🔒 Private — prepared for Briana and Cindi (RCB principals). Not for staff.
Overview / Session Notes / Karen — Machine Maintenance

Session Notes: Karen — Machine Maintenance

June 3, 2026 · Nate St. Pierre, Karen

Overview

This was a short focused session (about 10 minutes) that Karen initiated — she raised machine maintenance as a possible efficiency improvement before Nate asked. The conversation moved quickly from problem description into a generative brainstorm about what an AI-powered maintenance system could look like. Karen had clearly been thinking about this, and the session gave her a space to think it through out loud.

The Problem

RCB has roughly a dozen different machine types on the production floor, each with its own maintenance requirements and schedule. Some machines require daily maintenance; others have weekly, monthly, six-month, or yearly steps. The production team handles maintenance, but there's no consistent system for logging what was done, when it was done, and by whom.

The consequences Karen named: when a machine has a problem, there's no historical record to look back at for patterns. It's hard to tell whether a breakdown was preventable, or whether a recurring issue is actually a recurring issue. Accountability is also a gap — when maintenance doesn't happen, there's no trail to understand why or to follow up.

The UV printers are the highest-stakes machines in this regard. There are three of them, and they require daily cleaning — sometimes multiple times a day when running continuously — to prevent the ink heads from clogging. Each print head costs approximately $2,000. When a UV printer goes down, it's a significant operational problem. The lasers have their own maintenance needs (weekly cleaning, sensitivity to dust particles) but are somewhat less critical in terms of failure cost.

Beyond logging, troubleshooting is also manual and fragmented. When something goes wrong with a machine, operators rely on whatever documentation can be found — manufacturer manuals that exist somewhere as PDFs — and individual experience. Karen has done work to gather and organize this material, but searching through a 200-page manual in the middle of a problem is slow and frustrating.

There's also a maintenance calendar of some kind, but it's not functioning well as an accountability tool. The daily maintenance tends to happen. The less frequent steps — monthly, six-month, yearly — are easier to miss.

What an AI-Powered Maintenance System Could Look Like

Karen's thinking and Nate's sketching converged on a concept for a unified maintenance system — roughly described as a single app accessible from shared computers on the production floor, designed around how operators actually work.

The core elements discussed:

Machine-specific AI assistant. Each machine type would have its own AI-powered interface, loaded with that machine's manufacturer manuals, past maintenance notes, and any recorded operator knowledge. An operator who runs into a problem could ask a question in plain language and get an answer drawn from the actual documentation — without searching through a manual. The AI would know which machine it is and respond in that context.

AI-written maintenance logs. Rather than requiring operators to stop and write log entries, the system would let them describe what they did verbally — and write the log for them. Nate demonstrated how his own system works this way: he talks through what he's doing while he's doing it, and the AI captures and organizes it. Karen's response was immediate: "That sounds amazing." The goal is that maintenance logging becomes close to zero additional effort — the operator narrates, the system records.

Maintenance calendar. The app would show today's maintenance calendar on opening — which machines need what, based on their schedule. This creates a clear daily checklist and a record of what was completed.

Operator training recordings. Karen mentioned that stored video walkthroughs of how maintenance procedures are performed would also be valuable, especially for newer operators who haven't learned a particular process yet. Those recordings, combined with the AI Q&A, would reduce dependence on whoever happens to have the most experience with a given machine.

Karen acknowledged openly that the full vision is a significant build, and that the question of what to actually prioritize and fund is above her. But she was clear that machine maintenance matters — and that the UV printers in particular make this more than an organizational convenience problem.

Context on Nate's System

Nate briefly described his own AI operating system as a real-world example of voice-driven AI logging — specifically how he uses a voice transcription tool (WhisperFlow) combined with Claude to capture notes, update records, and manage his workflow by talking rather than typing. Karen was genuinely interested. This demonstration helped her see the voice-logging concept as something that exists and works, not just a hypothetical. She mentioned she'd like to see a live demo during the Day 6 training session.

Follow-Ups

  • Collect manufacturer PDFs for each machine type from Karen or from wherever they're currently stored
  • Confirm what shared systems exist on the production floor (browser access, installed software, etc.)
  • Determine whether Briana and Cindi see machine maintenance infrastructure as a priority investment
  • Ask Karen if any maintenance video walkthroughs have ever been recorded