How to pick the one job to fix first

The question that arrives in nearly every first conversation: where do we start?

The answer is one job. Not a category, not a department, not a platform — one recurring piece of work, fixed properly, end to end, before anything else gets touched. Businesses that do this get a second and third change almost for free, because the method transfers. Businesses that start everywhere at once get a set of half-configured setups nobody trusts.

Choosing which job is the whole decision, and the instinct most people reach for is wrong.

The trap: picking the most annoying job

The job that irritates you most is usually the one with the most judgement in it. That's why it's irritating — it can't be done on autopilot, it needs someone senior, and it resists being handed over.

Which makes it exactly the wrong first candidate. You'll spend three sessions discovering that the hard part was never the typing, conclude the tools are overrated, and stop.

Five tests that actually predict success

Run your candidates through these. A job that passes four is a good first job.

1. Frequency. Weekly or better. A quarterly job doesn't come round often enough for the setup to be refined while anyone still remembers how it works.
2. Stable shape. The same structure each time, with different facts. Tenders, monthly reports, standard explanatory letters, scopes of works, induction material. If every instance is structurally different, there's nothing to save.
3. Mostly writing or restructuring. Not deciding. The judgement can stay exactly where it is — the win is in the assembly and the typing.
4. No client information required. Not because it's impossible otherwise, but because the first job should produce a result rather than an argument about de-identification. There's plenty of work that qualifies; more on why in the client-data objection.
5. Currently done after hours. The clearest signal of all. Work that happens at night is work nobody has resourced, which means an hour saved is an hour returned to the person who owns the business.

The test that settles a tie

If two candidates both pass, ask this: if this job were half the effort, would anything downstream change?

Sometimes the answer is no. The report still gets sent on the same day, the client still renews or doesn't, and you've bought yourself a quieter Sunday — which is worth having, but it's a personal win rather than a business one.

Sometimes the answer is yes, and those are the jobs worth doing first:

• The quote that currently goes out tomorrow goes out today, and same-day quoting wins work in most trades.
• The monthly report gets specific instead of thin, and thin reports are how contracts get lost. That argument is set out in the report that keeps the contract.
• The second follow-up that nobody has time for actually gets sent.
• The tender you'd normally decline for lack of time becomes one you submit.

That's a different category of outcome from saving an hour, and it's the one that justifies doing this properly rather than dabbling.

What "fixed properly" means

Not a good afternoon with a chat window. A job is fixed when:

• The setup holds your context, your format and a good past example, so the result doesn't depend on who's typing. That's the reusable setups module.
• Someone other than the person who built it can run it.
• The checking step is written down, with a name against it.
• There's a note covering what to change when the template changes — the step that stops the whole thing being stranded with one person.

That's the shape of a workflow build, and it's deliberately narrow: one job, built with you, tested on the next live instance rather than on a practice run.

Then leave it alone for a month

The temptation after the first success is to do five more immediately. Resist it for a few weeks. You want to know whether the first one survives contact with a busy week, a changed template and a staff member who wasn't in the room — because whatever breaks it will break the next five the same way.

The businesses that end up with ten jobs running differently almost always got there one at a time.