AI Hope, Hype and Hazards

Some topics skew more neutral in our everyday discourse. Do you prefer mountain or beach vacations? Should the toilet paper roll dispense over or under in a Swiss bathroom?

One decidedly non-neutral topic is artificial intelligence (AI), based on its outsized relationship with individuals, businesses and society. The positive and negative corner cases for AI’s impact are extreme, and its adoption has been faster than any prior modern technology, including television, computers, the internet or smartphones.

Here are some perspectives on AI that have caught my attention, over the past few months, in terms of their relevance for supply chain leaders.

I’ll start with the positive end of the spectrum. I have been on calls with dozens of supply chain leaders implementing AI and generative AI (GenAI) solutions that led to positive business cases. In a recent survey with our global CSCO community on AI and GenAI adoption, nearly 60% of these large enterprise leaders had moved beyond pilots toward scaled solutions.

One industrial manufacturing CSCO noted that, despite the use of AI being more experimental than traditional technologies, his team was seeing a positive return on one out of every two implementations. In this case, the winners were paying back more than five times the level of investment, yielding a solid return for the overall portfolio.

A consumer product company is using AI and GenAI solutions in a variety of contexts (e.g., forecast improvement, manufacturing quality, etc.) and has democratized their use by making the tools and associated training available to the larger supply chain organization.

This team has also established a governance model that identifies the best ideas emerging locally and then scales them, with additional resources, across the broader organization. Employees are getting creative, even pointing GenAI inward to code data clean-up routines and summarize useful prompts that users might ask based on the data available in the large language model.

More recently, I’ve heard about AI being used in the context of navigating uncertain trade policies. Advanced users are accelerating strategic decision-making, particularly the ability to run what-ifs on the optimal product code classification or mix of sourcing locations for different commodities depending on which way tariffs land.

I debated on which side of the hope-hype line to place agentic AI, and you can see where I landed. As reference, AI agents are autonomous or semiautonomous software entities that use AI techniques to perceive, make decisions, take actions and achieve goals in their digital or physical environments.

More basic forms of generative AI are still in a state of lower agency within most companies. Put another way, most of you wouldn’t trust an AI agent to make an important customer decision or to communicate with them in an unsupervised way, unless it was operating within a tightly determined set of outcomes.

Future opportunities aside, the hype is naturally coming from the solutions market, as providers pilot AI agents at more advanced companies. The real question is whether companies with lower capability maturity and less developed data and process governance can replicate these use cases.

This hype phenomenon is a common feature of technology life cycles and Gartner has several Hype Cycle reports dedicated to the different solution types and business applications of AI.

Over time, I’d expect to see agentic AI push past the peak of inflated expectations for its ubiquitous use, slide downward toward a trough of disillusionment, and eventually land on a narrower set of use cases that drive true business value and productivity.

Whether your company’s favorite flavor of AI is agentic or more plain-vanilla generative, there are some hazards and pitfalls to navigate when it comes to your people.

An emerging risk relates to the potential for employees’ critical thinking skills to atrophy as they rely more on AI tools over time. It’s a large enough concern that even Microsoft published a report on this subject.

In your current state, AI tools were developed in part by practitioners steeped in the business context and operational rules of your supply chain.

Imagine a time five or more years in the future when a significant portion of your frontline workforce now relies on these tools to do their everyday jobs but would be at a loss to explain why they make the recommendations they do. This is where experiential learning alongside more tenured employees and developing “soft skills” such as structured problem solving become more critical.

Another looming hazard is the risk of accidental decision delegation. Using an autonomous vehicle analogy, it relates to knowing when a human should grab the wheel when the road takes a highly unexpected turn.

We saw an example of this during the pandemic when AI-based forecasting tools stopped generating feasible plans based on significant demand and supply disruptions. At that time, the planning organizations affected were staffed with people who knew how to plan in older, more manual ways.

In the future, we’ll need mechanisms to audit and monitor decisions so that we can “grab the wheel” before the car swerves off the road. And to my earlier point on skills, we’ll need a team that can troubleshoot their way to a temporary fix until driving conditions are more favorable.

Amid all the hope, hype and hazards, one thing is certain: AI is here to stay and will continue to evolve its role from tool to coworker in our organizations.

Stan Aronow
VP Distinguished Advisor
Gartner Supply Chain
Stan.Aronow@gartner.com

Podcast: The GenAI Productivity Paradox

2025 CEO Survey: 4 CSCO Priorities to Shape Supply Chain Strategy in an AI Era (subscription required)

Client Webinar: Overcoming Adoption Challenges With AI (subscription required)

Webinar: GenAI & Productivity: Where to Focus for Maximum Impact