The automation backlog that actually ships

Every failed automation program we have inherited started the same way: a spreadsheet of fifty ideas, ranked by enthusiasm. Enthusiasm is not a backlog. A backlog is a bet — and bets need odds.

Two axes, one week

Our week-one triage scores every candidate on two axes. Value: hours returned per month once live. Blast radius: what breaks if the automation misbehaves on day one. High value, small blast radius ships first — it funds the harder bets behind it and teaches the team how this particular business fails.

Anything with a large blast radius waits, no matter how shiny. A support autoresponder that can email the wrong customer does not ship in week one, even if the ROI slide is beautiful. Reversible first, irreversible later.

The three rules

• Score value in hours/month, not story points.
• Cap blast radius: read-only or reversible first.
• One owner per automation, named before kickoff.

What ships first

The winner is almost always boring: the report nobody likes compiling, the triage queue everyone dreads, the data sync done by hand every Monday. Boring is good — boring means well-understood, measurable, and easy to verify. The first release should be something the team can check in an afternoon and celebrate by Friday.

Automate the boring thing first. Excitement is a cost; boredom compounds.

Then the flywheel turns

Once the first automation returns real hours, the conversation changes. Stakeholders stop asking whether automation works and start asking what is next. That is when the bigger bets — the copilots, the voice agents, the workflow rewiring — get their funding and their patience. The backlog stops being a wishlist and becomes a pipeline: one shipped proof buys the next experiment.

Start small, measure honestly, and let the boring wins pay for the ambitious ones. That is the whole method — and it has never failed us.