AI Solutions Locations All Services Website Design Perth SEO Perth Google Ads Perth Facebook Ads Perth Content & Video Real Estate Videography Shopify Perth AI Services Perth Case Studies Free Strategy Call
AI Case Study. International Tier 1 Mining Operator

A tender scorer that cites its evidence

Deterministic pricing plus an AI scoring layer that quotes page referenced evidence, calibrated against a real evaluation panel.

Client
International tier 1 mining operator
Industry
Mining
Stages
Code prices, AI scores
Calibration
A real evaluation panel
Proven on
A real tender with five bidders
Scoring Stages
Bidders In The Real Run
100s
Pages Per Bidder
Under $1
Model Spend Per Run
01
Before

The problem

Tender evaluation meant reading hundreds of pages per bidder.

Bills of quantities arrived in inconsistent formats and had to be priced by hand. Scoring criteria were worked through by hand too.

The process was slow, inconsistent between evaluators, and hard to audit after the fact.

Hundreds of pages
Each bidder's submission ran to hundreds of pages, all of it read and scored manually.
Hard to defend
A score with no cited passage behind it is hard to audit and harder to defend.
02
The build

What we built

A two stage scorer, with a hard line between the stages.

Stage one is code. It prices every bidder it can from the bills of quantities, and states honestly why the others cannot be priced. No silent failures, no guessed prices.

Stage two is an AI layer. It scores each criterion, and it must quote the exact passage and page number behind every score. Code verifies every quote against the source documents before anything is shown to a human.

Deterministic pricing
Code prices what can be priced. Where it cannot, it says why instead of guessing.
Cited scoring
Every score must carry the exact passage and page number it rests on.
Verified before shown
Quotes are checked by code against the source documents before anyone sees a score.
03
Under the hood

How it works

Four rules keep the scorer honest.

  • One pinned, vetted evaluation source. Never a folder of drafts.
  • Every scored sentence is checked against the source documents.
  • The scorer quarantines its own output if it detects stale inputs.
  • Run against a real tender with five bidders.
One source of truth
The evaluation source is pinned and vetted once. The scorer never works from whatever happens to be in a folder.
Quarantine over confidence
If the inputs look stale, the scorer pulls its own output rather than publish a score built on the wrong documents.

The scorer was calibrated against a real evaluation panel, on a real tender with five bidders.

The AI layer's total matched the panel's score exactly for one bidder, and came within a point for another.

It also surfaced evidence the panel had not considered. That evidence was presented as additional consideration, never as a panel error.

Model spend for a full run: under a dollar.

Panel match
Exact
The AI layer's total matched the human panel's score exactly for one bidder.
Second bidder
Within 1 pt
For another bidder, the total came within a point of the panel's score.
Model spend
Under $1
The cost of a full scoring run across the tender.
Tender scoring is repeatable, cited, and auditable line by line.