A week to approve an hour

The message lands at 9:12. A sales rep wants help chasing invoices that slipped in August. The work is an inbox export, a list of names, and a set of follow-ups. It is an hour if someone sits down. Your calendar already has a Thursday block called “ops sync: follow-up process.” That block is how you used to buy safety.

You do not need the block.

Insurance you do not need

Most of the process in your company was written when the first serious attempt was expensive. An hour to clean an inbox export used to mean a vendor, a spec, and a risk review. Now it can mean one person, a spreadsheet, and a sample before lunch. The first try got cheap. The habit did not.

Hiten Shah (@hnshah), who has spent two decades watching small software companies decide things, named what is left over: “You can now spend a week deciding whether to spend an hour building.” That is the tax. It arrives as three meetings booked so nobody feels reckless.

Those meetings have a cost. They delay cash. They teach the rep that asking is slower than living with the unpaid invoices. They also create a fake kind of care. Everyone talked. Nobody touched the list.

This is not an argument for letting a script mail your customers with no eyes on it. It is an argument against treating an hour of work like a capital project. The old cost table said: first attempt is expensive, so decide in a room. The new one says: first attempt is cheap, so try it, then decide whether it deserves a process.

If the request touches customer data, billing, or an external send, you still approve the send. You do not need a week to approve the sitting-down.

Two paths on one request

Watch what happens to the same ask. One path staffs a project. The other path staffs an hour.

Figure 1

Sales rep

August invoices unpaidOwner

One of these

old habit: book a workshopProject manager

hour first: try todayBookkeeper

The project-manager path feels responsible. It is often just delay with a title. The bookkeeper path is uglier and faster. Someone actually opens the aging report. Someone actually writes twenty notes. You get a pile you can look at.

Hour first is a routing rule, not a personality. When the ask is “can we try,” default to a same-day spike. When the ask is “can we make this the way we work,” that is a different door. Staff, budget, and auto-send live on that door. They do not get to cut the line.

Write the rule where requests already live. Chat. The ticket queue. The top of the ops standup. One line is enough: if the first try is under two hours, we do the hour before we book the room.

Pull leaves a receipt

The risk after a good hour is the mirror image of the one you started with. The sample worked, so you make it permanent on the spot. A “follow-up owner” in the next hiring round. A second inbox tool. Thursday promoted from a block to a standing committee.

Julian Weisser (@julianweisser), who runs a program for people starting companies alone, holds that line at a much larger check size. The Solo Founders Program launched with no investment at all, to find out whether people would show up anyway. They did. Only then did it get capital and a standard check.

You are not running a fund. You are running a shop of eight to eighty people. The order still holds. First see whether the work pulls.

Pull is not a vibe in the Slack thread. Pull is a receipt.

For invoice chase, pull is money in the account, or a written promise with a date. For inbox cleanup, pull is a shorter queue tomorrow, counted, not described. For a support script, pull is tickets that actually closed without a rewrite from you.

If the spike produces no receipt, you just saved yourself a quarter of process. That is a win. Kill it in the same chat you approved it in. Do not schedule a retrospective to mourn an hour.

If the spike produces a receipt, you still do not auto-scale. You decide what to fund. Maybe it stays a Thursday hour for the bookkeeper. Maybe it becomes someone’s job with a clear yes/no on outbound mail. That second decision is allowed to be slow. The first try is not.

This is how you keep judgment without freezing the week. Prove the pull on a sample. Then argue about headcount.

The only slide that matters

Pedigree slides show up in small companies too. They just wear different costumes. The costume is a process workshop. The costume is a vendor logo on a one-pager. The costume is a three-week pilot charter for work that could have been twenty invoices today.

Paul Graham (@paulg), who has funded a few thousand companies at their earliest and worst-looking stage, put the outside version plainly: “No credential will get you more credibility with investors than growth. And if you do it well enough, you won't need investors at all.” Inside a company of twenty people, the credential is the deck about the new way of working. Growth is cash collected, tickets closed, and follow-ups that landed.

When you review the spike, lead with the ugly numbers. How many invoices were touched. How much cash moved. How many replies were wrong enough that a person had to rewrite them. If those numbers are thin, the idea is thin. A better narrative will not save it.

You still choose the tradeoffs. Faster collections can load support. A tighter chase can annoy a good customer. That is owner work. It is not a reason to skip the sample. It is a reason to look at the sample with the support lead in the room for fifteen minutes, not three hours.

None of this means “ship whatever.” It means the proof you bring to the later conversation is operational, not biographical. You did the hour. You have a receipt. Now you can argue like adults.

Count the decision meetings

Nobody has a credible figure for how many hours a company your size burns deciding whether to spend an hour. Yours is the only one that would mean anything.

Measure this:

1. Open this week’s calendar. Count every meeting whose purpose is to decide whether someone may spend less than two hours on follow-up, inbox, tickets, or invoices.
2. Write that count at the top of Friday’s notes. If the number is more than zero, you still bought leftover insurance.
3. Do not dress the tally up. Do not average it across last quarter. One week. One integer.
4. Count the hours of actual spike work that happened.

Two numbers on Friday: decision meetings, hours tried. The first should fall. The second should exist.

Kill the Thursday kickoff

• Pick the request. One already sitting in chat. Invoice chase is the cleanest. Inbox export is fine. A support follow-up script is fine. If it needs a week of alignment, it is the wrong one.
• Name the person who will sit for the hour.
• Name the owner who will look at the sample the same day.
• Name what “pull” means in one sentence. Cash in, queue down, or tickets closed.
• Put a stop on anything that mails a customer or writes to billing until you have seen the sample.
• Delete the Thursday block. The block called “follow-up process” is the work. Put the hour in its place.

You can run this yourself. Open the chat and answer with a time, not a meeting.