Framework

Not Every Workflow Needs an AI Agent

Most workflows don't need an AI agent. Some need a better prompt. Some need plain automation. Some need cleaner inputs. And a smaller number are genuinely agent-ready. This diagnostic helps separate them.

The seven signals

The diagnostic evaluates seven dimensions of agent readiness.

  1. 1

    Repeatability

    Does this workflow happen on a regular cadence?

  2. 2

    Information load

    Does it require gathering, reading, or synthesizing multiple sources?

  3. 3

    Judgment required

    Does it involve interpretation, comparison, or prioritization?

  4. 4

    Pain level

    Is the current process slow, tedious, or error-prone?

  5. 5

    Value potential

    Would faster or better output meaningfully help the business?

  6. 6

    Input accessibility

    Are the required data sources and documents actually available?

  7. 7

    Human review readiness

    Can a person review the AI's output before it's acted on?

Risk as an override

Risk doesn't add to the score — it overrides it. A high-risk workflow with no human review is flagged as advisory-only regardless of how well it scores on other dimensions. This prevents the diagnostic from recommending autonomous AI in situations where a bad output could cause real harm.

Precedence, not just points

The diagnostic doesn't simply add up a score. It applies rules in priority order. A missing workflow definition stops everything. A safety concern overrides a high score. A high-repeatability, low-judgment workflow gets routed to automation even if the total score is decent. This makes the recommendation trustworthy — not just mathematical.

Rules are evaluated top to bottom. First match wins.

1

Missing Inputs

Workflow name, trigger, or output is blank

2

Safety Override

High risk + no human review

3

High Risk Downgrade

High risk, even with human review

4

Strong Agent Candidate

Score ≥ 28 with human review ready

5

Needs Workflow Cleanup

Score ≥ 22 but inputs inaccessible

6

Simple Automation

High repeatability, low judgment and info

7

AI Assistant First

Score 16–27, no other overrides

8

Not Agent-Ready

Score below 16

The eight possible outcomes

  • Strong Agent Candidate

    High score, human review ready, manageable risk.

  • Good Candidate, Needs Guardrails

    High score but high stakes. Build with safety boundaries.

  • Good Candidate, Needs Workflow Cleanup

    Strong potential but inputs aren't accessible enough yet.

  • Use AI Assistant First

    Real value, but start with a prompt or assistant before building an agent.

  • Use Simple Automation Instead

    Highly repeatable, low judgment. Automation beats AI here.

  • Advisory Only

    High risk with no human review. AI can inform, not act.

  • Not Agent-Ready

    Low scores across dimensions. Improve the process first.

  • Define Your Workflow First

    Missing basic definition. Clarify before evaluating.

Ready to diagnose a workflow?

Start Free Diagnostic