AI Solutions Architect [WIP]

This guide explains how to use solutions architect prompts to generate a systems map and solution plan using evidence-based cross-repo delivery. This is a pattern where AI produces verified, cross-repository architecture context and then turns it into execution-ready PR plans using the tools like enterprise code search and GitHub.

If your system is small or well-documented, you can often jump straight to implementation. But in large orgs, changes routinely span services, shared libraries, configs, and infra where “tribal knowledge” and outdated docs can create huge misses.

This workflow solves that by cascading AI through two distinct stages:

AI starts from a concrete artifact (compose/helm/terraform/arch doc/repo path), discovers what’s actually connected across repositories, and produces a SYSTEMS_MAP.md that includes:

• service inventory (purpose/tech/ports)

• explicit call graph (who talks to whom)

• auth model summary (who validates tokens, where middleware lives)

• evidence references (file paths + line numbers) for every claim
  Anything that can’t be verified is explicitly marked Unknown.

With the systems map as the source of truth, AI generates one Markdown plan per recommended PR, each scoped to exactly one repo, detailing:

• proposed file/component/contract changes

• schema and data-flow impacts

• security/compliance considerations

• testing and rollout/migration steps

• risks, unknowns, and open questions

If evidence is missing, the plan proceeds with explicit assumptions and Unknown markers—so humans can step in, correct, and restart the flow without losing momentum.