Blue Star Labs · Case File 04 · Trace

The Buried Fix Problem

Somewhere in your SharePoint is the fix for the defect you're fighting tonight — and it's cheaper to re-solve the problem than to find it. So that's what your team does. Every time.

Ridgeline Instruments is fictional, and the file mess you're about to search is a synthetic re-creation — no real company's documents appear anywhere in this demo.
Scroll to start

Chapter 1 · Playable

9:40 PM. Find the fix. Timer's running.

SMT-1 is throwing solder bridges on HZ-340 — the flagship board. This exact defect was root-caused in 2023: an 8D report, closed and filed. It's in there right now. You have Ridgeline's actual document system, reconstructed click-for-click. Find it.

You're Marcus. The line is down. Somewhere in these folders is the root-cause fix for HZ-340 bridging. It genuinely exists — this is a fair (if cruel) game.

Every folder in there is a pattern we've met in real plants — the three libraries named Quality, the misspelled folder nobody can rename, the "DO NOT DELETE" no one remembers the reason for. If yours is a 15,000-file Z: drive instead of SharePoint: same disease, same audit — the Top-20 scorecard doesn't care where files live; only the scan plumbing changes.

Chapter 2

It's not you. It's eleven years of growth.

Scan the same system read-only and the mess turns out to have a shape. This isn't Ridgeline being sloppy — it's a growing company never having a spare week to stop and sort out what's canon. Recognize it?

How stale is it?

18,400 files by last-modified date. Two-thirds haven't been touched in three years — and a working answer might live in any era.

Who owns it?

Owner of record, all files. The biggest documented owner of Ridgeline's knowledge doesn't work there anymore.

The duplicate clusters

Same job, competing homes. Hover a cluster — nobody decided this; it accreted, one reasonable decision at a time.

3libraries named some flavor of "Quality"
3folders holding preventive-maintenance records
2,300files with FINAL in the name
34%of files owned by people who left

And no — the fix you're about to see is not a purge. Every file in this scan stays exactly where it is: archived and flagged, never deleted. History is evidence.

Chapter 3

The Copilot Trap

"We'll just point AI at it." Here's what happens when you do — watch closely, because this answer looks great:

MarcusWhat's the correct stencil printer setup for HZ-340? We're getting bridging.
AI assistant, pointed at the mess

For the HZ-340 controller board, set squeegee pressure to 11.5 kg with a print speed of 55 mm/s, per the process documentation.

📄 Source: Profile_Sheet_HZ340_2022.xlsx · Quality_2022
Confidently wrong

That's a real document, accurately quoted, with a citation — and it's the superseded 2022 sheet. The 2023 8D changed the squeegee-pressure spec precisely because 11.5 kg was causing the bridging. The citation doesn't make the answer safer. It makes it more convincing.

Copilot isn't broken.

It's retrieving exactly what you gave it. Three libraries named Quality, a superseded sheet that looks current, and no marker anywhere saying which document is canon — no model can tell your truth from your history. AI on top of a mess is just faster access to the mess.

Chapter 4 · Playable

The same question, after the audit

Same documents. Nothing deleted. What changed: a read-only scan, then an engineer — a person — tagged what's canonical, what's superseded, and what was a band-aid. Ask it what Marcus asked:

This demo box answers three canned questions — the real version is built on standard Microsoft Search in your tenant, so it works whether or not you pay for Copilot; and if you do have Copilot, this same curation is what makes it stop citing ghosts. Every stale document stays exactly where it was: archived and flagged, never deleted.

Chapter 5

The scoreboard

The 8D existed for three years. In that time the same defect was "solved" three times: a two-shift trial run ($7,900 — masked it for a week), an operator retraining cycle, and tonight. Total cost of not finding a document Ridgeline already owned: about $12,000 and five weeks — against a fix that takes one PM revision.

Industry math says this isn't a Ridgeline quirk: knowledge workers spend roughly 1.8 hours a day searching for information (McKinsey) — some estimates run higher still (IDC, ~2.5).

Chapter 6

How this works without moving your files

1 · Read-only scan

A scoped Microsoft Graph app registration — we name the mechanism because your IT will ask. Read-only, limited to the libraries you list (quality and engineering, not HR, not payroll) — an app registration your admin creates in your tenant and can revoke any afternoon. Two weeks.

2 · Tag & curate — a person

An engineer reads the candidates and tags them: canonical, superseded, band-aid, duplicate. The model doesn't decide what's true in your plant. An engineer does, and your team signs off. This is the work you're paying for.

3 · Answers layer

The tags ground Microsoft Search (and Copilot, if you have it) in your tenant. Ask a question, get the canon — with the trail of everything it supersedes still attached. Nothing deleted, ever.

Chapter 7

Two weeks to know exactly how findable your knowledge is

$1,500

File Findability Audit — 2 weeks, read-only

The headline deliverable: a Top-20 Critical Documents scorecard — you name the twenty documents your plant cannot run without, we measure whether your team can actually find each one, timed, from a cold start. Plus your own Sprawl Scan dashboard and a 90-minute readout with the fix roadmap. Fixed price, nothing installed, nothing moved.

$4,500

Answers Pilot — 30 days, one domain

Usually quality/CAPA. Scope stated up front so the price means something: one domain, measured by time-to-answer on ten questions your team writes before we start. If the ten questions don't get faster, you have our number.

The questions worth asking us first

"My office manager can clean up SharePoint."

For filing, absolutely. But the expensive failures aren't messy folders — they're a superseded spec that looks current and a root cause filed under a name nobody would ever search. Deciding what's canonical is engineering judgment, and it's the part the audit actually delivers.

"Microsoft will fix this eventually."

Microsoft keeps making search better at finding documents. Nothing in the roadmap decides which of your three quality libraries is the real one — that's not a feature, it's a decision, and it's yours. We make it once, write it down, and every tool you buy afterward inherits it.

"You're not reading our HR files."

Correct — we can't. The scan is scoped to the named libraries in the agreement, the app registration is read-only, and your admin creates it — and can revoke it — in your own tenant. NDAs, HR, payroll: never in scope, and structurally not reachable.

"We don't pay for Copilot."

Good — you don't need it. The answers layer grounds standard Microsoft Search, included in the 365 subscription you already pay for. If you add Copilot later, it gets smarter the day it arrives, because the curation is already there.

What can't your team find?

Every plant has its 8D — the answer that exists, filed where nobody will ever look. Thirty minutes, no deck: name the last question that took your team a day to answer, and we'll tell you whether an audit would have made it a ten-second search.

Book the 30 minutes

If it looks like a fit, the next step is the $1,500 audit — two weeks, read-only, fixed. If your files are actually fine, the scorecard will say so and you'll have proof.

Or just email john@bluestarcentral.com. The rest of the Ridgeline story: Case File 01 · Case File 02 · Case File 03.