Workforce gaps keep supply chain AI stuck in pilot mode

AI mostly stays in planning

The report scored AI adoption across 13 core supply chain functions on a scale from 1, meaning no AI use, to 5, meaning autonomous execution without human review. The average was 1.94.

Planning functions scored highest, led by demand planning at 2.28 and supply and inventory planning at 2.12. Execution-heavy functions trailed, and returns and reverse logistics ranked last at 1.59.

The report calls this keeping AI on a “short leash” and says companies doing it are using predictive tools as “little more than digital spreadsheets.” Eshkenazi reached for the same image.

“I think we’re using it much more on the analysis and on the predictions as opposed to the doing,” he said, calling it “more of a glorified spreadsheet.”

He sees room for AI to take on “routine transactional decisions where AI can operate within parameters,” with people keeping the higher-risk calls on customer commitments, sourcing, and major disruptions.

“We need to expand the AI authority, but we need to do so with confidence in the technology, and it’s not there right now,” he said.

Skills and ROI top the barriers

Two barriers tied for the top spot at 33% each: a lack of workforce skills and resources, and an unclear business case or return on investment. Poor data quality followed at 25%. Executive skepticism did not make the list of main hurdles.

Eshkenazi compared AI to enterprise resource planning (ERP) systems, which were also expected to revolutionize supply chains.

“The technology solution can’t improve bad process and it can’t improve bad data,” he said.

Companies should use value stream mapping and data analysis to find and fix problem areas before an ERP project starts, he said.

Headcount reduction is too simple a measure of AI’s return, in his view. The better one is response time: how quickly a company can recognize a disruption, weigh alternatives, and act.

“You can’t point to a system and say it made the decision,” Eshkenazi said. “Ultimately, somebody has to own that decision for the organization.”

Involvement drops below the C-suite

The survey found that 82% of C-suite respondents lead, implement, or scale AI initiatives. At the director level, 12% lead AI efforts directly, and half of directors said they are only aware of them. The aware-only share rises to 51% of managers and 70% of staff.

Those breakdowns rest on small samples. The survey collected 144 responses, and 107 met the criteria for analysis. Only 11 respondents were C-suite executives, and the report labels its results by job level as directional.

Few expect AI to replace workers

Only 7% of respondents predict AI will directly replace workers, while 51% expect it to move people toward higher-value decision-making.

Eshkenazi said that shift carries a cost. Supply chain professionals used to build judgment by running the analysis, building the reports, and working the spreadsheets, which taught them to spot exceptions and ask whether the numbers looked right.

“A lot of those foundational activities and experiences are now being taken over by AI, and we’re moving individuals into positions where they have to make decisions and manage without understanding the fundamentals,” he said.

Technology more powerful than companies are using, paired with workers who can’t critically think through what it produces, is “a very dangerous combination for organizations,” he said.

Cyber concerns outpace safeguards

Of those surveyed, 92% expressed significant concern about cybersecurity exposure through AI-enabled suppliers and trade partners. According to the report, the average organization had adopted 2.3 of six standard cybersecurity practices.

Eshkenazi traced the exposure to the visibility push that followed the pandemic, when companies began connecting suppliers, carriers, and brokers on shared platforms.

“It’s exposing them to significant ransomware and cybersecurity risks in their tier-three and their tier-four vendors, where predominantly they don’t have a lot of resources,” he said.

Food distributors on a conference panel in San Antonio described one way that plays out: a compromised carrier email that led to a fraudulent pickup of a cold-storage load.

Calling a tier-four vendor for inventory and delivery status isn’t an option, he said. “Huge opportunity but also significant risk,” he said.

Start with one process

Eshkenazi’s first recommendation is to pick one high-value process, such as a key customer process, supplier, or route, and involve the people who run it.

“Then establish measurable outcomes,” he said. “First, you have to identify where you’re starting from.”

That work has to reach below the leadership level, he said, so employees see AI as a path to more valuable work.

“Just because it’s a technology solution doesn’t make it autonomous and doesn’t make it better if you don’t have qualified processes, if you don’t have knowledgeable workers, and if you don’t have clean data,” Eshkenazi said.

Why it matters: Vendors are selling AI that can act on freight decisions, but most supply chain organizations still use it for analysis. The ASCM and DSCI survey ties that caution to unready workers, unclear returns, and thin cybersecurity safeguards.