AI for agriculture
AI for agriculture
Input costs cut where the data says so, records kept without the kitchen-table paperwork night. 5x ROI in 30 days, or we work free.
- Row-crop producers
- Orchards and vineyards
- Livestock and dairy
- Packers and shippers
- Ag retail and agronomy
Teams we build for
- Hoyes Michalos
- Nurse Next Door
- Fedi
- UBC Sauder
- Merchant House Capital
- Picton Investments
- Campbell Froh May & Rice LLP
- Barnakl
- Hungerford
- Breez
Teams we build for
- Hoyes Michalos
- Nurse Next Door
- Fedi
- UBC Sauder
- Merchant House Capital
- Picton Investments
- Campbell Froh May & Rice LLP
- Barnakl
- Hungerford
- Breez
Teams we build for
- Hoyes Michalos
- Nurse Next Door
- Fedi
- UBC Sauder
- Merchant House Capital
- Picton Investments
- Campbell Froh May & Rice LLP
- Barnakl
- Hungerford
- Breez
Proof, honestly
No published agriculture case study yet, and we say so. Founding clients get priority scheduling, direct founder involvement, and a co-published case study once the numbers are real. The 5x ROI guarantee applies from day one.
We guarantee 5x ROI inside 30 days of deployment, in writing, measured against a baseline you sign before we build. If the system misses the bar, we keep working for free until it clears.
Input costs eat the margin.
Chemical and fertilizer prices climbed, and blanket application pays for weeds you do not have.
Scouting cannot cover the acres.
More than 28,200 farm jobs went unfilled in the 2022 peak season, and employers put $3.5 billion of lost sales on the shortage (CAHRC, 2024, employer-reported estimate). Nobody is left to walk the field, so the problem shows from the truck a week after it started spreading.
Paperwork follows the field work.
Spray records, traceability and program forms, done at the kitchen table after dark.
The weather window and the filing deadline collide.
The seeded-acreage report is due in the same window the sprayer needs. The field wins, and the filing slips.
The bids move while you are in the cab.
The elevator posts, the basis narrows for a day, and the target you meant to hit passes while the combine runs.
The audit lands in the busiest month.
The buyer wants the mock recall and the water tests the week the crop comes off. The evidence exists. Finding it is the job.
Agriculture, before and after
The manual path is dashed: Blanket rates on every acre, Problems found from the truck, Records done after dark. The system path replaces it, and a person approves before anything ships: Variable-rate calls from data, Stress flagged from imagery, Spray records keep themselves.
Before: by hand
- 01Blanket rates on every acrehuman
- 02Problems found from the truckhuman
- 03Records done after darkhuman
After: the system
- 01Variable-rate calls from data
- 02Stress flagged from imagery
- 03Your approvalhuman
- 04Spray records keep themselves
The research
Farm operating expenses rose 5.1% in 2025 while cash receipts rose 4.7%, and realized net income slipped to $8.3 billion (Statistics Canada, 2026, national aggregate).
Statistics Canada · 2026 · National farm income accounts, released May 27, 2026
Run your numbers.
Your operations
Savings use the low end of our hours-reclaimed range. The math is conservative on purpose.
The math
Calculated at the low end of every range.
Every first build is covered in writing: 5x ROI in 30 days. Or we work for free.
The numbers in agriculture
Farm operating expenses rose 5.1% in 2025 while cash receipts rose 4.7%, and realized net income slipped to $8.3 billion (Statistics Canada, 2026, national aggregate).
Statistics Canada · 2026 · National farm income accounts, released May 27, 2026
See & Spray ran across 5 million acres in 2025 and saved more than 31 million gallons of herbicide mix (John Deere, 2025, vendor-reported).
John Deere · 2025
University field trials measured 43.9-90.6% herbicide reduction, averaging 76% (Iowa State / University of Arkansas, 2025).
Iowa State / University of Arkansas · 2025
What we build for agriculture
Watch the acres
Imagery-based crop monitoring.
Satellite and drone imagery flagged for stress and anomalies, so the scout goes where it matters.
Livestock health flagging.
Camera and sensor data watched for the early signs, flagged for the people who know the animals.
Grain marketing and price monitoring.
Bids, basis and your break-evens watched continuously, surfaced when a number is worth acting on.
Call the inputs
Input optimization analysis.
Yield, soil and application data analyzed for variable-rate decisions. Your agronomist makes the call.
Work orders and field tickets.
For ag retailers and custom applicators: the work order becomes the field ticket becomes the invoice, as-applied record attached, before the truck is back in the yard.
Keep the record
Spray and application records.
Application records kept automatically to the standard the regulator expects, signed by the licensed applicator.
Lot traceability for the packhouse.
Field, harvest date, pack date and shipment tied to one lot code, so trace-back is a query and the mock recall takes minutes.
Program and grant paperwork prep.
Sustainability programs, insurance forms and grant applications assembled from records you already keep.
One operations manager, four seasons
She calls the seeding window, not the software
MayhumanSoil temperature at dawn, moisture at seed depth, a forecast worth trusting for three days. She decides the drill rolls today. No system makes that call, and this one does not pretend to.
The record is made while the drill rolls
MayVariety, rate, treated acres and the as-applied map file from the monitor to the field's record before the tender truck is loaded again. Nothing waits for the kitchen table.
Two quarters come back flagged
JulyThe satellite pass shows stress in two quarters that looked fine from the grid road. A scouting proposal opens with the patches pinned, worst first, beside what the same ground did last year.
The agronomist walks the flags and makes the call
JulyhumanFungicide on one field, nothing on the other, and the reasoning goes into the record next to the imagery. The operations manager approves the work order. Spraying is a decision with a name on it.
The market is watched from the combine
SeptemberYield files land by field as the combine unloads, the bins report their temperatures through BinSense, and the elevator bids run against break-evens that update with every load. When a bid clears her target, her phone says so.
She prices a third of the durum from the cab
SeptemberhumanThe alert says the bid cleared her break-even with trucking in. She sells the tonnes she planned to sell, and the contract confirmation files against the field's cost record. The decision was always hers. Now it does not wait for a free evening in November.
The forms draft from the season, not from memory
JanuaryProgram and insurance paperwork drafts from the records the season already kept: seeded acres, applications, yields, storage. Every line cites the record it came from.
She reviews a season in an evening
JanuaryhumanThe forms go out under her signature after she reads them. The winter sit-down with the agronomist starts from five seasons of field records instead of anyone's memory. The calls stayed hers all year. The paperwork stopped following her home.
Who this is built for
The spray record finishes when the boom folds
Product, rate, treated acres, wind and temperature come off the monitor and the weather station at application time. The re-entry and pre-harvest intervals attach themselves to the field, and the licensed applicator signs from the truck. The record is complete when the job is, not at the kitchen table three weeks later.
The mock recall runs in minutes
Every lot code ties the field, the harvest date, the pack date and the shipment together, read straight from the line whether it runs on Famous or Produce Pro. Trace-back is a query. Trace-forward is the same query the other way. The binder the auditor wants is a report from records the season already kept.
Scouting starts where the imagery points
You cover more growers' acres because the passes are ranked before you leave the yard. The stressed patches are pinned, worst first, and your recommendation files to the grower's field record with your reasoning beside the imagery that triggered it. The rec stays yours. It always will.
Every bid lands beside your break-even
Your break-even is computed from your own cost record, by field, with trucking in. The elevator bids and the offers in your Bushel or Combyne feed run against it continuously, and you hear about a price the day it clears your number instead of the week after it did. Selling stays your call. It just stops hiding behind a free evening you never get.
What stays human
- She calls the seeding window, not the software
- The agronomist walks the flags and makes the call
- She prices a third of the durum from the cab
- She reviews a season in an evening
The steps the day below leaves to a person, by design.
The economics
Before and after economics
Line
- Admin load across the operation
Before
32 hours a week at the page defaults: 4 people, 8 manual hours each, $45 loaded cost. $1,440 a week of paperwork, chasing and copy-typing.
After
Records assemble as the work happens. What comes back is measured against a baseline you sign before we build.
- Spray and application records
Before
Reconstructed at the kitchen table from the monitor, the notebook and memory, weeks after the boom folded.
After
Complete at application time, weather attached, signed by the licensed applicator.
- Finding trouble in the crop
Before
Problems found from the truck, a week after they started spreading.
After
Imagery flags the change between passes, and the scout walks the flagged acres first.
- Selling the crop
Before
Bids checked when someone remembers. At harvest, never.
After
Bids watched against your break-evens continuously, surfaced the day one clears your number.
- The buyer audit
Before
A weekend of binders in the busiest month.
After
A report assembled from records the season already kept.
The only dollar figure in this table is arithmetic on this page's calculator defaults: 4 people, 8 manual hours a week each, $45 an hour loaded cost, which is $1,440 a week of manual load. Every other row is qualitative on purpose. There is no published agriculture case study yet, this page says so, and no row here is a client outcome. The calculator above prices your own baseline, and the 5x ROI guarantee is measured against a baseline you sign before we build.
Where the data comes from
Farm mgmt
John Deere Operations Center, Climate FieldView or AgExpert holds the operating truth: field boundaries, as-applied maps, yield files, treated acres and the input inventory. The build reads it through the vendor's export and API lanes and reconciles it with the spray log and the agronomy notes, so one field's season exists in one place for the first time.
Where it stops. Read-only, and rate maps never move themselves. A prescription reaches a drill or a sprayer only when a person loads it after an agronomist called it. Yield and input-cost data are commercially sensitive in a tight land market, so they stay in your accounts, unpooled, and train nothing.
Field imagery
Sentinel-2 passes on schedule, drone flights where resolution matters, a DJI multispectral when a block needs a closer look. The system reads change between passes, flags stress the grid road cannot show, and opens a scouting proposal with the patches pinned so the walk starts where the problem is.
Where it stops. Watching only. A flag opens a scouting proposal, never an application. Field boundaries and imagery archives are your commercial data and stay in your accounts, and no flag becomes a sprayed acre without a licensed applicator's decision in between.
Agronomy records
The spray log in Croptracker or Agworld, soil tests, scouting notes and the shop notebook's successors, indexed by field and by season. This is the corpus: the reason this year's call can see what the same ground did in the last five, and the reason program forms draft from records instead of memory.
Where it stops. The licensed applicator signs every application record, and the record carries what the regulator expects: product, rate, date, weather, re-entry and pre-harvest intervals. Nothing leaves for a regulator, a program or a buyer audit without a person signing it first.
How the system is built for agriculture
See the full capability mapRetrieval
Field records, application logs and agronomy notes indexed by field and by season, so this year's decision can see what the same ground did in the last five. Traceability paperwork assembles from the same records.
- pgvector
- Full-text BM25
- Per-field indexing
Agents and orchestration
Agents run imagery and sensor analysis on a schedule and raise a scouting or application proposal when a field changes. They do not commit inputs. An agronomist decides, and the record of why is kept.
- agent-worker
- Scheduled runs
- Proposal queue
Evaluation
Detection is graded against fields your team has already scouted on the ground. Published results for targeted spraying are strong and vendor-reported, so we measure yours on your ground before anyone changes a program.
- Ground-truth grading
- Precision and recall
- quality-worker
Models
Vision models on imagery for detection and classification. Classic agronomic and weather modelling for the forecasting. Frontier models only for the paperwork and the plain-language summary of what the models found.
- Vision models
- Agronomic forecasting
- Frontier models, one gateway, routed per task
Data boundary
Yield, input cost and field boundary data are commercially sensitive and stay in your accounts. Nothing is pooled into a shared dataset and nothing trains a model.
- Supabase row-level security
- Per-operation scoping
- No-training API terms
Farm mgmt, Field imagery, Agronomy records feed a hybrid index. The agent runtime works from that index, and every consequential action passes a human approval before it reaches Workflow builder, Task board, Cost per outcome.
Your systems
- 01Farm mgmt
- 02Field imagery
- 03Agronomy records
The system
- 01Hybrid index
- 02Agent runtime
- 03Your approvalhuman
Where your team works
- 01Workflow builder
- 02Task board
- 03Cost per outcome
One operations manager, four seasons
Operations manager runs the day through the built system: She calls the seeding window, not the software, The record is made while the drill rolls, Two quarters come back flagged, The agronomist walks the flags and makes the call, The market is watched from the combine, She prices a third of the durum from the cab, The forms draft from the season, not from memory, She reviews a season in an evening. Dashed steps stay with a person.
- 01Operations managerhuman
- 02She calls the seeding window, not the softwarehuman
- 03The record is made while the drill rolls
- 04Two quarters come back flagged
- 05The agronomist walks the flags and makes the callhuman
- 06The market is watched from the combine
- 07She prices a third of the durum from the cabhuman
- 08The forms draft from the season, not from memory
- 09She reviews a season in an eveninghuman
How we build it for agriculture
Get the record straight before the models.
Field records, applications and agronomy notes get indexed per field and per season. Traceability paperwork falls out of the same records for free.
Ground-truth detection on your own ground.
Published targeted-spraying results are strong and mostly vendor-reported. We grade detection against fields your team scouted by hand before anyone changes a program.
Automate the paperwork, not the agronomy.
Compliance and traceability documents assemble themselves. Input decisions stay with an agronomist and the reasoning is recorded.
What we will not automate
Committing inputs. The system raises a scouting or application proposal and a person decides.
Built around your rules
| Regime | What it demands here | How the system complies |
|---|---|---|
| Pest Control Products Act and provincial applicator rules | Products applied per label by or under a licensed applicator, with records of product, rate, date, weather and intervals kept for the period the province sets. | The record assembles at application time from the monitor and the weather data, complete to the checklist. A licensed applicator makes and signs every application decision. |
| Safe Food for Canadians Regulations | Packers and shippers moving food across provincial or national borders need licensing, preventive controls and one-step-forward, one-step-back traceability. | Lot records tie field, harvest date, pack date and shipment together, so trace-forward and trace-back are queries. The preventive control plan stays yours. The paperwork stops being the cost of it. |
| CanadaGAP and buyer audits | Produce buyers demand certification and audit evidence, mock recall included, and the audit lands in the busiest month. | Evidence assembles from records kept all season, and the mock recall drill runs from lot codes in minutes. The certification is yours and we claim none. The audit arrives as a report instead of a scramble. |
| Livestock traceability under the Health of Animals Regulations | Premises identification, animal identifiers and movement reporting for regulated species, on the regulator's clock. | Movement records are kept per event with premises IDs attached, and every submission to the registry is a person's, prepared not sent. |
| PIPEDA | Personal information handled under Canadian privacy law wherever it appears in the operation. | Most farm data is operational rather than personal, and we say so honestly. Where it is personal, crew records, a packhouse customer account, the ag retail counter's contact list, it stays in your accounts, scoped at the database layer, under no-training terms. |
| Farm data ownership | Producers have been burned by platforms that pooled their agronomic data. Yield maps and input costs are competitively sensitive in a tight land market. | Your agronomic data is yours. It stays in your accounts, is never sold, is never pooled, and never trains anyone's models. That is a contract term, not a promise. |
The regimes that govern agriculture, what each demands, and how the system complies
Regime
- Pest Control Products Act and provincial applicator rules
What it demands here
Products applied per label by or under a licensed applicator, with records of product, rate, date, weather and intervals kept for the period the province sets.
How the system complies
The record assembles at application time from the monitor and the weather data, complete to the checklist. A licensed applicator makes and signs every application decision.
- Safe Food for Canadians Regulations
What it demands here
Packers and shippers moving food across provincial or national borders need licensing, preventive controls and one-step-forward, one-step-back traceability.
How the system complies
Lot records tie field, harvest date, pack date and shipment together, so trace-forward and trace-back are queries. The preventive control plan stays yours. The paperwork stops being the cost of it.
- CanadaGAP and buyer audits
What it demands here
Produce buyers demand certification and audit evidence, mock recall included, and the audit lands in the busiest month.
How the system complies
Evidence assembles from records kept all season, and the mock recall drill runs from lot codes in minutes. The certification is yours and we claim none. The audit arrives as a report instead of a scramble.
- Livestock traceability under the Health of Animals Regulations
What it demands here
Premises identification, animal identifiers and movement reporting for regulated species, on the regulator's clock.
How the system complies
Movement records are kept per event with premises IDs attached, and every submission to the registry is a person's, prepared not sent.
- PIPEDA
What it demands here
Personal information handled under Canadian privacy law wherever it appears in the operation.
How the system complies
Most farm data is operational rather than personal, and we say so honestly. Where it is personal, crew records, a packhouse customer account, the ag retail counter's contact list, it stays in your accounts, scoped at the database layer, under no-training terms.
- Farm data ownership
What it demands here
Producers have been burned by platforms that pooled their agronomic data. Yield maps and input costs are competitively sensitive in a tight land market.
How the system complies
Your agronomic data is yours. It stays in your accounts, is never sold, is never pooled, and never trains anyone's models. That is a contract term, not a promise.
Your agronomic data stays in your accounts, is never sold or used to train models, and the records the regulators want are kept automatically.
The objections
The shop drawer already holds three ag tech subscriptions nobody opened after June.
Those tools died because they demanded data entry in the months you have no hours. This build reads what your equipment and your records already produce: the monitor's files, the spray log, the imagery passes. Nobody types anything in a cab in July. If a system needs the crew to feed it before it pays, it is designed backwards, and we do not build it that way.
We already pay an agronomist and an equipment subscription. What is left to build?
The subscription makes maps. Nobody has time to read maps in July. The build reads what your subscriptions already produce, turns it into flags, records and drafts, and puts the agronomist's name on every input call. What is left is the part nobody sells you: the paperwork, the watching and the chasing between the tools you already own.
May has no room for an IT project. Neither does September.
Agreed, which is why the build season is winter. Founding engagements are scheduled so the indexing and the first records run land before seeding, and nothing about the system changes mid-season without your say. The one thing we ask for in-season is what you already do: farm it.
Margins are cents a bushel. We cannot carry a failed software bet.
You are not asked to carry it. Engagements are fixed scope, quoted after a free audit, and the 5x ROI guarantee holds the risk: if the system misses the bar, we keep working free until it clears. If the numbers cannot work at your acre count, the audit says so and nothing gets built.
Half the crew is seasonal and lives in a tractor. Adoption dies here.
The crew never sees a dashboard. Their side of the system is what they already do: the monitor logs the pass, a photo goes in from a phone, a text asks one question and takes one answer. The office sees the system. The field sees fewer forms. That split is the design, not a compromise.
What that means in practice
Classic and predictive ML
Not every problem needs a language model. Predicting numbers, churn, demand, fraud risk, is usually solved better, cheaper and more explainably with proven statistical ML.
Where we stop. When the input is language, judgment or unstructured documents, classic ML underperforms and we say so. The discipline runs both ways: if your problem is a prediction problem, you will hear that it does not need an LLM from us before you pay for one.
Vision, documents and speech
AI for eyes and ears: reading documents, watching camera feeds, transcribing calls.
Where we stop. Bespoke computer vision is justified by volume and latency, not by novelty. Below that bar, a vision-language model on demand is cheaper to run and easier to maintain, and we will tell you which side of the bar you are on before anything is built.
MLOps and infrastructure
The plumbing that keeps AI running in production: deployment, scaling, monitoring and retraining, so it does not quietly degrade after launch.
Where we stop. We do not park your system on infrastructure only we understand. If your platform team runs Kubernetes on AWS, we deploy there. The architecture has to survive us leaving; that is the point of it.
Where your team works
Tour the platformWorkflow builder
Describe the workflow. Watch it assemble.
Say what should happen in plain language and the builder assembles the automation on a canvas you can read, run, and change.
Task board
People and agents, working the same board.
Every piece of work lives on one shared board, whether a person or an agent owns it. Handoffs between the two are explicit, so nothing falls in the gap.
Cost per outcome
Every dollar of spend traced to the work behind it.
Outcomes delivered, cost per outcome, value attributed, return on spend. The same measurement the guarantee is settled against, live on one page.
Ask your brain
Ask a question. Get the answer and where it came from.
Every answer cites the document behind it, so you can check the source yourself. When the brain does not have the answer, it says so instead of guessing.
The hard questions
Free · 3-5 days
Know your number in five days.
We map your operations, find the highest-ROI automations, and hand you a ranked plan with the payback math attached. Yours to keep, whoever builds it.
Prefer to talk first? Book 15 minutes with James. No pitch deck.