What Future for Supply Chain Knowledge Workers?

In the confusing landscape of rapidly evolving supply chain technology, Joe Shamir, the forward-thinking co-founder of ToolsGroup, long championed the idea of elevating planners from a position “in the loop” to a position “on the loop”. They would be observers rather than actors, sitting above the process, vigilant for any deviations from the control limits.

This, in his vision, represents the logical culmination of automating supply chain planning and was central to the “lights-out factories of the future” - a concept conceived in the 1990's to describe fully autonomous manufacturing plants which needed no light because they needed no humans.

The rate of change has not slowed down. Technological evolution is currently going through a revolution, fuelled by AI and, for supply chain in particular, Machine Learning. The media is awash with confusing, unsettling visions of how the future may look.

When it comes to the future of knowledge work (in our context supply chain planning) there is no doubt that technology will take on more and more human tasks in their entirety, but equally, a growing number of tasks will be shared between machines and humans.

In his article, "Decoding the Jagged Frontier of AI," Dan Martines describes a framework from Harvard Business School for classifying tasks. Tasks better done by humans and tasks better done by machines are separated by a “jagged frontier”. Tasks within the frontier are clearly done better by machines and those outside, better done by humans.

Outside the frontier, Martines discusses whether human-machine collaboration improves performance further (the answer from the BCG study he references is, not always). He goes on to describe a second framework of Centaurs and Cyborgs - Centaurs, where work is divided between humans and machines, and Cyborgs, where humans and machines seamlessly interchange roles, making it challenging to distinguish if the machine is working for the human or the human is working for the machine.

While this may be daunting for knowledge workers, it echoes a familiar concept for production workers. We might feel more comfortable in the role of a farmer driving a tractor, than a line attendant replenishing materials at the pace of the packing machine.

These emerging knowledge working models, the cyborg and the centaur, align with Joe Shamir's historical supply planner models—the cyborg within the loop and the centaur above the loop.

In this evolutionary process, the first consideration is whether the task lies within the jagged frontier. If it does, and once processes, technology, and data are robust, along with sensitive and fair redeployment of the humans affected, the operation can be entrusted to machines, with humans positioned clearly above the loop, monitoring, controlling, and intervening as exceptions arise. (After facing early resistance, the machines are now doing a great job, checking tickets at the gates of the London Underground, hardly noticed.)

Beyond the jagged frontier, the future model is far less clear, and clouded by social and ethical questions. How should we collaborate with our new artificially intelligent colleagues?

Right now, whether the planner is in the loop or on the loop depends largely on the level of autonomy and complexity of the system that executes the plan. For example, in a supply chain planning system that uses AI, the planner may be in the loop if the AI is not fully reliable or trustworthy, and the planner needs to verify or override the AI’s decisions. The planner may be on the loop if the AI is more reliable or trustworthy, and the planner only needs to supervise or audit the AI’s decisions.

The planner being in the loop or on the loop also depends on the preferences and skills of the planner. Some planners may prefer to be more hands-on and proactive, while others may prefer to be more hands-off and reactive. Some planners may have more expertise and experience, while others may have less.

According to the BCG study, centaurs and cyborgs will benefit from AI in different ways.?Centaurs tend to perform better on complex tasks that require creativity and judgment, while cyborgs tend to excel on simpler tasks that require speed and accuracy.

For now, at least, the planners themselves, their skills and personalities will help shape the definition of the next human machine collaborations.