Your Process, Every Conversation

Define it once. Every conversation follows it.

Your process lives in the workflow, not in a wiki nobody reads. The AI drives the conversation through it, step by step. Change a step — it appears everywhere. Remove one — gone.

Your Process, Every Conversation

What this pillar covers

  • 1

    No master prompt. Each step arrives when the conversation needs it.

    Instructions per step, not per conversation — no 40-page prompts.

  • 2

    Plain language a human can read. Structure the AI can run.

    Plain-language step definitions — the process reads like a checklist.

  • 3

    Building blocks you mix and match across playbooks.

    Reusable step fragments — define once, compose everywhere.

27
Ready-made playbooks
1
Edit to update every conversation
0
Retraining sessions needed
Step-at-a-time instructions

No master prompt. Each step arrives when the conversation needs it.

Your whole process does not get dumped into one giant prompt. Each step injects its own instructions at the moment the conversation reaches it. The AI never sees Step 5 while it is working on Step 2 — impossible to forget what it has not seen yet. (The technical term is Progressive Disclosure of Steps — the same principle that made software usable, applied to AI conversations.)

  • Instructions per step, not per conversation — no 40-page prompts.
  • The AI only ever sees the current step — fresh guidance, every time.
  • More steps, same focus per step. Performance stays flat as processes grow.
Readable process

Plain language a human can read. Structure the AI can run.

Playbooks are written in plain English, not trapped in a wiki or buried in implementation details. Five lines of readable process beats three paragraphs nobody opens. If you can write a bulleted list, you can define a playbook.

  • Plain-language step definitions — the process reads like a checklist.
  • Version-controlled like anything else — diffs show what changed and why.
  • Non-technical authors can edit a playbook without touching code.
Composable fragments

Building blocks you mix and match across playbooks.

Sales approval and procurement approval share the same sign-off step. Content publishing and blog posts share the same SEO review. Define the step once as a fragment, compose it into every playbook that needs it. Change the fragment — every playbook updates.

  • Reusable step fragments — define once, compose everywhere.
  • Change a fragment — every playbook that uses it updates instantly.
  • Team-wide consistency without copy-paste drift.
Change once

Edit a step. Every conversation on that playbook picks up the change.

No memo, no retraining, no wiki refresh. Change a step in the workflow — the next conversation that reaches that step sees the new instructions. In-flight conversations pick up the new version at the next step boundary.

  • Add a step — it appears in every conversation on that playbook.
  • Remove one — gone everywhere. No orphaned instructions lingering in prompts.
  • No retraining, no memo, no wiki update — the workflow is the source of truth.
Process outlives people

New hire, day one. Same structured conversations as your best performer.

Because the process lives in the workflow — not in someone's head, not in a wiki nobody reads, not in a master prompt someone forked two years ago — it does not leave when a person leaves. Organizational memory, not tribal knowledge.

  • Senior contributor leaves — the playbook keeps running.
  • New joiner opens a task — the AI drives the same process a veteran would.
  • Process is organizational memory. It outlasts the people who defined it.
See it in motion

See a playbook drive a conversation

Watch a blog post move through its playbook — each step surfacing its own instructions as the conversation reaches it. The AI drives. The author responds.

1
2
3
4
5
Claude
Watch the workflow in action
0 / 5

Master-prompt approach vs step-at-a-time

Master-prompt approach

  • All instructions upfront — one giant document the AI must re-read every turn.
  • The AI stops paying attention halfway through, loses guidance buried deep in the document.
  • More steps means a longer document, which means worse performance as work grows.
  • Changing one step means editing, re-testing, and re-shipping a monolithic file.
  • The process lives in the prompt. Edit it wrong and every conversation breaks.

Step-at-a-time (ConvOps)

  • Each step injects its own instructions at the moment the conversation reaches it.
  • The AI only sees the current step — impossible to forget what it cannot see.
  • More steps, same focus per step. Processes can grow without degrading performance.
  • Change a step in the workflow — it updates everywhere, in every conversation, instantly.
  • The process lives in the workflow. The AI's standing instruction becomes four words: follow the workflow.
Questions

Frequently asked questions

No. If you can write a bulleted list, you can define a playbook. The process is plain language the workflow engine understands. Non-technical process owners ship their first playbook on day one.

A wiki documents what the process should be. A playbook runs the process inside the conversation. The AI reads the current step, executes it, and asks for your input when it needs something. The wiki describes. The playbook drives.

Every new conversation on that playbook picks up the change immediately. In-flight conversations adopt the new version at the next step boundary. No memo, no retraining, no migration script. One edit, every conversation.

Yes. Steps are composable fragments — define a "legal review" or "SEO pass" once and drop it into any playbook that needs it. Change the fragment once and every playbook that uses it updates.

No. Each step's instructions arrive when the conversation reaches that step. The AI is never staring at a 40-page prompt wondering which paragraph applies now. It sees the current step — and only the current step.

The playbook keeps running. The process lives in the workflow, not in that person's head. New hires inherit the same structured conversations as your best performer — day one, no training.
ConvOps

Define it once. Every conversation follows it.

Your process lives in the workflow, not in a wiki nobody reads. The AI drives the conversation through it, step by step. Change a step — it appears everywhere. Remove one — gone.