Part Five: The Road and the Letter · Chapter 20 of 21
Chapter 20 The Reader’s Month
In this chapter
Nothing in the book asks the reader to believe it. It asks for one month and one vector, laid out as a calendar with the Monday as day one.
Week one: find the vector. Week two: build beside the old chain. Week three: switch. Week four: the Monday test, and a decision written in one line.
Nothing in this book asks the reader to believe it. It asks for one month and one vector. Here is the month, as a calendar, with the Monday as day one. It is the test the doctrine itself was first run on, at the scale of a single result, and it costs almost nothing.90
Week one: find the vector.
Monday. Pick one result your customer feels, from first word to cash: one order, one hire, one claim, one patient, one case. Not a department’s output; a result someone outside the company receives and would miss. Write it in one line. Then write, on one page, every desk it crosses today between the first word and the cash, in order. Count the handovers. That number is the one you will beat.
Tuesday. Take the last ten instances of that result and, from your own records, write down how many days each took, and where each waited. You will find that most of the days were spent between desks, not at them. Mark the three longest waits. Those are the leaks of chapter 3, and they are yours.
Wednesday. Mark the gates: the three or four points in the chain where an act becomes irreversible, where money leaves, a name goes out or a promise is made. Everything else in the chain is reversible and needs nobody’s approval. Chapter 17 is the rule; most readers find they have been approving twenty things and only four of them were gates.
Thursday. List the skills the chain needs, step by step, and beside each one write whether an agent that exists today can do it: answer the enquiry, draft the document, schedule, check the record, chase the payment. Be honest in both directions. In my experience, for most office results in 2026 the answer is yes for most steps; measure your own.
Friday. Name the pair. One aligner, who will hold the gates for the month, and one tuner, who will stand behind her and change the chain when it is wrong. Choose them by chapter 12: not the two most skilled people on the current chain, but the two who most want the result and will tell you the truth about the number. Tell them the month’s rule: they hold the gates, nobody else touches the result, and the record keeps everything.
Week two: build beside the old chain.
Monday to Wednesday. Build the chain: the agents from Thursday, in order, with the record in the middle so that every act is written, and the four gates with the aligner’s name on them. Use what exists. Nothing in this week should be invented; everything should be assembled. But do not build the chain on the department’s screens. An agent that fills a form designed for a clerk, or waits in an inbox designed for a manager, is the motor belted to the old shaft of chapter 4. Connect it to the record directly, and leave the screens to the old chain.
Thursday and Friday. Run the chain beside the department, on live instances, with nobody acting on what it produces. Let it answer the enquiry the department also answers, draft what the department also drafts, chase what the department also chases, into the record only. Compare. This is the fortnight chapter 13 gave the revenue chain; run it for two days at least, and do not switch until the chain handles the exceptions and recoveries you have seen as well as the routine. Fix what broke.
Week three: switch.
Monday. For this one result, the department stands down. Every new instance goes into the chain. The aligner holds the gates; the tuner reads the record every evening and changes what the agents did wrong; the record keeps every day. Nobody in the company intervenes on this result by any route except the gates, and anyone who tries is told, kindly, that the test is the test.
Tuesday to Friday. Hold. There will be a day when something goes wrong that the chain has no step for. It rises to you, as chapter 17 says it must, with the record of why. Add the step, name the gate, and go on. Do not add a person.
Week four: the Monday test.
Monday. Read the week. For every instance of the result that entered the chain and finished: how many days from first word to cash; for every one that has not finished, its age and where it waits, and keep watching it; against the ten you measured in week one; how many handovers, against the number you counted; how many gates, against the twenty you used to approve. Read the aligner’s week from the record, not from her: gates answered, how fast, what she caught. That is attention, read from the work, as chapter 11 reads it.
Tuesday to Thursday. Let it run, and watch the number move each day as the tuner’s changes take hold. The vector grows by design, and in its second week the design shows.
Friday. Decide, and write the decision in a record of one line. Either the vector stays open and the department never gets this result back, and you have a vector, a pair, a record and a Monday test, and the collapse of chapter 13 has begun at the one place it should: the result nearest the cash. Or the vector did not beat the department, and you write down in one line why, often, though not always, one of chapter 19’s rules for reading the road, and you fix that and run the month again, or extend it, or stop.
That is the whole test. One result, one pair, one month, one line at the end of it. The eight stations of chapter 19 took two thousand years to invent the pieces. The reader has them all, and a Monday.
Questions this chapter answers
How do you start an AI transformation in 30 days?
Pick one result your customer feels, from first word to cash. In week one map every desk it crosses and name a pair; in week two build a chain of agents beside the old one; in week three switch; in week four read the record and decide.
Which result should the first AI pilot carry?
One a customer outside the company receives and would miss, such as one order, one hire, one claim or one case. Not a department’s output.
How do you judge whether the pilot worked?
Read the week from the record: days from first word to cash against the ten instances measured in week one, handovers against the number counted, gates against the approvals used before. Then write the decision in one line.
Notes
- The month described here is the one the author’s own company first ran on a single recruiting result before any department was dissolved; chapter 6 shows the vector it produced. The test does not depend on it. ↩
From the text of Collapse or Collapse by Vinay Pasricha, shared here in full. © 2026 Vinay Pasricha. All rights reserved. Plain text (Markdown).