The Challenge
In early 2019, I joined Comcast as a Principal Architect. My organization was in the middle of a management reorg and looking to migrate applications from OpenStack to AWS. At the same time, Comcast's Chief Architect had mandated Concourse as the corporate CI/CD platform.
The problem: the internal Concourse platform had no golden pipelines. Every engineering team was literally reinventing the wheel—writing their own Terraform modules, figuring out how to call the Change Management API, solving security scanning independently. The same patterns duplicated hundreds of times across the enterprise.
ServiceNow and the Change Management API fell under my architecture oversight. I worked with the Change Management Process team—the team that ran CAB (Change Advisory Board)—to review pipelines before teams could get access to the Change API. Every Concourse team had to go through this review because there was no standardization.
Part 1: Building SEED
I was given an exception to the Concourse mandate. SEED had capabilities that Concourse didn't yet support—and I was building a platform that would actually solve the problem rather than perpetuate it.
Within nine months I was promoted to Senior Director of Architecture. My architects provided oversight for engineering teams transitioning out of OpenStack. We onboarded them to SEED, helped them restructure their applications, containerize them, and make them cloud-native.
SEED Platform Architecture
Jenkins Shared Libraries
Declarative pipelines exposed as reusable components
Terraform Under the Hood
Teams never touch IaC—platform handles it
Change API Integration
Baked-in change management—no CAB review needed
Pre-built Deployment Models
Lambda, ECS, static content—all ready to use
The key differentiator: SEED enforced how things were done. Testing and security were baked in. Design patterns were standardized. Teams using SEED got immediate access to the Change API because the pipeline itself enforced compliance.
Part 2: The War on Waste
This is when I saw it clearly: DevOps had become the new waste. The Concourse teams were all solving the same problems independently—and worse, they each had to go through CAB review to get Change API access because there was no way to trust their pipelines.
Concourse Teams (No Golden Pipelines)
- Every team writing their own Terraform modules
- Each team figuring out Change Management API independently
- Required CAB review before Change API access
- Sprints consumed by infrastructure vs features
SEED Teams (Golden Pipelines)
- Golden pipelines with testing and security baked in
- Immediate Change API access—pipeline enforced compliance
- Zero Terraform or CI/CD code required from teams
- Engineering teams back to building features
Part 3: Institutionalizing ADRs
Beyond SEED, I created and institutionalized Architecture Decision Records (ADRs) across the organization. This wasn't just my team creating them—engineers across the org could provide feedback on proposed patterns. Those ADRs directly informed how SEED's design patterns were built.
The ADR Flywheel
Engineers propose patterns
ADRs capture decisions
SEED enforces patterns
Teams ship faster
The insight: when standards are documented formally and enforced by infrastructure, teams stop reinventing the wheel. ADRs became the source of truth that SEED's golden pipelines enforced automatically.
This experience planted the seed for what I'm building now with OutcomeOps—where ADRs become queryable knowledge that AI uses to generate code matching your organization's patterns. Same philosophy: make the easy path the right path.
Key Lessons
DevOps is the New Waste
When every team builds their own pipelines, you're paying for the same work hundreds of times over.
ADRs Become Infrastructure
When decisions are documented and enforced by platforms, teams stop reinventing the wheel.
Make Easy = Right
The path of least resistance should be the path of compliance. That's what golden pipelines deliver.
Further Reading
DevOps Is the New Waste in 2023
The blog post that came out of this experience
OutcomeOps Technical Deep Dive
How ADRs become queryable knowledge for AI code generation
Anthropic Says Build Skills, Not Agents
Why organizational knowledge matters more than autonomous agents
How 3 ADRs Changed Everything
Spring PetClinic proof of concept showing ADR-driven code generation
