AI Studio × Atrium — the map, not the weeds.
One visual summary of what we're building, how the big platform and the HOCATT consolidation fit together, and how to review the 13 PRDs without drowning in them.
Ed's key architecture call: every client gets their own fully isolated instance — own database, own infrastructure. Nothing is shared between clients. "Multi-tenant" exists only at the top, where Result Labs runs the machinery that stamps instances out.
Result Labs' master panel. A customer signs up, pays, and the system auto-provisions their isolated Command Center — then watches, upgrades, and bills the whole fleet.
WHO SEES IT: Result Labs onlyThe per-client product: one trusted view of sites, analytics, CRM, ads, content, bots and AI agents — finds the leaks, proves what works, turns findings into profitable action. The 13 PRDs describe this layer.
WHO SEES IT: each client's team, roles-scopedThe hard, client-specific work of collapsing existing tools into one portal. Atrium (admin.hocatt.com) is the first one — 18 HOCATT systems become one role-based front door.
WHO SEES IT: HOCATT's team — Lisa logs in and sees only salesEvery PRD in one card. Chips show build triage (now / soon / someday) and size (S–XL as spec'd). Tiers are the recommended review order — spend your energy where feedback changes what gets built first.
One dashboard mapping every platform a client runs — flags what's broken or wasting money, findings become projects. The productized HOCATT control center.
Plug-in CRM connections — GoHighLevel first, read-only — into a clean warehouse with automatic data-quality flags (dupes, fake “$0 won” deals).
Logins, roles, invites, and audit trail inside each instance. Must reconcile with Atrium's “few roles, many segments” model — our biggest structural feedback.
Chat + voice agents trained on the client's own content, placeable per role (a sales bot only Lisa sees), with every AI call metered for billing. Faith's future home.
A rebuilt crawler that actually reads modern sites, served to AI assistants through each client's private MCP endpoint. Ask: build vs buy?
Privacy-friendly analytics that also shows which AI crawlers are reading your site — “ChatGPT read your pricing page.” Really two products in one PRD.
A per-client content brain: every page and doc as clean, versioned, searchable markdown that agents can cite. The knowledge base under everything AI.
Register and verify each site, install tracking, connect Google — plus the full ported SEO toolkit. The registry is now; full SEO parity is negotiable.
The one-button machinery that stamps out, upgrades, bills, and retires client instances. Ed's own roadmap says: hand-provision the first clients, build this later.
Client-built dashboards, calculated KPIs, scheduled branded reports — the Databox answer. Its “metric registry” foundation ships with v1.
The ~80-connector plan (Klaviyo, Shopify, CallRail…) shipped in demand-ranked waves on a shared framework, so no client is ever blocked.
Compose, approve, schedule, publish socials from the command center — tied to actual site visits and CRM leads, not just likes.
Draw the ideal customer flow, overlay what visitors actually did — leaks become quantified findings. Its event-schema groundwork ships with v1.
Track A ships Atrium in weeks — the HOCATT team feels the consolidation by end of July. Track B builds AI Studio in months on Ed's roadmap, entered only after our PRD review trims it 60/40. The bridge makes Track A the instance-zero of Track B — nothing we ship now is throwaway.
Atrium's shell calls the proven backends — Faith, onboarding, GHL sync, control-center — as services. No duplicated builds.
Ed's tokens + Atrium's role/segment navigation. Decisions made for Atrium land as PRD feedback the same week.
Everything in Atrium is either a disposably thin link/embed, or built clean enough to back out into the deployable base.
What the HOCATT team actually uses (and ignores) is evidence for the 60/40 PRD decisions — before Track B builds the equivalent.
Funding and team reality set the pace. Every phase has an explicit gate; no calendar heroics.
Each one is a real decision with an owner and a date — not a status meeting.
Green-light. Invoice sent, Ed looped on security fixes, greenfield-shell guardrail signed off.
Atrium demo to Daniella; go / no-go on the end-of-July announcement to Andre.
Scope freeze. PRD feedback consolidated, all-three session done, v1 base system defined — then design.
Continuity plan. Every August work item has a non-Jason owner before surgery on Aug 6.
The switch. HOCATT migrates from wrapped systems to native AI Studio modules — only when parity is proven.
Carry these into every page of the review. If a feature can't answer them, it's a someday.
Who's the buyer, and what problem does this uniquely solve?Filter and frame: micro-problem → macro-problem (money in, cost down, time saved, pain reduced).
Would any of the first five clients notice if this shipped six months later?The 60/40 test. If not — defer it.
What does this cost per client, per month?Isolation multiplies every infra, API, and AI cost by client count.
Does the simple version exist?The spec defaults to enterprise-grade. Ask what "startup mode" looks like.
Does HOCATT already have a working version?If yes: port or absorb it — don't respec it.
Spotted while cross-referencing the PRDs against this week's meeting decisions. These are feedback for Ed, not criticism — the set is otherwise remarkably complete.
A “baby ticket system” is a pre-launch requirement for Atrium (Faith needs somewhere to file support items) — and every future client will want one. Candidate PRD-14.
The my.hocatt.com concept — onboarding, training, shipment tracking, chat agent for customers — was adopted this week and generalizes to every client. Candidate PRD-15.
PRD-02 fixes 4 roles; Atrium runs 9 roles + segments (“SEO is a view, not a role”). One answer needed before either build hardens.
HOCATT's onboarding system exists and ports in as-is — but the PRD set has no module for a client-business onboarding/ops system. Decide where it lives.