Work our team has delivered · WHOLESALE · HORTICULTURE
A procurement agent that won back 58% of identified lost margin for a national wholesaler.
A national horticulture wholesaler was losing margin between sending a quote and the customer saying yes. Our team measured the leak, then built an agent that checks every open quote against live stock. Nine weeks after kickoff it was live, and 58% of the identified leakage is recovered.
↳ Some of this work was delivered by Reenva's founder and team before Reenva was formed. Client name withheld; some commercial figures shown as bands.
/01 · What changed
58% of identified procurement-margin leakage recovered: on the P&L, not in a slide.
Margin used to slip away between a quote going out and a customer saying yes. Now an agent watches every live quote against real stock. It flags the cheaper internal option just before the rep commits.
58%
Of identified external-procurement margin leakage recovered
0
New screens for reps to check: alerts arrive inside the ERP they already use
38%
Reduction in slow-moving inventory cycle time across the network, in the first six months
9wk
From kickoff to the agent live on every quote
/02 · The problem the team was asked to fix
When a customer asked for stock that wasn't on the floor, the rep priced it from an outside supplier. The customer then took a day or three to decide. In that window, the company's own stock often arrived: a branch transfer, a grower delivery or a returned order.
Nobody was watching. So when the customer said yes, the rep confirmed the outside order anyway.
Each of those lines gave away the gap between supplier price and internal cost. Meanwhile, the company's own stock stayed on the floor and aged towards a write-down.
Why the ERP didn't catch it
The ERP held stock on hand, but only as a report someone had to run. It recorded sales; it didn't watch a live quote against incoming stock. A quote raised on Monday against zero stock couldn't update when stock landed on Wednesday.
/03 · How the team worked
The team scoped this from behind the trade counter, not a meeting room.
For the first fortnight, the team’s engineer wrote no code. They stood at the branch counter beside the reps, watching where margin was lost.
- 01 · On the branch floor
The team sat where the decision was made.
The team’s engineer spent two weeks at the trade counter beside the quoting reps. They watched reps price an external line while the company’s own stock aged two aisles away. The blind spot was the timing of a decision nobody could see.
- 02 · Show the cost
A dollar figure on a habit nobody questioned.
The team rebuilt 18 months of quote lines and showed reps and the board what the habit cost. It was not a projection; it was their own orders, replayed. Once reps saw their own quotes in the number, they wanted the fix.
- 03 · Re-engineer the process
Change the decision first, then automate it.
The quote-to-confirm process was redesigned. The supply decision is now re-checked when the customer confirms, not locked in at quoting. The agent enforces that step inside the ERP.
By go-live, the sales team had already changed how it worked. The software made the new process permanent.
/04 · The cost, in numbers
A small percentage on a large number is not a small number.
Until the team ran this calculation across 18 months of orders, nobody could say what the gap was costing.
↳ The first phase did not predict savings. It put the leakage on one page.
Reconstruction · 18 months of orders
- External-procurement quote lines each month
- ~1,400
- Share where the company’s own stock arrived before the order was confirmed
- ~22%
- Average margin gap per leaked line (supplier price vs internal cost)
- 8–14%
- Median deal value on a leaked line
- $3,800
- Monthly leakage before the work started
- $93k–$164k
- Knock-on effect: own stock ageing towards write-down
- $420k+ a quarter
Source: anonymised analysis of quote lines in the company’s order data. Margin bands from supplier price files and the internal cost ledger.
/05 · What the team did
Measure the cost first, fix the process, then build.
- /01
Diagnose.
Quote-line analysis across 18 months of orders. The team matched every external purchase to the stock position when the quote opened and when it was confirmed. Now the blind spot had a dollar value.
Delivered: a quote-to-stock report, the size of the leak and an agreed first build
- /02
Prepare the data.
Stock on hand now streams live from the ERP into Microsoft Fabric. Incoming purchase orders, branch transfers and grower returns feed the same model. For the first time, stock sat in one place, refreshed every 30 seconds.
Delivered: a live stock model, six connected systems and a 30-second refresh
- /03
Build.
An agent watches every open quote line against stock for as long as the quote is live. When the company’s own stock would change the rep’s decision, the rep gets an alert. It arrives inside the ERP, not on a separate dashboard.
Delivered: the agent live on every quote, with no new screens to check
- /04
Run and improve.
A weekly leakage review with the sales team. Alert thresholds were tuned using reps’ feedback on false alarms. Edge cases, such as substitutions and advance bookings, were added one group at a time.
Delivered: a weekly leakage scorecard and ongoing tuning
/06 · Technology the team used
Data layer
- Microsoft Fabric as the single data store
- Live ERP feed: stock on hand, purchase orders, branch transfers
- Six source systems connected
- Shared metric definitions in dbt
What the team built
- Custom AI agent on Azure OpenAI
- Real-time monitoring of open quote lines
- Alerts sent inside the ERP
- Adjustable alert thresholds
Reporting layer
- Power BI leakage dashboard for the sales team
- Weekly leakage scorecard (automated)
- Inventory age report (refreshed hourly)
- Live stock view by branch
/07 · What changed for each team
Sales floor
Reps get an alert inside the ERP when the company’s own stock can fill a live quote. The choice between internal and external supply is now visible at the moment it matters.
Operations
Inventory cycle time for slow-moving stock improved 38% in the first six months. Stock transfers are now planned ahead, not arranged after a problem appears.
Board
Recovered margin, once invisible on the P&L, is now a tracked number. The board can see leakage fall as it happens.
“The work we'd been doing in spreadsheets for ten years suddenly had a line on the P&L. That was the moment the board stopped asking what AI was for.”
/09 · Your next step
Facing the same problem? Find out what fixing it would take.
If there is a fit, a two-week fixed-fee review follows, ending in a clear go or no-go.