By Nohtal Partansky, Founder & CEO, Sorting Robotics
Nohtal is a former NASA-JPL engineer, and serial entrepreneur building AI-driven automation solutions transforming cannabis manufacturing across North America with scalable industrial robotics.
Table of Contents
For most of cannabis manufacturing’s short history, automation has been treated as an upgrade rather than a foundational decision. It was something to introduce once volume justified it, margins allowed it, or labor became difficult to manage. That mindset made sense in an early market, when production systems were still forming and variability could be absorbed by people without destabilizing the operation. That phase of the industry is over, even if some of the decision-making logic has not yet caught up.
What cannabis manufacturers are confronting now is not a limitation of individual machines, but a limitation of system design. Hand-packed and semi-manual workflows that once felt adaptable are increasingly becoming structural bottlenecks. Adding machines one at a time may improve speed at specific steps, but it does little to address how variability moves through the operation. Judgment still fills the gaps, data remains fragmented, and compliance is still handled downstream instead of being designed into the process itself.
As SKU counts grow, infused products increase in complexity, and regulatory scrutiny tightens, those compromises stop scaling cleanly. They do not fail all at once. Instead, they accumulate quietly until the system becomes fragile under normal operating conditions.
Two Very Reasonable Paths, Two Very Different Outcomes
At the executive level, most automation decisions in cannabis do not feel reckless when they are made. They feel practical. Budgets are finite, headcount can be adjusted, and demand often feels uncertain. In that context, it is understandable why many leadership teams treat automation as something to revisit later, once volume stabilizes or margins widen.
That path usually unfolds the same way. Teams rely on people to absorb variability. Machines are introduced only when bottlenecks become impossible to ignore. Problems are addressed locally, one at a time, as they surface. Each decision makes sense in isolation. Each one solves a real issue at the moment. Over time, the operation stops behaving like a system and starts behaving like a workaround. Some areas get quicker, but the amount of oversight required to keep everything aligned grows in the background.
Other executives make a different choice early on, even under the same constraints. Instead of asking when automation becomes necessary, they start by asking how the operation will behave once SKU counts multiply, infused products expand, and regulatory scrutiny intensifies. Automation, data flow, and quality controls are treated as structural design decisions rather than tactical responses. As volume increases, the operation doesn’t just get bigger, it gets harder to manage, because issues that were once occasional begin to appear every day.
As the industry matures, the divergence between these two approaches becomes increasingly visible. The challenge is not that one group made bad decisions, but that early assumptions no longer hold. What once felt like flexibility begins to restrict scale, while systems designed with growth in mind compound into stability, predictability, and control.
When Automation Stops Being Incremental
In more mature regulated industries, automation decisions are rarely framed as innovation. They are framed as risk containment. Speed is rarely the primary constraint. What matters more is whether a process produces consistent outcomes when conditions are imperfect and pressure is high. Cannabis manufacturing has arrived at that same point, even though automation is still often discussed as a tactical upgrade rather than a structural one.
For teams following the incremental path, this is often where tension starts to surface. Layering isolated machines onto manual workflows creates the appearance of progress while leaving the underlying architecture unchanged. Each handoff continues to depend on operator judgment, and each adjustment requires retraining rather than reconfiguration. Regulatory updates introduce new points of inconsistency instead of being absorbed by the system.
These environments often appear functional on the surface. Orders ship. Output continues. The strain shows up elsewhere, in the form of workarounds, added oversight, and creeping headcount meant to compensate for complexity. The effort required to maintain acceptable performance steadily increases, until relatively minor disruptions begin to carry outsized consequences.
At that stage, automation is no longer about growth. It becomes about preventing instability that has already been designed into the operation.
Why Manual Intervention Becomes an Engineering Problem
Labor has always been central to cannabis production, but scale fundamentally changes the role it plays. At higher volumes, manual intervention introduces risk not because people are unreliable, but because they are being asked to compensate for limitations in system design.
Infused pre-rolls expose this faster than almost any other category. Weight consistency, dose accuracy, labeling requirements, and verification all converge in a single workflow. When those variables are managed primarily through human intervention, performance becomes increasingly sensitive to staffing changes, fatigue, and training gaps. The system can function, but it does not stabilize.
From an engineering standpoint, labor cost is only part of the equation. The deeper issue is variance. Every manual adjustment introduces a range of outcomes rather than a controlled one, which pushes quality assurance into a reactive role and makes compliance dependent on constant vigilance. Production continues, but consistency becomes harder to defend.
Manufacturing sectors with long regulatory histories encountered this problem years ago, when growing complexity made it clear that human intervention could not reliably stabilize outcomes at scale. Cannabis is now encountering the same limit.
Why Systems Thinking Changes the Equation
The most meaningful gains in manufacturing come from coordination rather than isolated improvements. When robotics, vision systems, and AI are designed as a connected system, variability is addressed at its source instead of being managed downstream.
In these environments, data moves across the line rather than stopping at individual machines. Problems don’t wait for a scheduled review to show up. They get noticed while the product is still in motion, which changes how much effort teams have to spend protecting against mistakes later.
This is not about replacing people. It is about building systems that do not rely on constant human correction to remain within acceptable bounds.
Operations designed this way handle change differently. Product changeovers become predictable rather than disruptive. Regulatory updates are implemented through configuration instead of retraining. Output stabilizes because constraints are enforced structurally, not informally. For teams that planned for this early, these advantages compound quietly over time.
Why 2026 Is the Real Decision Point
The importance of 2026 lies less in a single inflection event and more in the capital planning decisions already being made. The facilities and production architectures designed today will define how operators experience complexity for years to come.
Leadership teams are effectively choosing between two assumptions. One assumes that labor will continue to absorb variability and compliance pressure. The other assumes those pressures must be engineered out at the system level. Neither choice feels dramatic in the moment, but the gap between them widens as volume, regulation, and product diversity increase.
Newer markets make this distinction especially clear. Facilities coming online today are often bypassing fully manual phases altogether. The difference is not aggressiveness. It comes from having watched earlier markets take shape and learning where small decisions created long-term constraints.
Once those systems are in place, changing course becomes significantly more complex than most teams anticipate.
Cannabis as a Manufacturing Discipline
Cannabis is no longer developing its manufacturing logic in isolation. It is increasingly aligning with the principles that govern other regulated, high-variability industries.
That alignment does not require abandoning what makes cannabis unique. It requires accepting the same operational constraints and responding with the same discipline. Variability must be controlled at the system level. Compliance must be embedded into process design. Data must inform operations continuously, not retrospectively.
Automation, in this context, is not a signal of ambition or sophistication. It reflects a recognition of how complexity behaves over time.
Manufacturers who see this early are not betting on technology. They are acknowledging inevitability. Those who do not will find themselves managing costs, risks, and constraints that were avoidable earlier, but far harder to unwind later.
By 2026, the question will not be whether robotics and AI belong in cannabis manufacturing. It will be whether the operation was designed with them in mind from the start.
