About / Product Vision
ARB Copilot — a human-in-the-loop agentic AI product for enterprise architecture governance
Public demo pages use static sample data only — no live AI is called and no credits are consumed. Signed-in workspaces run live AI analysis via a managed AI provider (current model: google/gemini-3-flash-preview). ARB Copilot prepares evidence; architects make the final decision. See the AI Data Notice for provider details.
ARB Copilot is operated by Hari Krishna Bodapati / EA2.0, based in India. Built by Hari Krishna Bodapati, a TOGAF-certified enterprise architect. Contact hari@eabyea.com.
ARB Copilot is in early access. We are working with selected architecture teams using sanitized or non-confidential examples to validate ARB intake, routing, and decision evidence workflows.
Product summary
ARB Copilot is a human-in-the-loop agentic AI product for enterprise architecture governance. It helps CIO/CTO offices and enterprise architecture teams convert technology intake requests into structured review analysis, routing recommendations, architecture risk discovery, missing information detection, NFR checklist generation, reviewer task routing, and audit-ready decision evidence.
Problem
Architecture, security, procurement, cloud, data privacy, and AI governance reviews are fragmented. Requests are often incomplete, review routing is manual, decisions are poorly documented, and audit evidence is created late or inconsistently.
Agent pipeline
Specialist agents run in sequence, each producing structured output the next agent can build on. The ARB lead sees the consolidated decision pack and makes the final call.
- Step 1Intake AgentParses raw requests
- Step 2Classification AgentScores complexity & risk
- Step 3Risk DiscoverySurfaces architecture risks
- Step 4Missing InfoDetects gaps, drafts questions
- Step 5NFR ChecklistSecurity, privacy, ops checks
- Step 6Routing AgentAssigns ARB sub-teams
- Step 7Decision Evidence AgentDrafts ADR for the ARB
Human-in-the-loop principle
AI agents recommend. Human ARB leads decide. Final architecture decisions remain accountable to authorized human reviewers.
First MVP use case
AI / SaaS / software onboarding review.
Architecture direction for future enterprise deployment
Designed to grow from controlled pilots toward enterprise deployment. Each layer is chosen so the Copilot can evolve into the system-of-record for architecture governance — without re-platforming.
- FrontendReact / TanStack Start
Fast, deep-linkable UI that fits into existing enterprise portals.
- Agent OrchestrationBackend service / workflow engine
Coordinates specialist agents so reviewers see one consistent decision pack.
- AI LayerManaged AI provider (current model: google/gemini-3-flash-preview)
Live AI analysis runs through a managed AI provider. Alternative and private-VPC providers are on the roadmap for enterprise pilots.
- Data LayerPostgres (managed)
Durable record of every request, recommendation, and human decision for audit.
- StorageObject storage
Holds evidence artefacts and exported decision packs alongside the audit trail.
- WorkflowServer functions + queued tasks
Long-running review processes survive retries, escalations, and reviewer hand-offs.
- Enterprise IntegrationsServiceNow (work-note export today; live integration on roadmap)
Meets reviewers where they already work — no parallel inbox to maintain.
Early-access design partner program using sanitized examples. Let's scope a workflow review for your architecture team.