The ninety days

Why that, and not efficiency

Most automation work is sold on savings — hours back, costs down, productivity up. Those numbers are almost impossible to check afterwards, which is why they get promised so freely.

Ten days away is different. Either the phone stayed quiet or it did not. You will know by the second week whether you got what you paid for, and so will we.

It is also the thing owners actually want and rarely say out loud. Not a dashboard. The ability to be absent.

The ninety days, in order

1. Week 1

   Your signal

   Your customers are already telling you what is wrong with your business. Reviews across every platform, with dates, stored so they can be counted rather than skimmed — which technician generates callbacks, which service produces complaints, which competitor keeps coming up.

   First, because it is the only part that gets more valuable as everything else proceeds — and because it is arriving and evaporating right now.

2. Weeks 2–3

   The path from stranger to paid

   One line. Call arrives, qualify, schedule, dispatch, diagnose, quote, approve, perform, invoice, collect, follow up, ask for the review. Your version of it, with the real names, the real triggers, and the places it actually goes wrong. Not an org chart, not every variant — the common path.

   Hard limit: two weeks. Businesses that spend a month here never leave this stage, and we would rather cut it short than let it become the project.

3. Weeks 3–6

   Writing the systems down

   Each step on that path, written so someone else could do it. We record the work being done rather than sitting you down to describe it — describing a process from memory takes longer and produces the tidy version, the one without the workaround everybody actually uses. We start with the steps you do not touch, so six to eight systems tend to exist by week four having cost you almost nothing.

   Hard limit: thirty minutes per system. If it takes longer it is really two systems, and gets split. Exceptions do not get written down — they escalate to you.

4. Weeks 3–4

   The first automation, early

   One goes live in month one rather than month two: normally the review request after a completed job. Reversible, nobody enjoys doing it by hand, and it feeds the signal work from week one — so the first thing you get back also compounds.

   It comes from a short list we bring, not from whatever is annoying you most that week. An early win is exactly where the expensive mistake would slip in.

5. Week 6

   Deciding what a machine may do

   Before anything else is automated, every system on the path gets one of four answers. Half a day. The two questions that decide most of it are asked of other people rather than of you — whether your customers care that a human did it, and whether the person currently doing the work wants it taken away. Both get asked directly. Neither gets guessed.

   This is the step that gets skipped everywhere else, and the step that makes everything after it defensible.

6. Weeks 7–10

   Three automations

   Three. Not ten. Chosen for how easily they can be undone, not for how impressive they are. The usual opening set is a review request after a completed job, an appointment reminder with confirmation, and a follow-up on quotes that never converted.

   Each one has a written answer to: what happens when it goes wrong, who hears about it, how fast, and how it gets reversed.

7. Weeks 11–13

   You leave

   Ten consecutive working days. Every contact logged.

The four answers

Every system on the path gets exactly one of these, decided the same way every time.

Automate

The machine does it. Nobody signs.

Augment

The machine drafts. A person decides.

Attest

The machine may act. A named person answers for it.

Reserve

A person does this. Better models do not change that.

You can see this run on a worked example on the front page.

What you keep

Four things, yours whether or not you ever work with anyone again.

1. Your customer feedback, captured and queryable instead of scattered.
2. A written map of how your business actually delivers, 10–15 steps.
3. A decision, for every one of those steps, about what a machine may do and who stays answerable — with a name against each row.
4. Three automations running, each with a log of every time a human overrode it.

None of it is throwaway. If you later go further, this is exactly the input the next stage needs.

What makes this different

Every automation agency will tell you they can automate your processes. Most can. The tools are not the hard part any more and have not been for a while.

The hard part is knowing what a machine should be allowed to do — and being able to show your reasoning a year later when something goes wrong.

One example of what that prevents, because it is the most common own-goal in this trade: automating replies to customer reviews. It is visible, it feels like progress, and it is almost always the wrong call — the customer writing that review cares that a person read it, and generating the reply damages the exact signal everything else depends on.

The method catches that before it ships. Enthusiasm does not.

What we need from you

• Access to the people doing the work. The knowledge lives with them, not with you, and it gets captured by watching them work rather than by asking you to describe it.
• Three of your customers, briefly, to answer whether they care that a human handles a given interaction. We ask them. We do not infer it.
• An honest answer about who owns this. At your size it is usually you. If it is nobody, the work stalls in month four.
• Your time, and we will tell you how much before each stage. Every artifact costs you hours before it saves you any. We budget those hours per stage, say the number up front, and track it against what actually happens — because a cost nobody stated is the normal reason this kind of work stalls in month two. Being straight about the figures: they are not calibrated yet. No engagement has run, so the estimates come from the method rather than from experience.

When we will say no

If you do not actually want to be removable.

Some owners are the bottleneck and are content there. That is a legitimate way to run a business, and this engagement will fail inside it, because every step asks you to hand something over.

If you personally perform nearly every step that earns money.

Then there is nothing we can document without you, the work cannot be staged to spare your time, and ninety days is not enough. We will see this on the flow map in week three and say so then. It usually means hiring before systemising.

If nothing honest is coming in.

If your customers are not leaving feedback anywhere, there is no signal to build on, and the only thing worth doing is starting to collect it. That is a smaller engagement and we would rather sell you the right one.

If the work is not repeatable.

This assumes a service you deliver more than once, in more or less the same way.