01 the business
Food manufacturer
140 staff, 60 active suppliers
Quotes arrive as PDFs and spreadsheets with different units, pack sizes and freight terms. The buyer copies them into a spreadsheet when there is time, and often reorders from the usual supplier when there is not. Minimum order charges and short validity dates get missed.
02 before
The problem
Quotes come back in every layout imaginable, with different units, pack sizes, freight terms and validity dates. Whoever does purchasing copies them into a spreadsheet by hand to compare, and often just picks the usual supplier because there is no time. Hidden costs, like minimum order charges, get missed.
- Comparisons take hours per request.
- Different units make totals hard to compare.
- Delivery reliability is not part of the decision.
03 the build
What the AI does
- Step 1: Collects quotes for a request from the purchasing inbox in Outlook or Gmail, including PDF and spreadsheet attachments.
- Step 2: Pulls out unit prices, pack sizes, freight, lead times, payment terms and expiry dates from each quote.
- Step 3: Converts everything to the same unit and currency so the totals compare like for like.
- Step 4: Shows the quotes side by side in one screen, with each supplier's past delivery record from your ERP next to the price.
- Step 5: Highlights missing items, unusual terms and quotes about to expire, and drafts follow-up questions to suppliers.
- Step 6: Creates a draft purchase order in NetSuite, SAP Business One or Xero for the option your buyer picks.
paused: waiting for a person to approve
How it was put together
- Collected quotes for each request from the purchasing inbox.
- Extracted prices, pack sizes, freight, lead times, payment terms and expiry dates.
- Converted everything to the same unit and currency.
- Showed quotes side by side with each supplier's on-time delivery record from the ERP.
04 human in the loop
Where a person stays in charge
Your buyer chooses the supplier and approves every purchase order. Any follow-up email to a supplier is shown as a draft for them to send.
- The buyer chooses the supplier and approves every purchase order.
- Follow-up emails to suppliers are drafts for the buyer to send.
- Extracted figures link back to the page of the quote they came from.
05 rollout
From kickoff to live in 5 weeks
- Weeks 1–2Gather past quotes and agree the comparison view with the buyer.
- Weeks 3–4Run on live requests alongside the spreadsheet.
- Week 5Retire the spreadsheet for standard requests.
06 after
What changes
Buyers compare quotes on one page instead of rebuilding a spreadsheet for every purchase. Decisions take total cost and delivery history into account, not just the headline price.
Assume 50 quote requests a month with three quotes each, taking 90 minutes to compare by hand and 20 minutes in one view.
50 × 90 minutes = 75 hours a month today. 50 × 20 minutes = about 17 hours. Roughly 58 hours a month saved.
What to measure
- Buyer time per quote request.
- Share of requests with three or more quotes compared.
- Hidden costs caught before ordering.
- Price variance against the best quote.
07 evidence
What the research says
In a field experiment with 758 BCG consultants, those using GPT-4 on tasks within its abilities finished 12.2% more tasks, 25.1% faster and at over 40% higher quality, but did worse on a task outside its abilities.
Gartner research put the average cost of poor data quality at $12.9 million a year per organisation.
Gartner, Data quality: why it matters and how to achieve it (opens in a new tab) · 2020
08 lessons
Mistakes to avoid
- Comparing price only. Lead time and reliability matter as much.
- Missing expiry dates. A cheap quote that lapsed is not cheap.
- Automating the choice. Supplier relationships need judgement.
09 questions
Common questions
- Can it handle quotes in other currencies?
- Yes. It converts using the rate and date you choose, and shows the original figures too.
- Does it send requests for quotes?
- It can draft them from a template for the buyer to send.