tracebook
⌖ AI tech support · for machine OEMs

The 2 a.m. call answers itself.

Tracebook turns your shelf of PDFs, your YouTube backlog, and your resolved-ticket history into a sourced, cited AI tech support agent — running under your brand.

live customer Royal Master Grinders

Watch a real diagnostic session resolve →

support session Generation X · #SN-3648
live
tracebook.support · acme manufacturing
a real diagnostic, replayed

The same questions, again and again.

it's 2 a.m.

The customer wants help.

Your service manager is asleep. The right answer is in a 400-page manual nobody else on the team has read cover-to-cover. The customer ends up off-spec or off-line until morning.

scattered

Your knowledge is in eight places.

Manuals in Dropbox, drawings on a shared drive, training videos on YouTube, the real answers in a five-year-old email thread. Even your senior tech can’t find the one PDF where it’s documented.

looped

Same fix, twelve times a year.

“Linear motor not moving” comes in twelve times a year. The fix is always the cable transport module. Nobody wrote it down where the next customer would find it.

Three steps. No magic.

  1. Upload what you’ve got.

    PDFs of operating manuals, electrical schematics, mechanical drawings. YouTube links. Photos of HMI screens. Resolved ticket threads from past support work. Tracebook indexes all of it — per machine model, per customer.

  2. Your customer asks a question.

    Typed, or as a phone photo of a fault screen, or as a 30-second video clip of the noise the machine is making. The system reads it, retrieves the relevant pages of the relevant documents, and answers.

  3. Every claim is cited.

    The answer points back to the exact page of the exact document. If the AI doesn’t know, the chat escalates into a ticket routed to your team. Every resolved ticket becomes a new entry in the knowledge base, so the next customer with the same question doesn’t have to wait.

Built for the way you actually support customers.

grounded

Cited answers, not guesses.

Every claim links back to the exact page of the source document. Open the citation; the customer can verify for themselves.

scoped

Per-machine knowledge base.

A customer asking about their serial number sees their machine’s docs first — including private quotes, sales orders, and customer-specific drawings.

white-label

Your brand, your domain.

The customer-facing portal carries your logo, your colors, and lives at support.yourcompany.com. Tracebook is invisible.

multi-modal

Photos and video clips.

The customer attaches a phone shot of a fault screen or a 30-second video of the machine sound. The system describes it and uses the description to search.

closed loop

Tickets feed the loop.

Unresolved chats become tickets your technicians work on. When the ticket resolves, the fix becomes a new entry in the knowledge base — automatically.

isolated

Single-tenant, your data.

Each OEM is a separately deployed instance on Google Cloud. Your data does not cross tenant boundaries. Nothing trains third-party models.

Live with a real OEM, right now.

Royal Master Grinders — a precision-grinder OEM with a fleet of machines deployed across North America — runs their entire customer support flow on Tracebook. Manuals, drawings, training videos, and historical case files were ingested in two weeks. Their service team has been answering fewer 2 a.m. calls ever since.

“That’s exactly what I wanted to see. It’s just putting the pieces together — wow.” — John, Royal Master Grinders, watching the system diagnose a real Gen-X fault.

at-a-glance
Machine fleet indexed
50+ serials
Manuals + drawings
800+ docs
Customer chat sessions
live
Time to first answer
< 45s typical

Per-OEM SaaS. Quote-based.

Pricing scales with your machine fleet and document volume. Smaller OEMs land lower than you’d expect; bigger fleets get bulk-tier rates. Reach out for a quote — we’ll ballpark in 24 hours.

Get a quote

The real questions.

Where does my data live? +

In your dedicated tenant on Google Cloud. Each OEM is a separately deployed instance with its own database, file storage, and vector index. Your data does not cross tenant boundaries.

What happens if the AI doesn't know the answer? +

The chat escalates into a ticket. A real technician on your team takes over; the customer can attach files; the conversation continues until resolved. The resolved ticket becomes structured knowledge the next chat can retrieve from.

Do you train on my data? +

No. Your data is not used to train any third-party model. Inference happens on Google Vertex AI; retrieval happens against your tenant’s own vector index.

How long does setup take? +

Two weeks for a typical OEM with ~10 machine models and a few hundred documents. The bulk of that time is uploading and tagging content — we run alongside you on it.

What machines / industries does this work for? +

Anything where the customer is supported via written manuals, schematics, and historical knowledge: precision tools, packaging machines, food-and-beverage equipment, building systems. The first deployed customer is in machine-tool grinding; the same plumbing applies to any OEM with shipped machines under warranty.

Can I use my own logo and domain? +

Yes. The customer-facing portal lives at your subdomain (support.yourcompany.com for example) with your logo and brand colors. Tracebook does not appear in the customer experience.

What does it cost? +

Per-OEM SaaS, scaled by fleet size and document volume. Email cmisztur@mriiot.com for a 24-hour ballpark.

Want to see it on your data?

A 30-minute demo, plus a per-OEM ballpark within 24 hours.

Email cmisztur@mriiot.com