What automation actually costs

Automation is sold as a saving and is also a purchase, and small businesses routinely end up paying more in subscriptions and attention than the hours were worth. Worth looking at the full bill.

What you are actually paying

1. 01The subscriptions, which scale with volume. A connector platform priced per operation gets more expensive precisely when business is good.
2. 02The tier upgrades. The workflow needs a feature that is on the next plan up, in two different products, and the monthly total doubles for one capability.
3. 03The build time, which was an evening and became four.
4. 04The maintenance, which is permanent.
5. 05The debugging, which happens at the worst moment, because workflows break when volume is high.
6. 06The lock-in. Thirty workflows in one platform is a migration, not a cancellation, and it is priced accordingly.

Subscription sprawl

The pattern is consistent: each tool was justified alone, nobody has looked at the total, and a meaningful fraction of the monthly spend is on things nobody uses.

• List every recurring charge from your bank or card statement rather than from memory. The list is always longer than the remembered one.
• Beside each, write who uses it and for what. Anything you cannot answer is a cancellation.
• Look for overlap. Three tools that each send email, two that each store files, two that each track work.
• Check what you are on the higher tier for, and whether that feature is still used.
• Check annual versus monthly on the ones you are keeping, and whether a tool you already pay for covers the thing you added a tool for.

An honest review, twice a year

1. 01For each workflow: does it still run, is the thing it does still wanted, and does anybody notice when it stops.
2. 02Turn off anything that fails all three. A workflow nobody misses is pure maintenance.
3. 03For each subscription: cancel, downgrade, or keep, with a reason written beside it.
4. 04Add up what remains and compare it with the hours it buys back, at whatever your hour is worth.

Most people doing this the first time remove between a quarter and a third of what they are paying for, and lose nothing. That is not a criticism of automation, it is what happens to any system that is only ever added to.

Lock-in, and what it costs to leave

Worth examining before you depend on something, because the answer is rarely advertised. The question is not whether you can cancel but what you can take with you.

• Can you export your data, in a usable format, without asking somebody?
• Do the workflows themselves export, or would they be rebuilt from scratch elsewhere?
• What happens at the end of a subscription: is the account frozen, or is the data deleted?
• How much of your process exists only as configuration inside this product, undocumented anywhere else?

The last one is the real exposure. Thirty workflows that nobody has written down are thirty pieces of business logic held in somebody else's product, and that is a dependency with no price on it until you try to move.

Keep your records outside the machinery

Whatever you automate, the customer list, the order history and the accounting records should exist somewhere you control and could export this afternoon. Workflows can be rebuilt in a week; records cannot be reconstructed at all.

Which argues for an occasional export, kept somewhere boring. It costs a few minutes a quarter and it converts the worst possible outcome, losing the history of the business, into an inconvenience.

Counting it against your own hour

Put a number on an hour of your time, even a rough one, and the decisions get easier. A workflow that costs a monthly subscription and saves two hours a month is obviously worth it at one rate and obviously not at another. Without the number these arguments are conducted on feelings, and the feelings favour buying things, because the subscription is small and the hours are invisible.