get twelve half-started lanes and the same Friday you had last month.

Gabriel Petersson (@gabriel1), who left OpenAI research after watching the models get usable, put the squeeze in one line. Companies will want this for everything all at once, because it will soon work. There will not be enough specialists and consultants to add it fast enough.

> every enterprise will soon want ai for everything all at once because it will very soon work, and there wont be enough FDEs and consultants to add ai fast enough
>
> Gabriel Petersson (@gabriel1)

He means the hired people who sit with a team and wire the job. You do not have one on payroll. A shop of twenty was not going to get one this quarter anyway. After this wave, the consultant calendar gets worse, not better. Waiting is not a strategy. It is a place in line.

So you stop treating outside help as the install crew. The install crew is whoever already answers the inbox, prices the quote, or talks to the customer. Those people can run a trial this week on work that already exists. They cannot run twelve trials. They can run five if you name them.

> **Aside.** Nobody has a credible figure for how many of those specialists exist. Read the shortage as a reason not to wait, not as a hiring forecast.

## Start from the jobs that already have a queue

Your intake pile, your open quotes, and your status thread already have owners, even if the names are informal. Those jobs eat hours and they touch cash, which is why they keep showing up in the ask list. Add two more only if they already have a queue you can point at. A job with no live pile is a wish. It does not get a lane this week.

You rank by cash touched or hours eaten in a normal week, not by who asked loudest. The person who owns the week still makes that call. A model can sort a list. It cannot tell you which customer you cannot slip.

- [ ] Write every live job people asked to automate this month
- [ ] Rank them by cash touched or hours eaten in a normal week
- [ ] Keep five. Date the rest. Do not keep a maybe pile.
- [ ] Put one name already on payroll on each of the five
- [ ] Tell the rest of the company those five are the only open lanes until Friday

The parked list needs a date, not a shrug. "After Friday" is a date. "When we have time" is how twelve jobs crawl back onto the board.

Petersson is describing a rush. Your job is to refuse the rush a shape. Five named lanes is the shape. The rest of the company can keep asking. You can keep a dated park list. You cannot keep pretending every ask is in flight.

## The owner ranks, the lane owner tries, the closer sends

You rank on Monday. You do it in the same place you already run the week, the stand-up board, the ops chat, the sheet on the wall. You do not make a new system for the ranking.

Each lane owner takes one live item, not a plan. For intake, that is five real incoming messages, drafted with whatever you already use, sent only after a human reads them. For quotes, that is three open quotes. Follow-up text can be drafted. Price and promise date stay with the person who already prices. For status, the sentence that leaves the building still comes from the closer, the person the customer already trusts.

If the trial cannot touch a live email, a live ticket, or a live quote, it is not a trial. It is a meeting. Cancel the meeting and pick a smaller live item.

The closer still holds the customer promise. That is not a detail you bolt on at the end. If a draft would slip a date, miss a price, or sound unlike the shop, it does not go out. The lane owner can produce the draft. The closer still says yes. You still decide which five jobs even get a draft.

By Wednesday the five should each have one live item in motion. By Thursday you should know which lane is fake. A fake lane has a name and no work object. Kill it the same day. Do not carry a ghost into Friday.

```mermaid
flowchart TD
    A["Owner ranks five live jobs"] --> B{"Owner: cash or hours this week?"}
    B -->|No| C["Owner parks the request"]
    B -->|Yes| D["Lane owner runs the trial"]
    D --> E["Closer holds the customer promise"]
    E --> F{"Lane owner: did live work move?"}
    F -->|No| G["Owner kills the lane"]
    F -->|Yes| H["Lane owner keeps the job"]
    style A fill:#1e1b4b,color:#ffffff
    style B fill:#4c1d95,color:#ffffff
    style C fill:#064e3b,color:#ffffff
    style D fill:#1e1b4b,color:#ffffff
    style E fill:#064e3b,color:#ffffff
    style F fill:#4c1d95,color:#ffffff
    style G fill:#4c1d95,color:#ffffff
    style H fill:#064e3b,color:#ffffff
```

The handoff is short on purpose. Nobody is waiting for a consultant to join the thread. If a lane needs a person you do not employ, it was never a lane for this week.

## Friday, count the names

By Friday you should be able to point at five names and at live work those names actually touched. Empty cells until you count. Do not guess.

| Measure | Count |
|---|---|
| Ranked jobs with no named owner |  |
| Ranked jobs that touched live work this week |  |

If a lane still has no name, it is not a lane. If it has a name and no live item, it was theater. Drop both from next week's five. Promote one job from the dated park list only if you have a free owner. Do not grow the board to make the asks feel heard.

The first row is the one that tells you whether you actually staffed the week. The second row tells you whether staffing did anything. A full set of names and a zero on live work means you assigned spectators. Change the owner or kill the lane.

## Name five owners before the next stand-up

- [ ] Put the five jobs and five names where the week actually runs
- [ ] Each owner starts one live item before Wednesday
- [ ] You keep the customer promises, or the closer does
- [ ] Friday, fill the two counts. Drop any lane that still has no name or no live item.

You can staff the five without a specialist in the building.
