A tomato grower asked Harvie for next week’s irrigation plan. Harvie checked the forecast for that field and ran the water balance. About 32 mm (1.25 in) of effective rain was on the way, against a crop demand of roughly 15–16 mm (0.6 in) for the week. The advice was to skip irrigation, and to keep machinery off the field while the soil was saturated to avoid compaction.

Harvie is Agrio’s AI agent for agriculture. It works from your own fields: the crop, its stage, the local forecast and what you’ve already applied. And it remembers what you tell it, from one conversation to the next. It warns you when disease or weather risk builds, plans irrigation and fertilizer, follows up after treatments, and puts the agreed work on your calendar.

An AI agent that already knows your fields

Ask a general chatbot about your vines and you get textbook advice. Ask Harvie and the answer starts from your vineyard. For every field you’ve set up in Agrio, Harvie knows:

  • What’s growing: crop, variety, planting date and current growth stage
  • The field: location, area and soil type
  • The weather there: the local forecast, plus heat units (GDD) and chill hours accumulated since planting
  • What you’ve done: the irrigation, fertilizer and sprays you’ve logged, and the products in your inventory
  • What you’ve told it: observations, decisions and how you farm, remembered from past conversations
  • What’s coming: disease and weather risks for that crop over the next few days
  • How it looks from above: satellite scans of the field, on supported plans

So when you ask “should I spray this week?”, Harvie already has the crop, its stage, the forecast and the date of your last spray. You don’t have to explain it all again, this week or next season. Harvie is also multilingual, so you can ask in your own language.

Disease risk alerts with an agronomist’s judgment built in

Harvie watches the forecast on each of your fields for weather that favors disease. When risk turns high or severe, Harvie messages you. It tells you which fields are affected, what’s driving the risk (leaf-wetness hours, humid nights, temperature) and what to scout for.

The disease models tell you, hour by hour, when the weather is working for the pathogen. Turning that into a decision takes agronomy, and that’s where Harvie comes in. It reads the risk the way an experienced advisor would: against the crop’s stage, what’s at stake and what’s already been sprayed. You get a recommendation, not just a risk level.

Take a vineyard just after harvest. The weather that week was ideal for powdery mildew and Botrytis, and the models caught it. A simple alert would stop there, and the grower would be filling the sprayer. Harvie went further. The grapes were already harvested, but the remaining foliage still mattered for next season. Harvie noticed that copper had been sprayed just last week, making another fungicide application unnecessary for now. It recommended skipping the spray, avoiding redundant treatment, and continuing to monitor the foliage for signs of disease.

In a dormant peach orchard, the same kind of weather led somewhere else entirely. Brown rot conditions were building, but with no fruit on the trees there was nothing to spray for. So Harvie went after next season’s problem instead: pull the mummified fruit and mark cankered shoots for pruning, before they become the inoculum that hits at bloom.

Weather hazard alerts matched to crop and growth stage

Frost at bloom and frost during dormancy are different problems. Harvie’s weather alerts account for your crop and its current stage, and cover frost, cold, heat, wind, hail, excess rain and sun stress. You hear about a hazard when it’s serious for that crop at that stage, not every time the forecast shifts. And every alert comes with Harvie’s read on what, if anything, to do about it.

Wheat at anthesis is about as heat-sensitive as a crop gets, so a 30 °C forecast rightly triggers an alert. For one grower, Harvie then sized it up. A peak of 30.1 °C (86 °F) with only two hours above 30 °C is a brief spike, not a heat wave, so there was no need for emergency irrigation. The adjustment that mattered was spray timing: move the planned fungicide pass for Septoria and head diseases to the evening, when the product performs better and the flowers aren’t at risk of scorch. Read more about pesticide application and weather.

Growth stage tracking: GDD models checked against the field

Agrio tracks each field’s growth stage with heat-unit models, and with chill accumulation for crops that need it. When a field enters a new stage, Harvie tells you what that stage calls for.

Heat units tell you when a stage is due. The canopy tells you whether it has arrived. Where satellite data is available, Harvie reads both. On one mandarin block, heat accumulation had reached bud break, but NDVI showed no new flush yet. Rather than guess, Harvie gave the grower a quick field check (green tip on several trees) and a plan for either answer:

  • If bud break was confirmed: finish the remaining dormant potassium and phosphorus applications, keep irrigation steady and hold off on aggressive canopy sprays.
  • If the buds were still dormant: continue dormant nutrition and weed control.

When the stage is off, tell Harvie what you’re seeing, for example “we had bud break last Tuesday,” Harvie corrects the stage for that field and remembers it, so later stage calls build on what you saw.

For pest timing that runs from a biofix rather than from planting, ask Harvie to calculate growing degree days from any date, with your own base temperature and an optional upper threshold.

Irrigation and fertilizer planning built on your forecast

Harvie builds a weekly irrigation schedule from the crop, its growth stage, your soil and the forecast, including expected rain. You get the depth per irrigation and, if the field area is recorded, the total volume for the block. As the tomato grower found, the plan can also be “don’t irrigate this week.”

For fertilizer, Harvie estimates nitrogen, phosphate and potash requirements from the crop, its age and growth stage, scaled to the field area. Treat the estimate as a starting point. Check it against your soil analysis and what you’ve already applied this season, then work out with Harvie what goes into the next application.

Satellite crop monitoring: from index map to scouting route

Harvie reads your field’s satellite imagery and tells you what has changed, and where.

The value is in turning a map into a decision about where to walk. When an oil palm grower saw a change on the satellite map, the question was simple: is it just the edges, or the whole block? Harvie answered it, then turned the answer into a scouting plan: where to walk, what to look for, and when it’s fine to wait.

A change on the satellite map is a reason to walk the field, not a diagnosis. Send Harvie a photo or a quick note on what you find, and it takes it from there. It remembers the spot for next time, too.

Harvie remembers your farm

Harvie reads your logged irrigation, fertilizer and sprays, so it can answer “when did we last spray this block, and with what?” It also checks your pesticide inventory, so recommendations start from the products you have on hand.

What you tell Harvie stays with it, too. Mention that the northeast corner stays wet after rain, or that the soil test came back at pH 6.2, and Harvie keeps it with the field. The same goes for how you farm. If you’re certified organic, or you keep a strict pre-harvest interval for an export buyer, you only have to say it once. Months later, when that wet corner starts yellowing, Harvie already knows its history.

When you send photos, Agrio identifies likely diseases and pests. Harvie picks up the conversation with the photos and diagnosis in hand, and points you to Agrio’s treatment pages.

Reminders and follow-ups that close the loop

Once you agree on a plan, ask Harvie to schedule it. The tasks go on your Agrio calendar with the instructions attached. Reminders can be tied to a date or to a degree-day or chill-hour threshold, so a spray follows your pest model rather than the calendar.

After a treatment, Harvie follows up. A week after a zinc application on mango, it asked the grower:

  • Are the newest leaves coming out larger, with less yellowing between the veins?
  • Is there any leaf burn?
  • Exactly which product, rate and number of sprays were used?

It also set expectations: older, deformed leaves won’t recover, so look for improvement on the new flush.

Follow-ups work even when the plan wasn’t a spray. On a rice field with a grass weed problem at 50 days, Harvie and the grower had agreed to cut and pull the large clumps before they set seed, rather than spray the same herbicide group again, and to plan a stronger program for next season. Harvie remembered the plan and came back to ask how much had been cleared, whether any weeds were heading, and whether the rice had reached booting. The answers would decide whether the next priority was more weeding or feeding the crop.

Harvie doesn’t replace your judgment

Risk alerts, stage models and satellite indices tell you where to look. What you find in the field makes the call. Product rates come from the label and your local registrations.

How to get better answers from Harvie

  • Map your fields accurately. Weather, disease risk and satellite data all depend on it.
  • Keep crop, variety and planting date current. Stage tracking and risk alerts depend on them.
  • Log your applications. Harvie can only account for the sprays, fertilizer and irrigation it knows about.
  • Add your inventory, so advice starts from products you have.
  • Correct the stage when it’s off. One message fixes it for that field.
  • Tell Harvie what you see in the field. It remembers, so the next conversation starts from there.