Free checklist · no email gate
The agent-readiness checklist
Ten questions that tell you whether a task in your business is ready for an AI agent — the same filter we run in a Foundations Session. Print it, walk your workflows through it, keep it. No email required.
Johan Iskandar · 12 July 2026 ·
- 1 The work arrives in readable form. Emails, form submissions, documents, tickets — an agent can read anything your team reads on a screen.
- 2 The rules can be said out loud. If your best admin can explain how they'd handle it ('if the enquiry mentions X, do Y'), it can be specced. If every case is a judgment call, it's human work — leave it human.
- 3 It repeats. Daily or weekly, in volume. A task that happens twice a year doesn't pay for a build — say so and save the money.
- 4 The tools have doors. Your CRM, inbox, and calendar are mainstream enough to integrate with (almost all are). Locked-down legacy software is a flag to raise early.
- 5 A miss is survivable and catchable. The work can be reviewed before it matters — drafts approved, updates checked. Work where a single silent error is catastrophic needs gates designed first.
- 6 Someone owns the outcome. There's a person who'll approve drafts during the proving run and own the agent after handover. Orphaned automation rots.
- 7 The knowledge exists somewhere. Answers live in documents, wikis, or one person's head (extractable). An agent can't be grounded in knowledge nobody wrote down.
- 8 The volume hurts. The task costs real hours weekly — enough that recovering them changes someone's week. Payback math needs a numerator.
- 9 Data boundaries are namable. You can say what the agent may read and where data must not go. If that sentence is hard, the compliance conversation comes first.
- 10 You want fewer subscriptions, not more. If the fix you're imagining is 'another platform', an owned agent in your existing stack is probably the cheaper decade.
Scoring, honestly: 8–10 yeses — the task is probably a strong first agent; the remaining question is payback math. 5–7 — scopeable, with the gaps named up front. Under 5 — an agent is the wrong fix right now, and anyone who sells you one anyway is optimising for their invoice. The maybe-cases are exactly what a Foundations Session is for.
Ran the checklist? Bring the winner.
Bring your highest-scoring workflow to a Foundations Session — payback math, fixed quote, plan yours to keep.
Book a Foundations Session →30 minutes · no pitch · if AI isn't the answer, I'll say so