AI Risk
What Workflows Should Not Be Automated
A grounded look at automation boundaries, human judgment, sensitive data, and when manual review still matters.
Do not automate unclear processes
If the team cannot explain the current process, automation will usually make confusion move faster. Clarify ownership, inputs, outputs, and decision points first.
Be careful with sensitive decisions
AI should not make final decisions about hiring, lending, medical advice, legal conclusions, or other high-impact outcomes without expert oversight and a strong governance process.
Keep humans where trust matters
Apologies, complex customer concerns, relationship-heavy sales conversations, and unusual exceptions often need human judgment. Automation can prepare context, but it should not pretend to be care.
Avoid automating bad data
If source data is inconsistent, duplicated, or stale, automation may increase the cost of cleanup. Fix the data model or add validation before connecting more tools.
Automate support, not accountability
The best systems help people make better decisions and reduce repetitive work. They do not remove responsibility for outcomes.