Session Notes: General Discovery Conversation with Sandra

2026-06-02 · Sandra, with Nate St. Pierre · ~90 minutes, remote video call

Overview

This was a remote video call between Nate and Sandra, held the morning after Nate's full first day on-site at Merrick. Sandra had been working from home that week at her own pace. The conversation ran roughly 90 minutes and covered several topics: the process Janet manages for pricing updates and new product additions, the production checklist and labor-data capture system Sandra built, the email template workflow, the freight calculation tool Calloway had been working on, and a broader discussion about AI tools and what training for the team might look like.

This conversation provided significant depth on the production and operations side of the business, which complemented the order-entry and customer-facing context gathered during the Day 1 on-site sessions.

Janet's Process: Pricing Updates and New Product Addition

An early topic was the workflow Janet manages for updating vendor pricing and adding new products to the catalog. Sandra was direct about the situation: the process lives entirely in Janet's head and in her own spreadsheets. Neither Sandra nor Calloway have clear visibility into how it works in practice, and it has never been documented in a way that others could replicate.

When a vendor sends a pricing update, those changes need to flow into Ordova (the order management system), then get pushed out to both of the industry distributor catalogs Merrick lists on, and to the web store as well. Janet handles this chain, but the mechanics of how she does it — what spreadsheets she uses, how the margin calculations work, how the products get entered into each platform — is not written down anywhere. Sandra noted that she's asked Janet before to document the process so others could learn it. “It shouldn't be a secret,” she said. Janet hasn't yet created that documentation.

This matters beyond just pricing updates. Adding new products to the catalog goes through the same basic path, and Sandra identified this as “one of the biggest bottlenecks we have.” A specific example she gave: one of their vendors carries a wide range of crystal awards Merrick could be listing and selling. They're not in the catalog yet, primarily because the process to get them there is bottled up with Janet. Sandra noted that she personally has more quiet time than Calloway — she could sit down during slower periods and add product — but she can't do it without a documented process to follow.

The front-end of the product-addition workflow involves an Excel spreadsheet where vendor cost is entered, competitive pricing is researched by checking what other suppliers are listing similar items for in the distributor catalogs, and margins are calculated. For products with multiple components (like plaques that involve metal, tape, board, and other materials), additional component-cost fields are filled in. Once pricing is worked out, the product is built in Ordova, and then manually entered into both distributor catalogs. The auto-sync between Ordova and those platforms isn't working reliably, so both are done by hand, and they have slightly different data formats from each other.

Nate mentioned he planned to spend time with Janet the following morning to understand her process in detail. Sandra confirmed that a full walkthrough of how Janet works would be valuable to the team.

Production Checklists and Labor Data Capture

Sandra spent a significant portion of the call walking Nate through the checklist system she built for the production floor. The background: Sandra described herself as an impatient trainer — she said she has low tolerance for people forgetting steps and needed a way to step back from the floor without things breaking. So she started building detailed Microsoft Word documents listing every step of each production process, which eventually evolved into Microsoft Forms with structured data capture.

The system now has two modes for each process. The full training path walks a new operator through every step in sequence, with instructional detail and a required manager sign-off at setup completion. The trained-operator path collapses the instructional content and shows only the execution steps and data-capture fields. Both paths are on the same form; the trained-operator version just skips the baby steps.

What the forms capture is: a start time, a stop time, the quantity produced, and a notes field for anything unusual during the run. If a second operator joins a job, they submit their own time and quantity — because employees have different labor rates, and knowing who worked on a job for how long is necessary to calculate accurate job costs. The manager sign-off on the training path creates an accountability record: if something goes wrong with a setup, there's a record of who approved it.

The output from all the form submissions feeds into a shared Excel spreadsheet — one per process type. The sublimation spreadsheet, which Sandra showed on screen and emailed to Nate during the call, has accumulated data from all historical runs. Sandra has added calculation columns to the spreadsheet that compute pieces per hour from the submitted times and quantities. The spreadsheet is large — Nate estimated around 800 columns — because every checklist field generates its own column, even when trained operators aren't filling most of them.

The goal of all this data is to arrive at a reliable, per-product-type figure: how many pieces per hour can Merrick produce on a given product? That number is what drives accurate pricing on larger quantity quotes. Right now, getting that answer requires someone to open the spreadsheet and ask Copilot or Claude to analyze it. This works reasonably well, but Sandra noted it doesn't update on its own — every time she wants current numbers, she has to manually request an analysis. “I wish this would update live rather than asking AI to analyze each time,” she said.

Nate and Sandra talked through what a more complete version of this system might look like. Nate suggested a reference document organized by product type — a master list that shows the confirmed pieces-per-hour rate for each product once enough runs have accumulated to trust the number, with a confidence rating (something like red for still-gathering data, yellow for enough to estimate, green for solid). Once a product is rated solid, operators working on it would no longer need to track their time and quantity — they'd just be expected to hit the known production rate. Sandra responded directly: “The goal is to get to the point where we do have that solid data point and we go to those operators and we say, ‘You're expected to do this.’”

During the call, Sandra shared her screen to show the sublimation spreadsheet, and then sent Nate a copy via email.

Email Templates

Nate brought up the email template workflow, which Calloway had described during Day 1. The current state: roughly 30 email templates are stored in SharePoint. When a relevant email comes in, the person handling it reads it, decides whether a template applies, searches the SharePoint folder for the right one, copies it into a draft, customizes it with the relevant details, and sends it.

Nate described a version of this where AI assistance is built into the process. The idea: an AI reviews incoming emails, identifies which ones are candidates for a templated response, and pre-creates draft replies with the appropriate template already populated. The person handling email never loses control — they review and send, or edit as needed — but they start the process already at the draft stage instead of beginning from scratch.

Sandra's immediate reaction was to ask whether the same approach could apply to her accounting inbox. She said this somewhat in jest but clearly meant it. Nate suggested a potential path: build a working prototype in Claude first without worrying about the Microsoft ecosystem, demonstrate that the concept works and that the responses are accurate, and then look at a more integrated build — possibly including multiple inboxes and email types — as a later phase.

Freight Tool

Nate asked Sandra's perspective on the freight calculation tool Calloway had been building in Claude. Sandra's read on it: “It's a fair project, but there may be more important things to have.” She didn't dismiss it but also didn't flag it as high priority. Nate said he'd spend time reviewing what Calloway had already built to figure out whether it could be finished in its current form or whether it needed to be rebuilt more carefully.

Order Entry Coming Back In-House

Nate briefly referenced the order-entry situation that had come up in an earlier call — order entry is coming back in-house from an outsourced vendor. Sandra confirmed that it's largely handled: Merrick is moving a production-floor person into order entry, which addresses the backlog the outsourced arrangement had created and takes some pressure off the production floor without losing overall capacity. The internal solution is already in motion.

AI Tools and Training

The latter part of the conversation shifted to how Sandra and the broader team are using AI today, and what kinds of training or exposure would be most useful.

Current usage: Sandra and Calloway are primarily on Claude (the $20/month subscription). Sandra uses the free ChatGPT occasionally, has tried Perplexity a bit, and experimented briefly with an AI image tool for creative work (wasn't impressed). She described the team overall as still fairly early in working with AI but getting there.

Nate described the AI training session planned for later in the engagement — the goal is not a lecture or a demonstration of tools they don't need. Instead, the session would be reactive and personal: people bring what they're actually working on, Nate looks at how they're using AI, and he helps them do it better. Everyone sees what everyone else is doing, so there's incidental learning across the group. Sandra noted that for some of the team, AI will still feel quite new — the session needs to make people comfortable, not overwhelmed.

Separately from the general team session, Sandra confirmed that she and Calloway are both interested in seeing more advanced AI usage — the kind of back-end system work Nate has built for this engagement itself. She said they talk about it, and when Nate offered to show them that layer in a separate session, the answer was an unqualified yes. Nate shared his screen briefly to walk Sandra through the Rogue Agents website he built using AI-assisted code without writing any code himself — as an illustration of what's possible when you stop fighting for pixel-perfect control and let the tools do the production work.

On image generation and creative work: Sandra mentioned they produce some in-house marketing assets, including banner ads for their distributor-catalog listings. These are currently handled by Carla, their in-house designer. Nate noted that AI image generation has reached a level where standard marketing assets like banner ads can be produced quickly and at high enough quality for distributor platforms. Sandra was interested in this but was candid about Carla: “I don't know if we'll convince her of that. At the start.” She seemed to think it would take a concrete demo rather than a description to bring Carla around.

Day 4 Meeting Logistics

The alignment meeting is scheduled for Tuesday, June 9, 8–10am. The calendar invite is on the books at Merrick's office. Sandra hadn't yet accepted but committed to doing so. She said she'd check with Calloway about whether the team would prefer to hold it in-person or via video call. Nate said he's flexible either way; the main practical difference is that for in-person, he'd be presenting from his laptop rather than his home setup.

Both Nate and Sandra noted that six hours of in-person meetings on Day 1 felt manageable, while a two-hour video call can feel more draining — something worth considering in planning the format.

Follow-Ups

• Sandra: email the sublimation checklist spreadsheet to Nate (she downloaded it during the call but ran into file-sharing issues in the meeting chat).
• Sandra: confirm Day 4 format preference (in-person vs. video) after checking with Calloway.
• Sandra: accept the June 9 calendar invite.
• Nate: send Calloway and Sandra the session summaries from Day 1 meetings (Sandra was excited about receiving these).
• Nate: review Janet's process during Day 2 on-site and report back.