AI understands. Deterministic engines decide.
We don’t hide decisions behind a model. Language AI reads your brief; versioned, reproducible engines do the deciding — so every recommendation traces to a rule you can read.
Natural-language understanding.
Gemini reads a plain-English brief and extracts the product, formulation, packaging, certifications, volume, and geography into structured requirements — the input the engines reason over. Understanding is where the AI belongs.
- Requirement extraction, not guesswork
- You can inspect and edit every field
- Falls back to deterministic parsing — never blocks
Deterministic, versioned, explainable.
The choices that matter — can they make it, how well they fit, whether to recommend — are made by engines, not a black box.
- M5
Compatibility
Sixteen versioned rules confirm a maker can actually make your product — geography, capacity, certifications, process. Every rule is inspectable.
- M6
Scoring
A weighted, versioned model ranks the compatible makers and shows the drivers behind each score. Re-run it and get the same answer.
- M7
Recommendation
Tiers and rationale composed from the trace — a shortlist that arrives already justified, never a bare number.
A shortlist you can interrogate.
This isn't a mock-up. It's the live engine answering a real brief — scoring compatibility, ranking the fit, and showing exactly why. Every number below traces to a rule you can read.
“An FDA-registered gummy supplement manufacturer, MOQ under 10,000”
Replies understood. Deals advanced.
Inbound replies are classified by intent — interested, not, needs-info — and a deterministic workflow advances the engagement, with human approval at the moments that matter.
- Intent classification, thresholded
- One source of truth for every reply
- Human-in-the-loop by design
Smart where it helps. Accountable everywhere.
Re-run any recommendation and get the same answer.