Explore Agentic logo
Library · Playbooks

Playbooks, in the unfashionable sense

Step-by-step guides for the work people quietly need to do: measuring ROI, picking a gateway, running an internal eval. Each one ships with a downloadable asset and a named reviewer.

A playbook here is a procedure, not an essay. Each one is a sequence you can actually run: numbered steps, a named owner per step, the artifact that step produces, and the decision point that determines whether the next step happens at all. If a page cannot tell you what to do on Monday morning, it belongs in the insights library instead.

They are written for the person who owns the outcome and has to defend the number afterwards — a platform lead sizing an agent program, a procurement owner walking into a renewal, an internal audit lead assembling evidence. That means they assume you either hold budget authority or influence it, and they are explicit about which steps need someone else's signature.

Nothing here is tied to a particular vendor. Steps are written against the decision rather than the tool, so the same sequence holds whether you land on a hyperscaler platform, an independent gateway, or something built in-house. Where a step genuinely depends on a platform capability, we name the capability instead of the product so you can check it against whatever is in front of you.

  1. 01
    Governance 11 min Flagship · April 2026

    How to measure AI agent ROI

    The CFO-side framework we now recommend. Includes a spreadsheet and the three questions finance will ask on first review.

    Read
  2. 02
    Reference 5 min Published April 2026

    MCP gateway evaluation guide

    A reader's reference to the five capabilities that separate serious gateways from re-skinned API proxies. Drawn from the published product documentation of the gateways we scored, not from vendor briefings.

    Read
  3. 03
    Template Download DOCX

    A working AI policy template

    Three pages, written to be signable — structured around the evidence an internal audit function asks for.

    Read
How this library works

Frequently asked

  1. What counts as a playbook rather than an article?

    A playbook is a procedure someone can run: a sequence of steps with a named owner, an artifact produced at each step, and a decision point that determines whether the next step happens. Analysis, comparison and explanation live in the insights library instead. If a page cannot tell you what to do on Monday morning, it is not a playbook.

  2. Who are these written for?

    The person who owns the outcome and has to defend the number afterwards — a platform lead sizing an agent program, a procurement owner running a renewal, an internal audit lead preparing evidence. They assume you have budget authority or influence over it, and they are explicit about which steps need a signature from someone else.

  3. Do the playbooks assume a particular vendor or platform?

    No. The steps are written against the decision, not the tool, so the same sequence works whether you land on a hyperscaler platform, an independent gateway, or something built in-house. Where a step genuinely depends on a platform capability we name the capability rather than the product, so you can check whichever vendor is in front of you.

  4. How do I know a playbook is still current?

    Each carries a date and cites the sources its numbers rest on. Pricing and packaging in this category changed repeatedly through 2026, so re-verify any figure that will end up in a business case. The sequence of steps ages far more slowly than the numbers inside it — when something is stale, it is usually a price rather than the procedure.