Intelligence Platform Proposal — May 14, 2026
A phased plan to extract what G and Ori know, make it queryable by every employee, and build the data foundation that powers your push from days to hours — without adding headcount.
Discovery — May 13, 2026
53 years of operational knowledge, two people's heads, zero structure. These are the six pressure points you named — and each one has a direct solution path.
G and Ori are the single source of truth for virtually every operational question. The team interrupts constantly. The company's pace is capped by how many questions two people can answer in a day.
Ori said it plainly: "We've almost crippled our team." The fix isn't training — it's externalizing the knowledge into a system the team can query at 3x the speed of a human answer.
The SOPs exist. They're just not findable, not current, and not adopted. Ori is sitting on 6-8 years of documentation she can't audit manually against current software stacks.
Solution: ingest everything, filter against the active tool list, flag what's stale, and surface only live SOPs at the point of need — not as a PDF someone has to hunt for.
Critical files live in employees' "My Drive" — shared ad-hoc, never catalogued. Chris spent 13 months learning how to hunt. That institutional cost compounds across every new hire.
A living index automatically maps every document — what it is, who owns it, where it lives, when it was last updated — and feeds that map directly into the shared brain.
You know some suppliers produce more unsellable hats than others — but you have "almost no visibility" on the actual percentages. That blind spot is a margin problem hiding in plain sight inside Fulfill.io.
Supplier quality scoring pulls from order data and return/reject signals already in your ERP — no new data collection required.
G's goal is explicit: measure lead time in hours, not days. You already beat competitors on weeks-to-days. The next compression requires warehouse intelligence — pick routing, staff scheduling, product placement optimization — fed by real data, not intuition.
40% of revenue is DTC. Your fit quiz, review history, and order data contain a goldmine of customer signals — but they're not connected. Personalized product recommendations, review-driven product pages, and repurchase triggers are all sitting in disconnected systems.
The diagnosis
Every pain point traces back to the same root: the company's intelligence lives in people's heads and disconnected systems, not in a queryable structure.
A knowledge graph is a living, queryable database of everything your company knows — SOPs, Loom videos, order history, supplier performance, file locations, customer signals. It's the shared brain that doesn't forget, doesn't get interrupted, and retrieves the right answer in under a second.
The Loom videos Ori mentioned are exactly the right starting point. They're crude oil. Run through the right pipeline — transcript, structure, index, ingest — and each video becomes a node in the collective brain. Any employee can ask "how do I do X" and get G's actual process, in G's actual words, without paging G.
The brain compounds. Every interaction, every new SOP, every Fulfill.io update adds signal. In 6 months the system knows more than any individual. In 12 months it outperforms the tribal knowledge it replaced.
The engagement
Each phase delivers standalone value and unlocks the next. Phase 1 is the prerequisite — everything else is built on the foundation it creates.
Phase 1 — Weeks 1–6
Build the brain. The single deliverable that makes everything else possible.
What we need from you to start: Loom workspace access (view/export), Google Drive read access (service account), Fulfill.io API credentials or scheduled export, and 2-hour working session with Ori to map the SOP categories and department structure.
Phase 2 — Month 2 onward · $5,500/mo
The brain is running. Now give G and Ori their counterparts — agents that know everything they know, available to the whole team, 24/7.
GarBot is trained on G's decision patterns, his language, and 53 years of operational institutional knowledge extracted from Looms, meeting transcripts, and SOPs. It can answer the questions G gets interrupted for 20 times a day — without paging G. It flags the ones it can't handle with the right context already pulled.
Available in Google Chat. Accessible to any team member with the right permission tier. Logs every interaction so G can review — and every gap it surfaces becomes a knowledge-graph enrichment task.
OriBot knows every SOP, every process, and every tool the team uses. When someone is about to do something the wrong way, OriBot intercepts — via Google Chat nudge — before the mistake compounds. When someone needs "how do I do X," OriBot returns the live SOP, the relevant Loom, and the step count.
This is the IKEA model Ori described: no walls of text, no hunting. Point at the step, show the picture, done.
Every supplier scored automatically against order data: reject rate, unsellable rate, lead time variance, volume history. Refreshed weekly from Fulfill.io. G gets the supplier ranking he currently has almost no visibility on — without any manual analysis.
The data is already in Fulfill.io. This surfaces it.
Knowledge graph maintenance (new Looms, SOP updates, Drive changes auto-indexed) · Monthly ops intelligence digest for G and Ori · Knowledge gap log review (what the agents couldn't answer — backfill sessions) · 1x monthly working call (60 min) with your team · VPS hosting + agent uptime monitoring
Phase 3 — Month 4 onward · $7,500/mo
The operational brain is running. Now turn the same data layer toward your customers and your growth engine.
Connect Shopify, Amazon, and your fit-quiz data into a unified customer graph. Every customer has a profile: hat preferences (brim width, crown, material, style), purchase history, reviews written, fit quiz responses. The engine uses that profile to surface the right product, the right recommendation, and the right re-engagement trigger.
Product pages become dynamic — surfacing reviews from customers with similar fit profiles. Email sequences become personalized — based on what a customer actually bought and what customers like them buy next. The fit quiz stops being a widget and becomes an acquisition engine.
B2B order patterns analyzed for ICP signals: which wholesale buyers reorder fastest, highest volume, lowest returns, longest relationships. Build the ideal customer profile for the Fair.com channel — and use it to target outreach, prioritize account management, and flag at-risk accounts before they churn.
Chris mentioned this was on the active list. The data exists in Fulfill.io + your CRM. The intelligence layer connects and surfaces it.
Order velocity data by SKU, time of day, and channel — used to recommend optimal product placement, pick-route sequencing, and staffing patterns. The same data G is currently running on intuition, made explicit and automated.
This is the path from days to hours. Not through adding headcount — through removing friction from the picking and packing workflow itself.
Investment
Simple structure. No long-term lock-in. Phase 1 is fixed-fee; Phases 2 and 3 are month-to-month with 30-day cancellation.
| Phase | Investment | Term | What you get |
|---|---|---|---|
| Phase 1 — Knowledge Foundation | $22,500 | One-time · 6 weeks | Knowledge graph built, Looms + SOPs + Drive ingested, first agent live in Google Chat |
| Phase 2 — Ops Intelligence | $5,500/mo | Month-to-month · 30-day notice | GarBot + OriBot, supplier quality dashboard, SOP enforcement, monthly digest + call |
| Phase 3 — Revenue Intelligence | $7,500/mo | Month-to-month · 30-day notice | Everything in Phase 2 + customer intelligence, B2B ICP engine, warehouse optimization |
| VPS Hosting | Included | Ongoing | Dedicated server — your data never touches shared infrastructure. Included in monthly retainer. |
Our stack
No dependency on Claude, Gemini, or GPT. The platform is model-agnostic — it works with any underlying AI, which means no vendor lock-in and no disruption if the AI market shifts.
Your dedicated VPS — one server, one tenant. Your Loom transcripts, SOP archive, order data, and customer signals never touch shared infrastructure. MERIDIAN runs the knowledge graph (Neo4j) and the agent layer on a dedicated node you can verify.
Google Workspace native — agents surface in Google Chat, the tool your team already uses. No new app. No onboarding curve. The first thing employees notice is that answers come faster, not that a new system appeared.
Fulfill.io connected — we integrate with your existing ERP via API or scheduled export. No disruption to current workflows. The intelligence layer sits on top of what you already run.
On-premise option available — if you want the server on-site rather than in the cloud, the architecture supports it. We've built for edge computing. That's a conversation for Phase 1 scoping.
Next steps
The follow-up call is May 14 at 2:00 PM. Here's what closes Phase 1 and gets the clock started.
Appendix
The server is dedicated to American Hat Makers — no shared tenancy. You have read access to the server at any point. All data ingested (Loom transcripts, SOPs, order data) stays on that server. We don't send it to third-party AI services — the model inference happens locally or through your existing Google Workspace Gemini subscription if you prefer. On-premise is also an option if you want the hardware in your facility.
The first demo of the knowledge graph — with your Looms ingested and queryable — happens at the end of Week 2. The first agent in Google Chat goes live at the end of Week 6. Phase 2 agents (GarBot + OriBot) come online 4-6 weeks after Phase 1 completes. The supplier quality dashboard appears on the same timeline. ROI on Phase 1 is measurable by Month 2.
Phase 1 is fixed-fee — once the agent is live and the knowledge graph is built, that work is yours. The monthly retainer (Phases 2 and 3) is month-to-month with 30-day written notice. We don't believe in long-term contracts that outlast the value delivered. If it's working, you'll stay. If it's not, you should leave.
Phase 1: one 2-hour kickoff session with Ori, plus credential sharing for Loom, Google Drive, and Fulfill.io. That's it. No weekly check-ins, no homework. We build; you review the demo at Week 2 and sign off on the agent at Week 6. Monthly retainer: one 60-minute call per month. Everything else is async.
No — and in fact, it complements it. MERIDIAN's platform is model-agnostic. If Gemini is your preferred AI layer through Google Workspace, we can route agent inference through your Gemini subscription. The knowledge graph and the agent architecture are independent of the underlying model. You can run Gemini today and swap to a different model in 12 months without rebuilding anything.