AgriLens — by Apex Intelligence

Precision agriculture you can trust.

In English and French, built on open public data.

Schedule Consultation Watch the tour

66 s

drawn boundary to analysed field

0.89 R²

yield, held-out years

2

languages, the product itself

6

open data sources

What it does

A whole field's per-pixel crop-health map, with the land around it

Field health from space

Sentinel-2 passes become a per-pixel crop-health map — cloudy pixels omitted, never painted over.

Yield forecast broken down into its drivers

Yield forecast, with the reasons

Every forecast comes with a SHAP breakdown of what raised or lowered it, and a 90% interval.

The low-NDVI nitrogen zone: its rate, the calculation behind it and its OMAFRA citation

Nitrogen zones for your sprayer

Zoned rates for a field, every rate cited to OMAFRA Pub 811 — or it abstains.

A question, then the assistant's grounded answer and its first cited source

Agronomy answers with the page cited

Ask in English or French; answers come from indexed OMAFRA guides, cited by section and page.

Built from nothing

One minute from drawn boundary to analysed field.

66 seconds, measured live — fresh account, nothing seeded, cold caches.

Setting up a farm: name and country, with imagery and NDVI confirmed available — Name your farm and draw the boundary on live imagery — area computed as you draw.

Proposed field boundaries on real farmland, to confirm before use — Or let the model propose boundaries — nothing saved until you confirm.

The NDVI card in its first honest state: pending, queued for a Sentinel-2 scene over the new field — The real pipeline starts; every product shows its honest state while it runs.

The field health map, ready — About a minute later: health map, soil profile and yield forecast ready.

Why you can trust it

Four commitments, in the product's own words.

Says so when the guides don't cover it

“No grounded recommendation available — the agronomy corpus didn't cover today's priority directly. The priorities engine above still drives the day.”

In the product: Home — “Today’s grounded recommendation” card

Estimates are labelled estimates

“Modelled estimate (SoilGrids 2.0, ~250 m) — not a soil test of this field.”

In the product: Field detail — Soil card

Refuses photos it can't read

“This doesn't look like one of the corn leaves I was trained on.”

In the product: Disease detection — out-of-distribution decline

Every number shows where it came from

“Numbers come from the database, never from the LLM's memory.”

In the product: Assistant — grounded answer with cited sources

See the whole walkthrough

The full tour, module by module, on real data. 1 min 33 s.

In detail

Fields table that names its sources beside its heading: Sentinel-2 and Modelled — The table names its sources.

Climate trajectory for the nearest projection cell — Climate scenarios, uncertainty preserved

Carbon estimate with its modelled-estimate chip — A carbon estimate that states its limits

The fields table in French, with the same source chips — The Home dashboard, in French

The NDVI card in French — Honesty copy translated, not dropped

Ontario climate cells coloured by projected heat units — Ontario-wide climate to 2100

Built on open public data — Sentinel-2 via Microsoft Planetary Computer, SoilGrids 2.0, Open-Meteo, NASA POWER, CanDCS-U6 and OMAFRA Pub 811, with field boundaries from the open FTW models — each attributed inside the product.

Full attributions

Want this kind of system for your data?

AgriLens is the proof, not the product. If your organisation needs satellite, climate or field intelligence built with this level of honesty, the next one gets built for you.

Schedule Consultation Download the one-pager (PDF)