Your lead estimator
gives notice.
Ten years of pricing logic
walks out the door.
Company Brain ingests the documents your company already has — quotes, job notes, emails, vendor files — and makes them queryable. Ask a question, get an answer with a source citation.
The behavior change is on our side — we do the ingestion, tuning, and ongoing curation. Your team just asks questions.
Two to four weeks of scoped setup, then ongoing tuning. Not a self-serve tool.
The owner is still the system.
Your operations manager gives notice. A decade of vendor relationships, job history, and field judgment walks out the door.
Your new PM asks which subs were on the Millbrook project — and who to never use again.
A foreman needs the spec from a job three years ago. Nobody knows where to look.
Everyone routes it to you — because you're the only system left.
As long as you're the system, you're also the ceiling. The departure is the reason to start. The owner stepping out of routine answers is the reason to keep it.
Your company
already knows
the answer.
Ten years of knowledge is sitting in your sent folder, quote files, and job notes right now. Nobody can find it fast enough to use it — so they ask the person who already knows. Company Brain reads everything that's already there and returns answers with citations back to the source.
What it reads
How it flows
Answer with source citation
Ten years of documents also means ten years of outdated pricing and superseded SOPs. Part of what we do is teach the system which sources to trust, and flag conflicts instead of guessing.
Every other solution
asks you to change first.
The tools already exist. The problem is they all require someone to start putting things in before they return anything useful. Company Brain starts with what's already there.
| Solution | Behavior change first? | Works on existing mess? | Conversationally queryable? |
|---|---|---|---|
Shared drive / Notion / wiki Someone has to write it first | Required | No | No |
Hire SOP consultant Major project; output still isn't queryable | Required | No | No |
NotebookLM / Claude Projects Works up to ~50 docs. No live inboxes, no ongoing curation | Required | Partially | Partially |
Company Brain Ingests what's already there | Not required | Yes | Yes — with citations |
For a 3-person shop with a clean Google Drive and 200 documents, NotebookLM is probably the right answer — it's free and it works. Company Brain is for the 14-year company with thousands of documents across scattered systems that no consumer tool was built to handle.
Ready in weeks.
Built around how
you actually work.
Not a login we throw over the fence. The first version is scoped around real questions, connected to real sources, and tuned with the people who need it most. The hard part isn't the technology — it's the judgment calls about what to ingest, what to exclude, and how to handle conflicts. That's the consulting.
Where we start
“What happens if your operations manager leaves tomorrow? We start there — build her vault while she's still here, capture the vendor relationships, job history, and field judgment that never got written down. Highest-risk gap first. Then expand.”
Pricing
Setup: $4,000–$8,000 depending on scope and document volume. Monthly retainer: $500–$1,500 for re-ingestion, tuning, and team support. The retainer is what keeps the system current — without it, it decays. We'll land on both numbers in the first call.
Map the knowledge mess
We get access to the inboxes, drives, and job systems that matter. We find the landmines — the "OLD pricing DO NOT USE" folder that's been the reference for three years, the shared inbox nobody monitors, the estimator who keeps everything in a personal Gmail.
Build and tune
We pick one high-value slice — the last 3 years of accepted quotes, the vendor email threads — and build the first working index. Then we sit with your operations manager and key leads and have them ask real questions from their actual workday. Half the answers will be incomplete on first pass. We fix it.
Handoff and guardrails
We train the team on how to ask questions that work, how to flag wrong answers, and how to add new documents going forward. We set up a feedback loop so we can tune it again in 30 days.
Monthly tuning and curation
New documents need to be ingested. Stale ones flagged. The team finds new failure modes. If you're not in a retainer relationship, the system decays. If you are, it compounds. This is why we price it as an ongoing engagement.
Built for the company
with years of context
and nowhere to put it.
The signal is accumulated complexity in specific people — not years in business. A 4-year company whose lead estimator has written 500 quotes has more at-risk knowledge than a 12-year company that's kept clean records.
See If You Fit5–25 person HVAC, plumbing, electrical, roofing, GC, or similar service team
A lead estimator, PM, or owner holding years of pricing logic, vendor history, and job context no one else has
Owner still answers routine operational questions most days
Knowledge spread across email, notes, drives, and job folders — never formalized
Who this isn't for
If no single person is holding critical knowledge the rest of the team can't easily access, or your documents already live cleanly in one system — you probably don't need us. Try NotebookLM or Claude Projects first. They're free, and for a tidy operation they'll do the job.
We'll tell you this on the first call if it applies. The people who do fit trust the rest of what we say because of it.
Your documents
aren't for training
someone else's AI.
Most AI tools are consumer products. Uploading your quotes, contracts, and job history to them carries real risks that most business owners don't read the fine print on.
- Consumer AI tools can use your inputs to train the model. Your pricing strategy, customer details, and internal processes become part of the training set.
- No committed deletion timeline. Documents sit on third-party servers indefinitely unless you request removal — and that removal isn't always clean.
- "Private" doesn't mean no one sees it. Providers reserve the right to review conversations for safety purposes. Sensitive contracts and personnel notes have no privilege protections once uploaded.
Air-Gapped
Nothing leaves your building.
The AI model runs entirely on your hardware using open-source models. No data ever touches an external server. Best fit for regulated industries or highly sensitive documents.
Tradeoff: lower model quality than cloud AI.
API — Secure by Design
Commercial-grade. Not the consumer chatbot.
We use Claude's commercial API — a contractually different product from Claude.ai. By default: your data is never used to train the model, retained for 7 days only, then deleted. No opting in required.
Best quality. Contractually clean. This is what we recommend for most clients.
Hybrid
Docs local. Queries go out, not documents.
Your source documents stay on your machine or internal network. When someone asks a question, only the question and the matched excerpt — not the full file — are sent to the API.
Best of both: local control with full model quality.
We build it until your team
asks the system
before they ask you.
Your team gets answers in seconds. You get pulled in for judgment calls, not routine lookups.
Book a Setup Call