AI in Cultivation: MJBizCon Panel Stresses Fundamentals First

Artificial intelligence (AI) officially has entered the cannabis grow room.

At MJBizCon 2025 in Las Vegas, a panel about AI in cultivation drew operators, technologists, and curious skeptics looking for clarity on what is real, what is hype, and what might actually move the needle in an oversupplied, margin-squeezed market.

Why environmental control comes before AI

On stage, the technology spanned everything from canopy-scanning robots to tissue-culture automation and leaf-level plant monitoring. But throughout the conversation, Cannatrol founder and Chief Executive Officer Jane Sandelman served as a kind of counterweight, repeatedly reminding the room that no amount of AI can compensate for poor environmental control.

“Our core competency is environmental control,” she said early in the session. “If your HVAC and room conditions aren’t consistent, AI can’t fix that for you. You’ll just end up in a ‘garbage in, garbage out’ situation.”

Her point landed, because the other panelists’ success stories depended on exactly the kind of stability she described.

How data-driven cultivation is rewriting SOPs

Gary Holland, chief innovation officer at Endless Biotech, described a data-heavy approach that would have been unthinkable in the early medical days. His team invested in scientific instrumentation that clips onto individual leaves, measuring what is happening inside the plant almost down to the cellular level and generating 45 different data points per sample.

Those thousands of data points are fed into AI systems that analyze growth across propagation, veg, and flower. Studying how plants respond to different stimuli — light spectra, irrigation patterns, environmental changes — helped the team rewrite the company’s standard operating procedures. The result, he said, was the ability to consistently hit roughly 100 grams of flower per square foot, up from the 25–55 grams per square foot he sees in many of the facilities with which he consults. At the same time, the operation pushed terpene levels from the 2–3-percent range into the 4–6-percent range, with occasional runs approaching 9 percent.

Holland explained AI’s biggest impact is coming through robotics. Working with a company called Bud Scout, his team is bringing an automated robot into vertical racking systems to continually scan the canopy. Rather than relying on two or three wall-mounted sensors, the robot reads environmental and plant data at the level of individual square inches, and it does so for the full life cycle of each plant. Over time, that dataset reveals patterns that would be nearly impossible to keep straight in a human-managed logbook.

He also described a tissue-culture lab where robots, trained using a combination of virtual-reality sessions and human demonstrations, are learning to recognize explants, understand plant polarity, and perform repetitive lab tasks around the clock. The next phase is even more science fiction–adjacent: semi-mobile robots with two arms, designed to move through flower rooms and eventually handle defoliation and trimming.

Why stability matters more than advanced sensors

Against that backdrop, Sandelman’s focus sounded almost deceptively simple.

In her view, the industry’s rush toward AI can succeed only if facilities solve long-standing environmental issues first. It does not matter whether the AI is optimizing irrigation, suggesting setpoints, or issuing commands to a post-harvest system. If the room cannot hold the set conditions, the algorithm is just shouting into the void.

“I think a lot of people think AI is a magic bullet, and it is certainly not,” she said. “If you have an environmental situation — if your HVAC isn’t up to snuff and your rooms are not consistent either in drying or cultivation — you’re going to have problems. The AI can tell your room to be at a certain [vapor-pressure deficit], but if your room cannot achieve that consistently, it’s not going to make things better. It could create more problems.”

Sandelman emphasized that post-harvest is where product value can evaporate if rooms drift in and out of target conditions. In a market where oversupply is common and inventory often sits longer than intended, keeping flower in a stable environment is less a luxury and more a survival strategy.

Environmental discipline, she argued, is also the bridge between cultivation and branding. As regulations evolve toward eventual federal oversight, she expects consumer expectations to shift from fixation on THC percentage to a demand for consistent performance and experience from SKUs. AI and machine learning can help identify the inputs that drive repeatability, but only if the underlying systems do what they are told.

What AI can reveal that growers can’t see

Across the panel, there was broad agreement that AI’s real strength is not in replacing growers, but in seeing patterns they cannot.

Holland shared how AI-driven analytics helped him uncover the root causes of issues like bud rot and powdery mildew. By analyzing temperature swings, vapor-pressure deficit (VPD), and root-zone moisture, AI models highlighted factors such as sudden drops in temperature that pushed rooms to the dew point, leaving leaves and buds wet long enough to fuel botrytis. The same tools pointed to overly aggressive VPD targets that were stressing plants, closing stomata, and creating conditions where powdery mildew spores could take hold.

Building cultivar-specific roadmaps with machine learning

Corey Waggoner, CEO of Higher Yields Consulting, stressed how AI changed the way his team approached “genetic roadmapping.” By correlating plant performance with the number of days in veg, defoliation intensity, transplant timing, substrate choice, and lighting technology, his group is building what he described as cultivar-specific “recipes” that can be adapted across facilities. For multistate operators juggling single-tier high-pressure sodium rooms, four-tier LED builds, soil, rockwool, and everything in between, that level of insight has become essential to any serious attempt at standardization.

Sandelman listened to these examples and consistently brought the crowd back to the precondition for all of them: good data. Without well-maintained rooms, calibrated sensors, and clear targets, she said, AI cannot sort signal from noise.

“Go slow,” she cautioned. “The computer is not going to fix your world. You need to teach the computer — you need to teach the robot — to fix your world. You need to lead the machine, not let the machine lead your business.”

Why AI won’t replace growers anytime soon

The panel did not sidestep the question most employees worry about when they hear the letters “AI”: Will this take my job?

Sandelman’s answer was quick and unequivocal. She does not believe AI will trigger mass layoffs in cultivation facilities. Instead, she sees the technology shifting people away from repetitive tasks and into more valuable, and more interesting, work.

She pointed to the dry room as an example. In traditional workflows, a staff member might spend an entire shift manually burping jars. Automation and better environmental control can take over that job, freeing the employee to work in quality control, data collection, or higher-skill plant work.

Holland offered a longer historical view, comparing today’s anxiety circulating around AI to the fear that greeted early tractors. Blacksmiths and field hands worried mechanization would erase their livelihoods. Instead, agricultural output soared and new categories of work emerged around machinery, logistics, and agronomy. In cannabis, he suggested, AI likely will reduce the need for brute-force labor while creating demand for data-literate head growers, systems integrators, and specialized technicians.

That does not mean adoption will be painless. Capital expenditure remains a hurdle in a tight funding environment, and several panelists acknowledged that operators are understandably skeptical after years of vendors overpromising and underdelivering. There is also cultural resistance: One technologist described winning grants that would have placed dozens of robots in facilities at no cost, only to have operators ask not how much data they would receive, but how much money they would be paid to participate.

A practical roadmap for adopting AI in cultivation

By the end of the session, a loose roadmap for AI in cultivation had emerged.

Panelists urged operators not to “go all in” overnight. Instead, they suggested piloting a single solution in a specific niche, watching how it performs, and then iterating. That might mean trying a canopy-scanning system in one flower room, using AI-driven analytics to refine standard operating procedures in a single facility, or adding smarter controls to a post-harvest space where environmental drift has been a recurring headache.

Moderator David Johnson, chief commercial officer at Chorus, closed by noting one practical entry point is to bring in an expert — whether a consultant focused on AI in cultivation or a vendor solving a well-defined problem — and test one system at a time. That incremental approach echoed Sandelman’s philosophy: Fix the fundamentals, define what success looks like, then layer in automation and intelligence where they actually can deliver.

For now, AI in cannabis cultivation is in its early days. The technology on display at MJBizCon suggested a future of highly instrumented rooms, robotic tissue-culture labs, and genetic roadmaps tailored to each facility. But if the panel made anything clear, it was that the operations most likely to benefit are not the ones that chase every shiny new tool.

Successful operations will do what Sandelman kept emphasizing: Get the house in order first. Then let machines help keep it that way.

---

The Truth About AI in the Grow Room