Build-Out Timeline – 12-Week Implementation#

Week-by-Week Schedule#

Weeks 1–2: Discovery and Infrastructure Design#

Infrastructure Workstream

• CISO alignment meeting: define “no code leaves network” (Level 1, 2, or 3)
• Design AWS account structure for Bedrock isolation
• Design VPC endpoint + PrivateLink architecture
• Evaluate LLM gateway options (LiteLLM vs. Kong vs. Portkey)
• Begin Terraform for VPC endpoint and networking

Platform Engineering Workstream

• Inventory top 5 codebases by developer count
• Interview 2–3 senior engineers per codebase: “What do you always tell new developers? What patterns do people consistently get wrong?”
• Identify Cohort 1 candidates (25 power users)

Change Management Workstream

• Stakeholder mapping (CISO, VP Eng, team leads, finance)
• Draft communication plan for rollout announcement
• Establish success metrics with leadership

Weeks 3–4: Infrastructure Build + Platform Design#

Infrastructure Workstream

• Deploy VPC endpoint with PrivateLink
• Deploy LLM gateway (internal service)
• Configure gateway: SSO auth, rate limiting, logging
• End-to-end validation: developer laptop → gateway → VPC endpoint → Bedrock → Claude
• CISO review and sign-off on data flow

Platform Engineering Workstream

• Draft managed-settings.json baseline
• Draft org-level managed CLAUDE.md (~30 lines)
• Design context budget worksheet for each major codebase
• Design agent_docs/ directory structure template
• Draft initial 5–8 org-wide skills
• Configure Mobile Device Management (MDM) deployment for managed settings

Change Management Workstream

• Prepare Cohort 1 onboarding materials
• Set up Slack channel for Cohort 1 support
• Brief Cohort 1 candidates on timeline and expectations

Weeks 5–6: Cohort 1 Launch + Platform Iteration#

Infrastructure Workstream

• Monitor gateway under Cohort 1 load
• Tune rate limits and timeouts based on real usage
• Resolve Bedrock inference profile gotchas
• Validate prompt caching behavior

Platform Engineering Workstream

• Cohort 1 developers write first project CLAUDE.md files
• Cohort 1 tests and iterates org-wide skills
• Discover which instructions Claude follows well vs. ignores
• Iterate: rephrase instructions, move content between layers
• Begin agent_docs/ drafting for top codebases

Change Management Workstream

• Daily Slack support for Cohort 1 (first week)
• Twice-weekly stand-ups with Cohort 1
• Collect qualitative feedback: “makes me faster” vs. “gets in the way”
• Identify emerging champion candidates

Weeks 7–8: Standardize + Cohort 2 Preparation#

Infrastructure Workstream

• Scale gateway for 100-developer load
• Configure per-team token budgets
• Deploy CloudWatch dashboards (token usage, latency, costs)
• Instrument OpenTelemetry metrics

Platform Engineering Workstream

• Finalize managed CLAUDE.md based on Cohort 1 feedback
• Finalize org skills library – package as plugin or shared repo
• Create agent_docs/ template and style guide
• Create .claude/rules/ template with path-scoping examples
• Write developer onboarding guide: “How to write effective CLAUDE.md and skills”

Change Management Workstream

• Brief Cohort 2 teams
• Prepare Cohort 2 onboarding session (led by Cohort 1 champions)
• Publish internal documentation from Cohort 1 learnings

Weeks 9–10: Cohort 2 + Cohort 3 Preparation#

Infrastructure Workstream

• Monitor gateway at 100-developer scale
• Validate cost model and projections
• Address rate limiting and capacity issues
• Evaluate provisioned throughput need

Platform Engineering Workstream

• Each Cohort 2 team writes project CLAUDE.md and team skills using templates
• Platform team reviews and coaches – first pass usually too long, too vague, or mixing declarative/procedural
• Iterate based on developer feedback and token usage metrics
• Finalize all configuration for full rollout

Change Management Workstream

• Bi-weekly feedback surveys with Cohort 2
• Weekly office hours
• Prepare Cohort 3 materials
• Brief department leads on rollout plan
• Cost projections reviewed and approved by finance

Weeks 11–12: Cohort 3 Full Rollout#

Infrastructure Workstream

• Scale gateway for 500-developer load
• Activate alerting for capacity thresholds
• Final CISO review on security posture at scale

Platform Engineering Workstream

• Support remaining teams writing project CLAUDE.md
• Address edge cases and non-standard configurations
• Document all platform team processes for ongoing operations

Change Management Workstream

• Department-by-department rollout (not all-at-once)
• 15-minute onboarding sessions led by Cohort 1/2 champions
• Self-service documentation for async onboarding
• Monitor adoption metrics (daily active users, weekly active usage)
• Prepare leadership report: productivity data, cost actuals vs. projections, adoption rates

Week 12+: Ongoing Operations (Phase 3)#

• Quarterly CLAUDE.md and skills review cadence
• Monthly cost and usage reporting to finance and leadership
• Ongoing skills library development based on developer requests
• Model version updates as Anthropic releases new versions on Bedrock
• Continuous refinement of deny rules and security policies
• Annual CISO review of the security architecture

Effort Distribution#

The infrastructure is 30% of effort but often gets 90% of attention. The other 70% – platform engineering and change management – is where the engagement has the most impact.