Skip to main content
Agents.mu

Decider / 7 questions

Should you automate this task?

Not everything should be an agent. Pick ONE task you are actually thinking about, answer seven questions honestly, and get a verdict: keep it human, let an agent assist, or test a carefully scoped pilot. The reasoning shows as you answer.

  1. 1.Could you write down the rules for doing this task correctly?
  2. 2.How often does the task repeat in a typical week?
  3. 3.Are the inputs digital (email, files, forms, systems)?
  4. 4.If the task were done wrong once, what would it cost?
  5. 5.How often does the way this task is done change?
  6. 6.Could software be given access to the data it needs, safely?
  7. 7.How much do people dislike doing this task?

    Not scientific, but the best first automations are the ones nobody will miss.

0 of 7 answered

Method note: this is an original educational scoring model, not consulting, legal, security or investment advice. It cannot assess your systems or controls. Last reviewed: 2026-07-18.

Frameworks behind the questions

  1. AI Risk Management Framework National Institute of Standards and Technology, 2023A broader framework for governing, mapping, measuring and managing AI risk.
  2. OWASP Top 10 for Agentic Applications OWASP Foundation, 2026Security risks remain relevant even when a task scores as a strong candidate.
  3. Data Protection Act 2017 Data Protection Office, Mauritius, 2017Personal data use requires a separate legal and operational assessment.

Last reviewed: 2026-07-18

Verdict suggests assistance or a pilot?

Sketch the pipeline on the workflow builder to see what the agent would actually do, step by step, and where the human approval gate belongs.

Build the workflowDecode the jargon first