PULSE · v17.7
Applied irrigation
A trained regression model predicts how much irrigation was applied.
Detailed evaluation available in private diligence materials.
For investors
AI that knows when to irrigate. Your adviser makes the call.
AgWaterAI combines trained models, field measurements and agronomic rules to help California permanent-crop growers and agronomists make block-level irrigation decisions and keep tamper-evident records.
The labeled corpus that turns permanent-crop sensors into a defensible water decision every week.
The problem
California permanent crops cannot skip a season.
Under SGMA, groundwater agencies allocate water and charge for overuse. Almonds, pistachios, citrus and vines are the wedge: a 1,000-acre pistachio operation running 1.5 acre-feet per acre over its allocation in East Kaweah GSA faces $750,000 in a single year, before state fees.
Illustrative example
$750,000
1,500 acre-feet × $500 per acre-foot, AY2026 tier, verified August 2026.
The product
Agronomist-reviewed block-level decisions with a tamper-evident record.
A Monday briefing and a Thursday water order for each block, on the web and in the iPhone app. Crew capture pressure-chamber readings in the app; the agronomist reviews the exceptions and writes the order in pump hours, inches and acre-feet against allocation. No trained model's score enters the order on its own.
Reading summaries are signed into a tamper-evident record; the signature shows whether a summary changed since signing, not that a measurement was correct. Patent-pending. On the farms in our order pilot, the draft order is held back and logged in a hash-linked record before the adviser makes the call, and each issued order is kept with the readings behind it.
Managed · $80 per acre per year
We install and own the sensors and run the field program.
Platform · $20 per acre per year
The grower's own sensors, after we confirm the connection.
The corpus
Daily multi-depth soil data since 2016 on the oldest site.
70 instrumented sensor sites
68 active as of August 30, 2026
11.4 million sensor readings across 320 site-years
Pressure-chamber labels come from our own field crew, dated and paired to the probes. The network includes long-term training sites beyond the commercial deployments.
The models
Trained models in the decision path.
Three trained AI models are live in the irrigation decision today (SENTRY, WATCHDOG, PULSE); more are trained and staged behind promotion gates. Language models help only at the edges, in briefings and Scout, never in the irrigation decision.
SENTRY · v54
Same-day plant water stress
A trained classification model helps the agronomist choose which blocks to check between visits; local chamber history is required.
Detailed evaluation available in private diligence materials.
WATCHDOG · v20
Stress without chamber readings
A trained classification model estimates tree water stress from soil, weather and rootstock data on blocks without recent chamber readings.
Detailed evaluation available in private diligence materials.
PRESCRIBE ships mined rules with trained successors in shadow; SALT FLUSH is trained but gated; a physics planning alpha projects days to the depletion line. No published number is claimed for these.
Validation
Time-forward and transfer results, reported separately.
Each trained model has time-forward results and results on sites it never saw, reported separately and never blended. They are in the private diligence materials. Model outputs on a new farm depend on that farm's data and promotion status.
Research · September 3 to 4, 2026
On a new farm, stress calls come from that farm’s own pressure chamber readings, each judged against its block’s stress line. Models trained on other farms have not shown a head start there.
Research result, not a customer promise. The research and its qualifiers are available in the investor data room.
Commercial · August 2026
Commercial deployments on about 4,000 contracted acres in Kern and Madera counties as of August 2026, subject to execution.
Further acreage discussed verbally depends on funded sensor installation. Cost-share programs can cover part of sensor deployment; recovery is not guaranteed.
A verifiable water record is the intended second product line for districts, buyers and registries: a direction, not a claim of their acceptance. No credits have been issued; no credit volume, price or revenue is established.
The team
Nick McGill
CEO · Certified Crop Adviser #384333
About 22,000 acres under his crop-adviser practice, of which about 4,000 are on AgWaterAI. Consulting practice since 1984; his chamber program produces the labels.
JP Montgomery
CTO
Built the models, pipelines, dashboards and iPhone app. Company founded 2025.
Financing
Series Seed. Delaware C-Corp.
Current financing terms, deployment budget and assumptions are in the investor data room.
Book a 15-minute call
Open the data room · One form, no NDA. Questions and data-room access: info@agwaterai.com.
Not an offer to sell securities. Forward-looking statements are not guarantees. No water or dollar savings are attributed to the models; no crop ROI is claimed; no water credit has been issued.