Portal Mill Operations

Start here: open Rover Data.xlsx from Graham’s email in Excel desktop with Copilot enabled. Send the simple prompt, reply Go when ready. About 15 minutes to first insight.

You will have in ~15 minutes: a column map, board-type list, scrap ranked by CELL and Board, and top replacement part numbers — all cited to the sheet.

Try it once — scrap and board overview

1. Open Rover Data.xlsx and click Home → Copilot.
2. Copy and send the simple prompt below. Wait for Copilot to reply Ready.
3. Say Go (e.g. Go. The data is on the active sheet.).
4. Compare Copilot’s board and CELL read to a quick pivot or filter you run yourself.

I have Rover CNC / AlphaCAM export data open in this workbook (Rovers A1 and A2).

When I say "Go", analyze the active sheet and tell me:
1. What columns exist and what each appears to mean
2. How many rows and what date range StartDate covers
3. All distinct board types (Board column) and row count per board
4. Average Scrap % by CELL and by Board — flag the top 3 highest scrap combinations
5. For replacement jobs (Split column contains "Replacement"), list the top 5 Part Numbers by QTY with their CELL

Rules:
- Read-only — do not modify the workbook
- Cite sheet name and column headers when stating findings
- Flag outliers instead of treating them as normal
- Do not invent board types, scrap rates, or part numbers not in the file

Do not analyze yet. Reply "Ready — say Go when the Rover file is open." when you understand these rules.

I have Rover CNC / AlphaCAM export data open in this workbook (Rovers A1 and A2). When I say "Go", analyze the active sheet and tell me: 1. What columns exist and what each appears to mean 2. How many rows and what date range StartDate covers 3. All distinct board types (Board column) and row count per board 4. Average Scrap % by CELL and by Board — flag the top 3 highest scrap combinations 5. For replacement jobs (Split column contains "Replacement"), list the top 5 Part Numbers by QTY with their CELL Rules: - Read-only — do not modify the workbook - Cite sheet name and column headers when stating findings - Flag outliers instead of treating them as normal - Do not invent board types, scrap rates, or part numbers not in the file Do not analyze yet. Reply "Ready — say Go when the Rover file is open." when you understand these rules.

Stop and check: Copilot named the right columns; scrap figures trace to the Scrap % column; it did not edit the workbook.

Using the Rover CNC data in this workbook, go deeper:

1. Compare scrap performance across CELL15, CELL17, and CELL23 — which cell drives the most total scrap volume?
2. For the highest-scrap board type, break down Scrap % by whether the job is a replacement (Split contains "Replacement") vs other splits
3. Pick one replacement job (one Job Number) and narrate the route: Part Number, CELL, Split/nest name, QTY, and StartDate — what would you ask the shop floor about this job?
4. List two data-quality issues that could mislead a scrap analysis

Rules:
- Read-only — do not modify the workbook
- Every number must trace to a column in the file
- Mark uncertain interpretations as [Verify with shop floor]
- Do not recommend process changes without stating the evidence is correlational only

Evaluation rubric — Rover CNC

Start here: open May_Operations Scorecard template_FY26_Final.xlsm from Graham’s email. Use tabs Scorecard and Scorecard (By Plant). About 15 minutes to a plant comparison you can use in a monthly review.

You will have in ~15 minutes: Reliability MTD/YTD vs Target for key international plants and one sanity-check headline.

Try it once — plant Reliability comparison

1. Open the Scorecard .xlsm in Excel desktop (enable macros if prompted — Copilot reads values, not VBA).
2. Copy and send the simple prompt. Wait for Ready, then say Go.
3. Spot-check United Kingdom, Dongguan, and HM India rows on Scorecard (By Plant) yourself.

I have the Operations Scorecard workbook open (May_Operations Scorecard template).

When I say "Go", using the Scorecard and Scorecard (By Plant) tabs:
1. List the key metric rows available (Safety, PPM, Reliability, Inventory Turns, Labor, etc.)
2. For Reliability on Scorecard (By Plant): compare MTD and YTD for United Kingdom, Dongguan, and HM India against their Target values
3. Name which plants are below Target on Reliability YTD
4. One plain-language headline a plant leader could sanity-check before a monthly review

Rules:
- Read-only — do not modify formulas or structure
- Cite tab name and row labels when stating values
- If a target or plant name is missing, write [Not in file] — do not invent it
- Do not analyze until I say "Go"

Reply "Ready — say Go when the Scorecard file is open." when you understand these rules.

I have the Operations Scorecard workbook open (May_Operations Scorecard template). When I say "Go", using the Scorecard and Scorecard (By Plant) tabs: 1. List the key metric rows available (Safety, PPM, Reliability, Inventory Turns, Labor, etc.) 2. For Reliability on Scorecard (By Plant): compare MTD and YTD for United Kingdom, Dongguan, and HM India against their Target values 3. Name which plants are below Target on Reliability YTD 4. One plain-language headline a plant leader could sanity-check before a monthly review Rules: - Read-only — do not modify formulas or structure - Cite tab name and row labels when stating values - If a target or plant name is missing, write [Not in file] — do not invent it - Do not analyze until I say "Go" Reply "Ready — say Go when the Scorecard file is open." when you understand these rules.

Stop and check: values match the By Plant tab; plants below Target are named correctly; nothing invented outside the file.

Using the Operations Scorecard workbook, build a cross-plant leadership view:

1. On Scorecard (By Plant), rank the international plants (United Kingdom, Dongguan, HM India, Brazil, Naughtone) on Reliability YTD vs Target — who is furthest from target?
2. On the main Scorecard tab, show the last 3 months of trend for Inventory Turns and Direct Labor % of Production — note direction only, cite month columns
3. Draft a 5-bullet "Monthly ops review" narrative for Graham's network — mix wins, risks, and [Needs input] where the file does not support a claim
4. List which comparisons are hard in this workbook and would need a cleaner data model

Rules:
- Read-only
- Every metric must cite tab + row label
- No invented targets, plant names, or trend claims
- Label anything beyond 12 visible months as [History not in this file]

Evaluation rubric — Operations Scorecard

Two wins in this path: (A) Excel on the UK extract — ~15 min; (B) Outlook or Copilot chat on the weekly Inventory Update email — ~10 min.

Part A — Excel: UK inventory extract

1. Open the extract in Excel desktop with Copilot.
2. Send the simple prompt, wait for Ready, say Go.

I have the UK inventory extract open in this workbook (HM-InventoryExtract).

When I say "Go", analyze the active sheet and tell me:
1. What columns exist and row count
2. Summary by Warehouse: item count and total Qty on Hand
3. Top 10 items by Qty on Hand with Item, Description, Warehouse, and Location
4. Items where Qty on Hand is more than 10× Safety Stock (or Safety Stock is 0 but Qty on Hand > 100) — flag as review candidates
5. One headline about inventory concentration risk

Rules:
- Read-only — do not modify the workbook
- Cite column names; do not invent SKUs or quantities
- Do not analyze until I say "Go"

Reply "Ready — say Go when the inventory extract is open." when you understand these rules.

I have the UK inventory extract open in this workbook (HM-InventoryExtract). When I say "Go", analyze the active sheet and tell me: 1. What columns exist and row count 2. Summary by Warehouse: item count and total Qty on Hand 3. Top 10 items by Qty on Hand with Item, Description, Warehouse, and Location 4. Items where Qty on Hand is more than 10× Safety Stock (or Safety Stock is 0 but Qty on Hand > 100) — flag as review candidates 5. One headline about inventory concentration risk Rules: - Read-only — do not modify the workbook - Cite column names; do not invent SKUs or quantities - Do not analyze until I say "Go" Reply "Ready — say Go when the inventory extract is open." when you understand these rules.

Stop and check: warehouse totals and top items match filters you run manually; no invented SKUs.

Part B — Outlook: weekly Inventory Update

Open Ian’s Inventory Update email from Graham’s thread. Copy the plain-text body (retail vs contract totals, vendor CHANGE table). Paste into Copilot in Outlook or Copilot chat.

When I paste the body of a weekly Inventory Update email in the next message, produce:

1. Executive summary (3 bullets): total retail vs contract movement, direction vs prior week, one risk flag
2. Top 5 vendor changes by absolute £ CHANGE — only from the pasted text
3. Suggested follow-up questions for the inventory team (max 3)

Rules:
- Use only figures and vendor codes from my paste — never invent £ amounts or vendor names
- If a week column is unclear, write [Needs input]
- Keep under 250 words

Do not draft yet. I will paste the email text in my next message. Reply "Ready — waiting for the inventory update." when you understand these rules.

When I paste the body of a weekly Inventory Update email in the next message, produce: 1. Executive summary (3 bullets): total retail vs contract movement, direction vs prior week, one risk flag 2. Top 5 vendor changes by absolute £ CHANGE — only from the pasted text 3. Suggested follow-up questions for the inventory team (max 3) Rules: - Use only figures and vendor codes from my paste — never invent £ amounts or vendor names - If a week column is unclear, write [Needs input] - Keep under 250 words Do not draft yet. I will paste the email text in my next message. Reply "Ready — waiting for the inventory update." when you understand these rules.

Stop and check: every £ figure in the summary appears in the pasted email; vendor names are not hallucinated.

Evaluation rubric — Inventory

Start here: Customer Care receives dealer emails to change order dates in Siteline (ERP). Today that is manual. Copilot can triage — not update the system.

Try it once — triage a dealer request (~15 min)

1. Go to Copilot chat or Copilot in Outlook.
2. Send the simple prompt. Wait for Ready.
3. Paste either a redacted real dealer email from your queue or the practice example below.

When I paste a dealer email requesting an order or delivery date change in the next message, produce a Siteline action checklist:

1. Order / CO number(s) mentioned
2. Line-level changes requested (product, qty, old date → new date)
3. Ambiguities that need human clarification before ERP update
4. Suggested Siteline fields to verify (order header, line dates, hold codes) — generic field names only
5. Draft internal note for the agent handler (not a customer reply)

Rules:
- Extract only from the pasted email — do not invent order numbers or dates
- Mark missing CO/line detail as [Needs clarification]
- Do not claim the ERP was updated — this is triage only
- Flag if the request needs escalation (partial ship, credit, cancellation)

Do not process yet. I will paste the dealer email in my next message. Reply "Ready — waiting for the dealer email." when you understand these rules.

When I paste a dealer email requesting an order or delivery date change in the next message, produce a Siteline action checklist: 1. Order / CO number(s) mentioned 2. Line-level changes requested (product, qty, old date → new date) 3. Ambiguities that need human clarification before ERP update 4. Suggested Siteline fields to verify (order header, line dates, hold codes) — generic field names only 5. Draft internal note for the agent handler (not a customer reply) Rules: - Extract only from the pasted email — do not invent order numbers or dates - Mark missing CO/line detail as [Needs clarification] - Do not claim the ERP was updated — this is triage only - Flag if the request needs escalation (partial ship, credit, cancellation) Do not process yet. I will paste the dealer email in my next message. Reply "Ready — waiting for the dealer email." when you understand these rules.

dealer order change request — practice example (fictional)

From: orders@example-dealer.co.uk
Subject: CS00451234 — delivery date change

"Hi Customer Care,

Please move order CS00451234 (Line 2 — 24× Aeron chairs, graphite) from requested delivery 14 July to 28 July. Customer site closure 10–20 July.

Also confirm whether Line 1 (sit-stand desks) can stay on 14 July.

Thanks,
Alex"

Stop and check: CO number and dates match the email only; ambiguities listed before any Siteline step; output says triage — not “updated ERP.”

Make it reusable — minimal agent (~20 min)

Open Agent Builder on your computer (not on your phone — Agent Builder needs a desktop browser):

1. Go to microsoft365.com/chat .
2. In the left menu, click Agents, then New agent, then Skip to configure.
3. You will work on the Configure tab (fill in the fields), then Try it (test your agent), then click Create when you are ready to save.

1. Name: Dealer Order Change Triage
2. Description: Structures dealer emails into Siteline action checklists for human review.
3. Paste the minimal instructions below.

You triage dealer emails about order and delivery date changes for International Operations Customer Care.

Input: pasted dealer email text.
Output sections:
- Orders affected (CO numbers and lines)
- Requested changes (table: line, product, qty if stated, old date, new date)
- Clarifications needed before Siteline update
- Internal handler note (max 5 bullets)

Rules:
- Never invent order numbers, dates, quantities, or customer names
- Use [Needs clarification] for ambiguous requests
- Never state that Siteline was updated — human confirms and enters ERP
- Escalate credit, cancellation, or partial-ship requests explicitly
- Professional tone; no customer-facing reply unless asked

On the Configure tab, paste each section from Copilot's output into the matching box:

Evaluation rubric — Dealer triage

Graham flagged production and material planning as a priority but does not yet have a shareable dataset. Use 10–15 minutes in the closing block to answer these scoping questions — capture answers for a follow-up workshop.

Scoping questions

• What decision does planning make weekly that takes the most manual Excel time?
• Which systems hold truth today (MRP, ERP, plant spreadsheets) — and where do they disagree?
• What one export could Matt or the planning team pull that mirrors the Rover / Scorecard pattern?
• Which metrics would Graham trust Copilot to summarize vs which need a formal planner sign-off?
• Who owns the follow-up session once sample data exists?