Should this workflow
become an agent?

Turn workflow characteristics into a recommendation for agent fit, autonomy, controls, architecture, and estimated operating leverage. Built to force the product decision before the model decision.

WORKFLOWJUDGMENTCONTROLVALUE

01 / DEFINE THE WORK

Workflow inputs

Workflow name

PRESETS

Weekly volumeHow often does the workflow repeat?150 Minutes per caseHow much human time does each run consume?30 Rule clarityCan good decisions be described consistently?5 / 5 System / data accessCan the product reliably retrieve the context it needs?4 / 5 Action reversibilityCan a bad action be stopped or undone?4 / 5 Failure consequenceWhat happens if the system is wrong?2 / 5

02 / PRODUCT DECISION

Operations run setup

STRONG FIT

82/100

AGENT FIT

RECOMMENDED AUTONOMYSupervised agent

Good structure and system access support an agent, while a human approval boundary preserves control.

CAPACITY RETURNED49 hrs / wk≈ 2,548 hrs / yr

SYSTEM PATTERNAPIs + typed actionsapproval before consequential action

CONTROL POSTUREHuman approvallogs · deterministic validation · rollback

WHY THIS RECOMMENDATION

01Assistretrieve / summarize

02Copilotrecommend actions

03Supervisedact with approval

04Bounded autonomousact inside policy

03 / SYSTEM SHAPE

Fit is only useful if it changes how you build.

For this workflow, the recommendation is a supervised agent: retrieve live context, call bounded domain APIs, validate deterministic rules, then request approval before consequential action.

CONTEXTRetrieval / MCPlive state + policy

TOOLSDomain APIsnarrow capabilities

ACTIONSTyped callsbounded parameters

CONTROLApproval gatedeterministic checks

EVALSBehavior + QAbefore autonomy expands

Should this workflowbecome an agent?

Workflow inputs

Operations run setup

Fit is only useful if it changes how you build.

Should this workflow
become an agent?