Overview
Quoting a messy lighting tender takes an estimator around four hours of manual catalogue hunting and Excel formatting. This engine reads the tender, matches each line to your catalogue, and flags only the uncertain ones — turning that four-hour rebuild into a focused three-minute review, without sacrificing a single line of accuracy or control.
At a glance
| The problem | An estimator spends ~4 hours per tender cross-referencing specs, chasing part numbers, and reconciling brand variants by hand. |
| What it does | Reads the tender, matches every line to your Delta catalogue, and surfaces only the lines that genuinely need a human. |
| How it stays safe | It never substitutes across fixture categories and never guesses on ambiguous lines — those route to your estimator, with the reason shown. |
| Proven on | A real Delta tender and 64 adversarial stress cases — measured, not claimed. |
See it
A three-minute narrated walkthrough on a real Delta tender — drop the document, watch it tether to the source, match against your catalogue, and export a branded quote.
Prefer to read it? The full flow is in 4.0 How it works.
Try it
The working prototype is live, built on your real catalogue. Drop in a BOQ and watch it work — this is the real engine, not a mock-up.
Open the prototype →Your catalogue, already structured
Real fixtures, lifted from your catalogue and made matchable — a glimpse of the structured data behind every quote.
How it works
A deterministic, guardrailed engine — built for accuracy and trust, not a black box.
- Read the tender. Vision models extract every line item, spec, and quantity, and tether each value to the exact spot on the source document.
- Match to your catalogue. A hybrid of deterministic spec-compatibility rules and semantic search ranks the closest products — the spec differences shown.
- Review only what needs it. Confident matches pass; anything uncertain is flagged with a plain-English reason for your estimator.
- Export the quote. Adjust quantities or margins live, then generate a branded, client-ready PDF.
Three honest outcomes per line
| Auto | High-confidence match — passes through. |
| Review — your call | Ambiguous or below threshold — flagged with the reason; the human decides. The honest flag is the feature. |
| No match | Nothing in the catalogue fits — said plainly, never a guessed substitution. |
What earns the trust
| Category guardrail | Never substitutes across fixture categories — the rule behind 0 unsafe substitutions in 64 stress cases. |
| Accuracy method | Measured on a real gold-standard tender (88%) and a live messy-BOQ harness (0 hallucinated products), not a vendor claim. |
| SAP-ready | Reads pricing/catalogue through an abstraction layer, so a read-only SAP Business One adapter drops in later — not required for v1. |
| Your data, yours | Catalogue and pricing stay isolated per client; nothing written back to SAP until proven. |
Working together
A staged engagement built around how you actually quote — delivered in steps you see and approve, and you own the result.
Free prototype
Proof on your real data, at no cost.
Core build
Full catalogue ingestion, the complete matching engine, human review, branded export.
Phase 2
SAP-ready read layer and added catalogues, when you want it.
How you're protected
| You own everything | Instance, data, configuration, branding, and source — yours at handover. |
| In your name, day one | Cloud accounts are yours from the start; you never rent access. |
| Proven first | Validated on your real data before you commit; money-back if it misses the agreed accuracy. |
| Staged | Nothing at signing; work delivered in stages you see and approve. |
Next step
Share one real BOQ and your monthly quote volume, and we'll size your exact return — then present terms tailored to your numbers.
Open the prototype →