Insights/Decision science

Automation debt is real. Here's how to pay it down

Priya RaghunathanPrincipal, Decision Science
March 18, 2026 · 12 min read
Copy linkXLinkedIn

On this page

The problem with undocumented callsWhat a decision record actually containsWhere AI genuinely helpsStarting this quarter

Want the decision-record template we use with clients?

Get the template →

Automation debt compounds the same way technical debt does, but it is harder to see: the script still runs, so nothing looks broken.

Inventory before you refactor

Start with a list of every automated job, its owner, and what happens when it fails silently. Most organisations cannot complete this list, which is itself the finding.

The script still runs, so nothing looks broken.

Prioritise by blast radius

Fix the jobs whose silent failure would take longest to detect, not the ones that break loudest. Loud failures are already handled.

Budget for it

Reserve a fixed share of automation capacity for maintenance — a fifth is a reasonable starting point — and protect it from roadmap pressure.