Skip to content
AnthropicClaude Academy playbooksLockedIn Labs implementation companion

Claude Academy
playbooks.

Bring Claude into the software development lifecycle with clear outcomes, accountable people, and evidence behind each decision.

Explore the playbooks

Anthropic publishes the source courses. LockedIn Labs provides independent application exercises; this companion is not endorsed by Anthropic.

AI-native software deliveryRead → apply → review

One lifecycle.
Six places to improve it.

  1. 01PlanA useful outcome
  2. 02DesignA clear boundary
  3. 03BuildA working change
  4. 04TestEvidence to decide
  5. 05DeployA bounded release
  6. 06MaintainA learning system
Human ownership and review across every stage

The source library

Official playbooks, connected to practice.

Reviewed September 8, 2026. Our review found one playbook course and two tutorials that describe themselves as playbooks. Explore the official reading, then apply it to an original LockedIn Labs exercise.

01

Claude Academy
Course · 60 min

Open at Claude Academy (opens a new tab)Learning and course progress stay with the original publisher.
02

Claude Academy
Tutorial · 15 min

Scaling workflows with Claude Cowork at your organization

Plan Cowork adoption around owned workflows, local champions, support, and evidence from a pilot. Decide how useful practices become shared capabilities.

Who it helps
Transformation leads, department heads, program managers, and adoption champions.
Put it to work
A focused pilot with clear workflow ownership and a defensible expansion decision.
Open at Claude Academy (opens a new tab)Learning and course progress stay with the original publisher.
03

Claude Academy
Tutorial · 20 min

Claude Enterprise Administrator Guide

Coordinate access setup, launch, role-specific enablement, and ongoing adoption. The guide includes a directory of current enterprise configuration resources.

Who it helps
Enterprise administrators, security partners, enablement leads, and program owners.
Put it to work
A launch readiness record that connects technical access with support and learning.
Open at Claude Academy (opens a new tab)Learning and course progress stay with the original publisher.
Explore the full Claude Academy catalog (opens a new tab)

Original LockedIn Labs exercises

Reading is the start.
Your decisions are the work.

Take a concrete scenario through a first plan, a review, and a changed constraint. Bring the result to a teammate or facilitator and explain what you would do.

01Design the handoff before you automate itThe AI-Native SDLC Playbook

Fictional practice brief: an operations team wants an assistant that proposes a category and next action for internal support requests. The pilot uses synthetic requests. A person approves every response, and the assistant cannot send messages or change customer records. Your pod must explain how one change travels from the original problem to a monitored release.

Produce something reviewable

  1. Write a one-page intent: whose problem you are solving, the observable outcome, excluded actions, and questions the product owner must resolve.
  2. Sketch six handoffs: plan, design, build, test, deploy, and maintain. For each, name the input, output, accountable person, and evidence required to continue.
  3. Describe one reusable instruction and one enforceable control. Explain how you would prove the control blocks an unauthorized action, including a route outside the agent's usual tools.
  4. Specify three evaluation cases: an ordinary request, a request containing hostile instructions, and a request the assistant should escalate. Write the expected behavior before trying a model.
  5. Record a release decision with an owner, acceptance criteria, a rollback route, and the operational signal that should send the work back for investigation.

Review the reasoning

  • Another teammate can follow each handoff without the original conversation.
  • Generated recommendations, observed test results, and human approvals are distinguishable.
  • The release cannot gain permission merely because a prompt or a test says it is ready.
  • Unknown results remain unknown; reading the source or completing this worksheet does not prove the workflow works.
Then change one constraint

A sponsor now wants the assistant to send responses automatically. Identify the changed permission boundary, evidence you still lack, who must approve it, and the smallest useful test you would run before accepting the request.

02Earn the next rollout waveScaling workflows with Claude Cowork at your organization

Fictional practice brief: product and delivery teams have access to Cowork, but most people have only tried a short chat. Leadership wants broader adoption. You can support one small pilot and must show that the work improves before inviting another team.

Produce something reviewable

  1. Choose one repeatable workflow for each role, such as synthesizing synthetic interview notes or preparing a delivery-risk review. State the output and the person who will check it.
  2. Prepare a first-session brief: approved inputs, required access, a concrete task, an example of useful output, and the route to help when it fails.
  3. Assign the workflow owner, champion, administrator, and support contact. Define who can create a skill and who can approve a shared version.
  4. Compare a baseline attempt with an assisted attempt on similar work. Record quality, reviewer corrections, time, and cost; label estimates and observations separately.
  5. Write the pilot decision: expand, revise, or pause. Name the evidence behind it and the next review date.

Review the reasoning

  • First-session success means producing a useful, checked output rather than merely logging in.
  • Each shared workflow has an owner who can correct or withdraw it.
  • The pilot tests quality and operational burden alongside adoption.
  • Participants know how to get help and when to stop a task.
Then change one constraint

Usage rises, but reviewers spend longer correcting the output. Decide whether the pilot should expand and propose one change that tests your explanation for the extra work.

03Make the first day work for everyoneClaude Enterprise Administrator Guide

Fictional practice brief: a mixed pod of engineers, product managers, and project managers is about to begin AI training. You are responsible for a usable first day. Some participants need coding tools, others need approved connected information, and all need a clear owner for access problems.

Produce something reviewable

  1. Create a role-to-access matrix showing the minimum tools and information each role needs, the approver, and how access will be tested and removed.
  2. Define a pilot sign-in check using the organization's intended account and permissions. Keep passwords and tokens out of the record.
  3. Prepare a role-specific first task and name the facilitator, office-hours contact, and escalation owner.
  4. List the evidence required for launch: an access test, a checked task output, the support route, and the approval of any unresolved exception.
  5. Specify an adoption review that considers completed useful work, quality, support demand, and spend. Give every measure an owner and a collection method.

Review the reasoning

  • Learning access and production-system access are treated as separate decisions.
  • A successful administrator login is not used as proof that every member's access works.
  • Configuration claims are checked against the current provider documentation and the actual organization.
  • The launch record names who resolves blockers and who can decide to postpone.
Then change one constraint

An engineer can open chat but cannot use the coding tool needed for the first assignment. Show how your launch check catches the gap, routes it to an owner, and preserves a useful learning session while access is resolved.

Make the learning your own.

Use a focused worksheet to define an outcome, set boundaries, assign ownership, and defend your release decision.

Open the implementation workspace

Build the foundations

Useful alongside the playbooks.

These supporting Claude Academy courses can fill a specific gap before you begin applying a playbook.

LockedIn Labs perspectives

From AI adoption to delivery practice.

Our implementation perspective on AI transformation, forward-deployed engineering, and the operating habits behind useful software.