Doing this in the right order

Given a week and no particular constraint, there is an order that produces the most benefit for the least maintenance. It is not the order the tools suggest, which starts with whatever that tool does.

First: stop doing things

Costs nothing to run and nothing to maintain. The report nobody reads, the field nobody uses, the step that exists because of a situation that resolved two years ago, the weekly meeting that could be three lines of text.

Worth doing properly before anything else, because everything you automate here is machinery built around work that should not exist.

Second: write things down

The five saved replies, the checklists, the forms that replace a conversation. An afternoon, no subscription, no failure mode, and for most small businesses the largest single reduction in daily friction available.

Third: the connections you already have

Tools you pay for that can already talk to each other. Switching on a built-in integration costs nothing extra and adds nothing to maintain, and it covers a surprising share of the copying between systems.

Fourth: the one workflow that hurts

One, not six. The highest frequency, lowest judgement, most clearly repetitive job from your week of logging. Build it, run it for a month, and find out what it is actually like to own one of these.

Fifth: notice when things break

Before the fifth workflow rather than after the twentieth. Failure alerts, a nothing-happened alert on anything daily, and a written list of what runs and what it depends on. Dull, and it is what makes the rest trustworthy.

Sixth: the rest, slowly

• One at a time, each justified by the three numbers rather than by enthusiasm.
• Each documented on the one-page list as it goes in.
• Each with a failure alert before it handles anything real.
• Each reviewed at the twice-yearly check, where some of them get switched off.

What this looks like after a year

Not an automated business. A business with perhaps eight to twelve things running, all of which somebody understands, each of which does one job, with the judgement still in human hands and a known route to doing anything by hand when something is down.

That is a considerably better outcome than forty workflows nobody can explain, and it is reachable by one person doing the steps above in order, which is more than can be said for most of what this subject promises.

The first week, concretely

1. 01Monday: log tasks as you do them. Nothing else.
2. 02Tuesday and Wednesday: keep logging. Cancel one recurring thing nobody needs.
3. 03Thursday: save the five replies you write most often, as you write them.
4. 04Friday: read the week's log, rank by frequency times duration, and check whether the top item has a built-in integration you are already paying for.
5. 05The following week: build one thing, or nothing, if the honest answer is that the templates covered it.

A reasonable number of people get to Friday and find that the deletions and the templates removed most of the problem. That is a successful week, not a wasted one.

What changes when somebody joins

The order holds, and the emphasis shifts. Written-down processes stop being a convenience and become how the work gets transferred, which means the one-page documents matter more than the workflows.

It also becomes worth automating things you tolerated, because explaining a fiddly manual step to somebody else is more expensive than doing it yourself was. And every connection that depended on your personal account needs moving to a business one before it breaks in a way nobody but you can fix.

What to resist

Two things, consistently. The first is the demonstration: a tool shown doing something impressive with a process that is not yours, which produces a workflow in search of a problem. The second is the backlog of ideas you accumulated while enthusiastic, which will be built in one weekend and abandoned by the following month. Keep the list, build from it one item at a time, and let most of the items expire unbuilt.