Editor’s Note: This article is an answer to the blog that Mike Lessiter posted on February 6, 2026, based on an AI theme about tractor intelligence and when an advanced machine may “dictate” the farm practices used. We asked Randall Reeder, retired ag extension engineer and renowned expert in conservation tillage practices, to weigh in on the subject. Here’s what Reeder had to say …

Farmers already accept that ‘technology’ often makes better, or quicker, decisions than they can do themselves. Take the latest visual/spray technologies, for example. It’s impossible for a person to identify weeds and turn a spray nozzle on/off at the speed of the sprayer, covering 120 feet wide. So, if the technology is accurate, then why not let the technology (AI) "decide" which fields are suitable for no-till — or strip-till.

A few years ago, Randy Raper, now at Oklahoma State Univ., developed a sensor to determine how deep a tillage implement should run. Instant, with the sensor on the front of the tractor, automatically adjusting depth, it had several pre-set depths that the sensor could choose from.

Today, a smart sensor, with AI, could measure many, many soil characteristics. With the right conditions, the sensor could ‘tell’ the implement to remain in the raised position as the tractor goes across the field — no-tillage.

Millions of Acres’ Data Could Tell A Different Story

Forget the tractor for a moment. 

When can a farmer look at a field, study all the collected soil and yield data, and decide whether this field is ready for no-till? Maybe an ATV with ‘sensors’ can determine if the soil is ready for no-till. If it senses a compaction layer in places in the field, the advice could be to sub-soil those parts of the field. After that "corrective" measure is taken care of, then perhaps a light, leveling tillage operation could be ordered for that part of the field. From then on, continuous no-till.

I think no-till research results across the country, and successful no-till farmers, can provide millions of acres and years of data for AI. Apparently about 80% of our major crop producers feel their land is not suitable for continuous no-till. But AI, with access to the no-till research would likely show the opposite: that 80% of the land IS suitable for no-till. 

Scott Shearer of Ohio State Univ. gave an example of determining N rates with AI instead of using the tri-state recommendations. AI can search more than 1 million research articles and data, select for the soil in a location, and give a precise recommendation for that field. Would that rate determined by AI be more accurate than a recommendation that's a general number for farmers in 3 states?

AI & Unemotional Memories

We still need human intelligence. However, the ‘human’ may remember and place incorrect “weight” on the one bad year when he tried no-till (or a certain cover crop, or a lower N rate, no fungicide, etc.) and had a poor crop yield. With only that one "data point" the farmer lacks the understanding that one result could be a fluke, an outlier – and should not be the sole determination for future plans. 

As farmers get accustomed to AI (i.e., ChatGPT), they will ‘ask’ for recommendations. Then using their human intelligence, they may decide to use the AI recommendation on at least part of their land and compare it with their old practice. 

A key for no-till is to get AI to properly consider the first year of no-till. Or only use the first year after a cover crop has been planted (and perhaps other ‘corrective’ practices have been completed). Even the typical 3-year research project seldom gives enough data. The soil that's been plowed for decades needs more than 3 years to recover and regenerate.

Related Content: The Machine Age: What If Your Fully-Aware Tractor Won’t Allow a Tillage Pass?