SAMPLE: a fictional metal fabrication and building-products shop. The shop and everything about it below, its people and every number included, are fictional. Not a client of Adwaizer.
AI Audit report: Kestrelford Metal Works Ltd. (sample)
Prepared by: Adwaizer AI Consulting Inc. · Engagement: Adwaizer AI Audit (2 weeks, fixed price CAD 2,100 plus GST/HST for a Canadian client) · Report version: sample, October 2026
What this page is. This is the report a client keeps at the end of the Adwaizer AI Audit, shown for a metal fabrication and building-products shop. Kestrelford Metal Works is invented to show the format and depth; the shop does not exist, and neither do its people or numbers. In a real audit, every figure comes from the client's own records or interviews, and every estimate says what it rests on.
How to read the numbers. Nothing in this report was measured. Figures marked (owner's estimate) are what the fictional general manager or their staff told us in interviews. Figures marked (estimate) are ours and name their assumption next to them. Costs are in Canadian dollars (CAD), before tax.
The area we agreed at kickoff. The audit covers one process area: the office side of an order, from a request for quote to a shipped and invoiced job. That takes in quoting, shop-floor scheduling and the paperwork around each job. Section 11 lists what it leaves out, including the shop's manufacturing execution system (MES), CNC programming and safety systems.
1. Executive summary
Kestrelford Metal Works is a fabrication shop in southwestern Ontario with about 60 people. It makes custom sheet-metal and light structural parts for equipment makers, and makes a small line of building products for contractors: steel stair and railing kits, and flashing and trim. Two estimators answer about 25 requests for quote a week, and about one quote in five is redone because the customer sent a new drawing revision (owner's estimate). A production planner rebuilds the week's schedule by hand whenever a rush job comes in. The office re-keys the same job into the quote sheet, the MES, the paper job packet, the packing slip and the invoice. The general manager's concern is that quotes take three to five business days and the shop loses work it could have won on turnaround.
We mapped 10 workflows and scored 6 AI opportunities. Our recommendation:
- Start with a shipped-versus-invoiced check (opportunity O2) that compares what the MES says was completed and what the packing slip says was shipped with what was invoiced, and lists mismatches for the office manager. It changes no record, uses data the shop already keeps, and gives the office back an estimated 3 hours a week.
- Next, request-for-quote intake with a drawing-revision compare (O1): AI reads the incoming request and its drawings, fills the first lines of the quote sheet, and on a new revision lists what changed since the last quote. The estimator checks every line and sets every price. Estimated 8 estimator hours a week back.
- Then a rush-job impact list (O3) that shows the planner which booked jobs a rush would push, for the planner to decide.
We advise against letting AI set a quote price or a delivery promise on its own, and against any AI step in machine programming or safety (sections 5 and 11). The 90-day plan in section 6 delivers the first opportunity in weeks 1 to 4 and the second in weeks 5 to 10. Estimated cost of the three recommended opportunities: CAD 9,500 to 16,500 one-time plus CAD 1,800 to 4,400 a year to run (section 7).
2. The shop (fictional)
| Item | Detail |
|---|---|
| Name | Kestrelford Metal Works Ltd. (invented) |
| Location | An industrial park in southwestern Ontario |
| Sector | Custom metal fabrication (laser cutting, press-brake forming, welding, powder coating sent out) and a small building-products line sold to contractors |
| People | About 60: general manager, 2 estimators, 1 production planner, 1 shop foreman and 2 lead hands, 1 purchaser, 1 office manager, 1 bookkeeper, 1 shipping and receiving lead, about 49 shop staff on two shifts |
| Volume | About 25 requests for quote a week, about 40 open jobs at any time (owner's estimate) |
| Revenue | Not disclosed for this sample |
| Current tools | Email for requests and drawings (PDF and CAD files); a quoting spreadsheet per estimator; CAD and nesting software in engineering; an MES on the shop floor for job status and labour time; paper job packets (traveller plus printed drawings); Sage 50 for accounting; Microsoft 365 |
| Who we spoke with | General manager, lead estimator, production planner, office manager, shop foreman, shipping and receiving lead (six 30-minute interviews) |
3. Current-state findings
F1. Quotes wait in the estimators' inboxes. Requests arrive by email with drawings attached, sometimes ten or more files per request. The estimator first opens each drawing and copies part numbers, revision, material, thickness and quantities into the quoting spreadsheet, about 20 to 30 minutes per request before any costing starts (owner's estimate). Turnaround is three to five business days (owner's estimate).
F2. Drawing revisions mean starting over. About one quote in five is redone because the customer sends a new revision (owner's estimate). The estimator compares the old and new drawings by eye to find what changed. Twice in the past year a revision that changed only a hole pattern or a bend was missed and the quote went out on the old one (owner's estimate).
F3. The same job is typed five times. A won quote is re-keyed into the MES as a job, printed into a paper job packet, written onto the packing slip at shipping and typed again into Sage 50 for the invoice. The office manager finds 2 to 3 invoice corrections a month, mostly short shipments invoiced in full (owner's estimate).
F4. The schedule lives in the planner's head and a whiteboard. The MES records what each job has done, but the forward schedule is a spreadsheet plus the whiteboard by the laser. Three or four rush jobs a week each trigger a rebuild, about 6 hours a week of the planner's time (owner's estimate). Customers whose jobs slip often find out when they call.
F5. "Where is my order?" calls. The office takes about 20 status calls and emails a week (owner's estimate). Each answer means checking the MES, then walking to the floor or calling the foreman.
F6. Job packets can carry an old drawing. Packets are printed when the job is released. If a revision arrives after release, the office reprints it and someone swaps it on the floor. Once last year a batch was cut to the superseded revision and scrapped (owner's estimate).
F7. No AI policy exists. One estimator already pastes customer specifications into a free consumer chatbot to summarise long requests. That puts customer drawings and terms into a tool the shop has not reviewed, and some customers' purchase terms restrict who may see their drawings.
Readiness (our scoring, 1 = weak, 5 = strong)
| Dimension | Score | Why |
|---|---|---|
| Data | 3 | The MES holds job status and labour time; quotes and revisions sit in spreadsheets and email folders |
| Tools | 3 | MES and Microsoft 365 already paid for; quoting and scheduling live in spreadsheets |
| People | 4 | Estimators want to quote faster; the general manager wants turnaround to win work |
| Process | 3 | A clear path from quote to invoice, but each step is re-keyed by hand |
| Governance | 1 | No AI policy; unreviewed consumer tool in use with customer drawings |
These scores are our judgement from the interviews, not a measurement.
4. Workflow map and scored AI opportunities
4.1 Workflows mapped (10)
| # | Workflow | Hours a week, staff (owner's estimate) | Error risk | Data sensitivity | Fit for AI |
|---|---|---|---|---|---|
| W1 | Request-for-quote intake: reading emails and drawings into the quote sheet | 10 | Medium | Medium (customer drawings) | High, with estimator check |
| W2 | Costing and pricing a quote | 20 | High (margin) | Medium | Low (the estimator's judgement) |
| W3 | Re-quotes after a drawing revision | 6 | High (missed changes) | Medium | High as a compare only |
| W4 | Order entry: won quote into an MES job | 3 | Medium | Low | Medium |
| W5 | Job packet printing and revision control | 3 | High (scrap) | Medium | Medium as a check |
| W6 | Weekly schedule and rush-job rebuilds | 6 | Medium | Low | Medium as a planner aid |
| W7 | Order status answers to customers | 4 | Low | Low | Medium |
| W8 | Material purchasing (sheet, tube, hardware) | 5 | Medium | Low | Low (a reorder list does this) |
| W9 | Shipping documents: packing slips, bills of lading, material certificates | 4 | Medium | Low | Medium |
| W10 | Invoicing and matching to shipped quantities | 4 | High (billing errors) | Medium | High as a checker only |
4.2 Scored opportunities
Scoring: Value (time saved or errors avoided), Effort (how hard to build and run; 5 = easiest), Risk (what could go wrong for customers or for the shop; 5 = safest). Each is 1 to 5; total out of 15. Scores are our judgement, with the reasoning in the last column.
| # | Opportunity | Value | Effort | Risk | Total | Reasoning |
|---|---|---|---|---|---|---|
| O2 | Shipped-versus-invoiced check. Each evening, compare MES completed quantities and packing slips with draft invoices; list mismatches for the office manager. | 4 | 4 | 5 | 13 | Reads only; a person corrects every invoice; catches short shipments before the invoice goes out |
| O1 | Request intake and revision compare. AI reads the request and drawing title blocks, fills part number, revision, material and quantity lines in the quote sheet, and on a new revision lists what changed. The estimator checks every line and sets every price. | 5 | 3 | 4 | 12 | Largest time gain and the turnaround the general manager asked for; drawings are varied, so the build needs testing on past requests |
| O3 | Rush-job impact list. When a rush job is entered, list the booked jobs it would push and by how much, from MES status and the schedule sheet; the planner decides. | 4 | 3 | 4 | 11 | Saves rebuild time and tells the office which customers to call first; the schedule sheet's format is the main build risk |
| O4 | Job packet revision check. Before a packet reaches the floor, and again when a revision arrives, check that its drawings match the latest revision on the order. | 3 | 4 | 4 | 11 | Small build once O1 reads revisions; prevents the scrap in F6 |
| O5 | Status reply drafts. Draft an answer to "where is my order?" from MES status, for the office to check and send. | 3 | 4 | 3 | 10 | Answers F5; depends on O3 for a forward date, and a wrong date reaches the customer |
| O6 | AI that prices quotes or promises delivery dates on its own. | 2 | 2 | 1 | 5 | Price and delivery are commitments the shop must stand behind; they stay with the estimator and the planner |
Top 3 for Kestrelford: O2, O1 and O3, in that order. O4 rides along with O1 once the revision compare works.
4.3 What could go wrong, per top opportunity
| # | Main risk | How we would contain it |
|---|---|---|
| O2 | The check misses a mismatch, and the office stops its own review | It runs beside the office manager's monthly review for the first 6 weeks; both lists are compared before the manual review is dropped |
| O1 | A misread drawing puts the wrong material or quantity in a quote | The estimator checks every filled line against the drawing before costing; AI never sets a price; the intake is tested on 30 past requests before go-live, and customer drawings stay in reviewed tools |
| O3 | The list misses a job, or a promise is made from it without the planner | The list reads live MES status and the schedule sheet; the planner approves every change; no date goes to a customer from the list alone |
5. Do not do yet
| Idea | Why not now | Revisit when |
|---|---|---|
| AI setting quote prices or margins (O6) | The price is a commitment the shop must honour; it rests on the estimator's knowledge of the floor | Not recommended |
| AI promising delivery dates to customers | A date reaches the customer before the planner has checked capacity | O3 has run for a full quarter and the planner trusts its lists |
| Replacing the quoting spreadsheets with a new quoting system | A larger change than an AI step; the estimators' cost tables would have to move first | O1 has run for six months and the shop has chosen a system |
| AI writing material certificates or changing them | Certificates come from the mill and the supplier; the shop passes them on as received | Not recommended |
6. 90-day roadmap
| Weeks | Work | Owner (Kestrelford) | Measure of success |
|---|---|---|---|
| 1 | Write a one-page AI use policy; move the estimator off the free consumer chatbot to a reviewed business tool; check which customers' terms restrict their drawings | General manager | Policy signed by office and engineering staff |
| 1-2 | O2 design: export a month of MES completions, packing slips and invoices; agree what counts as a mismatch | Office manager, bookkeeper | Mismatch rules signed off |
| 3-4 | O2 live beside the monthly invoice review | Office manager | Mismatches found before invoices go out, tracked against the monthly review |
| 5-6 | O1 design: agree the quote-sheet lines AI fills; test intake and revision compare on 30 past requests | Lead estimator | Test results signed off |
| 7-10 | O1 pilot with both estimators; every filled line checked before costing | Lead estimator | 80% of filled lines need no correction (target, not a forecast) |
| 11-12 | O3 design: schedule sheet export, rush rules, planner approval step (build starts after day 90) | Production planner | Rules written and agreed |
Measures are set by the client in week 1. We do not promise a saving; we agree how it will be measured before work starts.
7. Costs (CAD, estimates)
All figures are estimates for this fictional shop. A real proposal replaces them with vendor quotes and a fixed Adwaizer price.
| Item | One-time | Yearly running | Assumption |
|---|---|---|---|
| AI Audit (this report) | CAD 2,100 plus GST/HST, fixed | none | Adwaizer's published audit price for a Canadian client (PROPOSAL-RULES §6) |
| O2 shipped-versus-invoiced check | 2,500 to 4,500 | 300 to 900 | 2 to 4 days of build at a small-firm consulting rate; running cost assumes model usage on about 2,000 shipments a year |
| O1 request intake and revision compare | 4,000 to 7,000 | 900 to 2,000 | 4 to 7 days including tests on past requests; running cost assumes a business AI subscription for the estimators plus document reading on about 1,300 requests a year |
| O3 rush-job impact list (build after day 90) | 3,000 to 5,000 | 600 to 1,500 | 3 to 5 days; running cost assumes a shared board or messaging tool for the planner's approvals |
| Total, three opportunities | 9,500 to 16,500 | 1,800 to 4,400 | Sum of the rows above; the audit fee is separate |
Value side (estimate). If O1 cuts the 20 to 30 minutes of intake per request to about 5 to 10 minutes of checking, on about 25 requests a week, that is about 8 estimator hours a week (estimate: 25 requests from the owner's estimate; the checking time is our assumption, to be confirmed in the O1 pilot). If O2 replaces most of the office manager's invoice matching with a same-day list, that is about 3 office hours a week (estimate: from the 4 hours in W10). These are the numbers the week-1 baseline will confirm or correct.
8. Tool choices
Kestrelford already pays for its MES, Sage 50 and Microsoft 365, and is not asking to switch. Every opportunity above is designed to work with those tools first.
| Need | Option we would put in front of Kestrelford | Why it fits |
|---|---|---|
| O2 mismatch list and O3 rush-job approvals, if the office wants one board for them | monday.com (work management), as an option next to the MES and spreadsheets | One shared board for mismatches and schedule changes, each with an owner and a due date, without changing the MES |
| A customer request form that collects drawings, revision and quantities in one place | Jotform (Forms), as an option | Requests arrive complete, which makes O1's intake simpler |
| O1 drawing intake and revision compare | A Claude or OpenAI build by Adwaizer, with the shop's quote-sheet format and an estimator check step | Varied drawings and the shop's own sheet need a small build; customer drawings stay in reviewed tools |
We're a monday.com partner, so we may earn a referral fee; the recommendation stands on fit. New to monday.com: https://try.monday.com/q1hi6cfy4tr6. We are experienced with Jotform. Kestrelford can stay on its current spreadsheets and email for all of the above; the partner tools are options, not conditions.
9. What Kestrelford would buy next
| Next step | What it is | Indicative scope |
|---|---|---|
| Build package 1: O2 | Shipped-versus-invoiced check, handed over tested with a one-page runbook | Fixed scope and fixed price, about 1 week (estimate) |
| Build package 2: O1 + O4 | Request intake, revision compare and job packet revision check, with the estimator check step | Fixed scope and fixed price, about 2 weeks (estimate) |
| Build package 3: O3 | Rush-job impact list with planner approval | Fixed scope and fixed price, 2 to 3 weeks (estimate), after packages 1 and 2 |
| Staff training | Two 90-minute sessions on safe AI use and checking AI output, built on the AI use policy from week 1 | Half a day (estimate) |
Each package is quoted separately after this audit. Kestrelford can also take this report and build any of it with its own team or another provider.
10. How the work would be delivered
- Certified delivery: PMP and PMI-ACP certified; fixed scope and fixed price.
- Weekly progress: small deliveries every week, each shown in a live demo.
- No surprises: a risk and issue log shared with you, and written acceptance criteria you sign off.
11. Not covered by this audit
The audit looked at the office side of an order only (the area agreed at kickoff). It did not review:
- the MES itself or a replacement for it; the opportunities above only read from it;
- CNC programming and machine control (nesting included);
- safety systems, guarding, welding procedures or any health and safety duty;
- engineering design work and the drawings themselves.
Any of these would be a separate engagement, scoped on its own.
Sample report. Kestrelford Metal Works is fictional; any likeness to a real shop or business is a coincidence. Adwaizer AI Consulting Inc. promises no savings or results; every figure above is an estimate or an owner's estimate as labelled.