Client Discovery

Discovery turns an ambiguous consulting kickoff into a concrete map of systems, risks, and decisions. Run it before prescribing tools.

First-week inventory

Capture the following for every product, service, and shared platform:

• Business owner, technical owner, on-call owner, and escalation path.
• Source repositories, branching model, release cadence, and deploy path.
• Runtime environments, cloud accounts, regions, clusters, and networks.
• Datastores, queues, third-party services, and operational dependencies.
• CI/CD systems, artifact registries, image registries, and package feeds.
• Secrets stores, certificate flows, identity providers, and break-glass access.
• Existing Terraform, Helm, Kubernetes, policy, and runbook assets.
• Compliance scope, audit obligations, data classes, and retention rules.

Brownfield assessment

For brownfield work, score each area from 1 to 5:

• Delivery safety: tests, approvals, rollbacks, and release visibility.
• Infrastructure control: drift, state isolation, and ownership clarity.
• Security posture: identity, secrets, vulnerability management, and least privilege.
• Reliability posture: observability, SLOs, incident review, and capacity.
• Developer experience: onboarding speed, local workflows, and paved roads.

Use low scores to create the first risk-reduction backlog. Avoid starting with a platform rewrite unless delivery risk is already controlled.

Greenfield assessment

For greenfield work, define decisions before implementation:

• Cloud provider, account model, network boundaries, and environment map.
• Runtime target, such as Kubernetes, serverless, managed containers, or platform-as-a-service.
• Infrastructure-as-code standard and state backend.
• Deployment strategy, promotion model, and release gates.
• Secrets, identity, logging, metrics, traces, and alerting defaults.
• Policy baseline for access, data handling, images, and supply chain.

Discovery artifacts

End discovery with a one-page current-state diagram, a target-state principles page, and a ranked backlog. Each backlog item should identify risk, owner, expected impact, and the first reversible step.