What makes a workflow ready for an agent (the readiness checklist)

Most processes fail an agent handoff for the same few reasons. Here is the checklist to run before you pick a first workflow.

Most processes fail an agent handoff for the same few reasons

Today

A ranked wish list, picked on instinct

  • You have a list of processes you would like to hand off, and no reliable way to rank them.
  • The obvious candidates are usually the loudest ones: the month-end scramble, the inbox nobody owns, the report someone rebuilds every Friday.
  • Loud and ready are different properties.
  • A process can consume half your week and still resist a handoff, because the rules live in one person's head, the inputs arrive in six shapes, and nobody has written down what a correct result looks like.
  • Picking wrong costs you a quarter and a good deal of internal credibility.
With agents

Six checks, run before you pick a first workflow

  • You choose on evidence.
  • Preparation stops being invisible.
  • The first handoff is survivable.

The readiness checklist

Readiness is a property of the process, and you can inspect it before you commit to anything. Run the workflow you are considering against these six checks.

A workflow that passes all six is a candidate. One that fails a check is not disqualified; the failing check names the preparation work.

01

Stable inputs

It reads from named places in known formats. If the input is "whatever the vendor emailed," the first work is normalizing intake.
02

Written decision rules

Someone can state the rules in sentences. Rules that exist only as a person's judgement are discoverable, and discovering them is a separate engagement.
03

A definition of correct

You can describe what a finished, correct output looks like well enough that a second person could check it.
04

Observable state

The process leaves a trail: a record updated, a file written, a ticket moved. Work that leaves no trace cannot be verified, and work that cannot be verified should not run unattended.
05

A bounded blast radius

You know which systems it touches and what the worst realistic mistake would cost. Sending money, sending email to clients, and changing permissions each raise the bar.
06

A human gate on the consequential step

The agent assembles, reconciles and flags; it routes the irreversible action to a named person for sign-off. The gate is part of the design rather than a promise made about it.

Results

What changes when you run the checklist first

You choose on evidence

The checklist turns a ranked wish list into a shortlist you can defend, because each item passes or fails on an observable property rather than on how painful it feels.

Preparation becomes a scoped task

A failed check is a scoped task with an owner, so the work of writing down rules and normalizing intake gets planned instead of discovered halfway through.

The first handoff is survivable

Bounding the blast radius and gating the consequential step means an early mistake is caught and corrected rather than propagated.

Where this fits

Workflow / Automate a Workflow

This is the sorting step ahead of a Tier 1 engagement. One process, end to end, is the lowest-commitment way to see the pattern work, and the checklist tells you which process to pick. Once a workflow runs and holds, the question becomes what else follows the same shape, which is the path toward a full role.

Read more on Automate a Workflow, and on how gates and review surfaces are built in How It Works.

Tier 1 / Prove
How It Works

Questions, answered

The checks are about the process, not the vendor. The relevant question per tool is whether it exposes a way to read and write the data the workflow needs, and whether your access policy permits it. An audit confirms fit before anything is built.

AI Opportunity Audit

Run the checklist on the process you have in mind

Bring the two processes you keep arguing about and run them through the checklist together.