Create a do-not-automate list

A guardrail playbook for naming the tasks AI should not own yet, so teams can move faster on safer work without blurring accountability.

Teams hesitate because AI feels too broad, or they move too fast because nobody has named which decisions, actions, and customer promises stay human-owned.

  • People worry AI will touch sensitive decisions but cannot point to a clear boundary.
  • The first automation idea includes pricing, discounts, refunds, promises, clinical or legal guidance, eligibility, safety, finance, complaints, or unusual exceptions.
  • There is no shared list of tasks that are off limits, draft-only, or approved for automation.
  • Identifying high-risk tasks from workflow descriptions and historical examples.
  • Separating support tasks from authority tasks: gather, summarize, draft, classify, recommend, approve, send, update, or decide.
  • Drafting a clear do-not-automate list with allowed AI support and required human owner for each item.
  • Humans define the final off-limits list and decide when a boundary can be revisited.
  • AI should not override human judgment, policy, compliance, care, safety, financial authority, employment decisions, or relationship context.
  • List tasks that involve pricing, legal, clinical, safety, finance, complaints, eligibility, refunds, hiring, staff performance, or unusual exceptions.
  • Mark each task as human only, AI may summarize, AI may draft, AI may recommend for review, or AI may automate under defined rules.
  • Review the list before any workflow goes live and attach it to the workflow owner, not a generic AI policy.
  • The team knows where AI is not allowed to act and where it can safely support humans.
  • Safer use cases move faster because boundaries are explicit.
  • Stakeholders trust the rollout more because high-risk authority remains visibly human-owned.
  • Skipping the off-limits list because it feels negative, then letting fear slow every safe use case.
  • Writing vague boundaries that do not help people make real decisions under time pressure.

DIY works for obvious boundaries. Get help when the off-limits list touches compliance, care, finance, employment, safety, refunds, eligibility, legal exposure, or customer trust.

Get grounded AI ops notes, not generic AI hype.

Join the Chalcas list for practical checklists, workflow ideas, and short notes on where SME owners are actually seeing time or margin back.

Free AI action plans for SME owners, then analysis, design, and implementation support for the workflows that can create measurable cash and time value.