The People Paradox: Why supply chain AI readiness cannot be measured by training alone
On October 23, 2026, Marina Mayer reported in Supply & Demand Chain Executive on findings from The People Paradox in Supply Chain AI Integration, a joint research initiative by the Association for Supply Chain Management (ASCM) and the Digital Supply Chain Institute (DSCI). The underlying research was released on October 14, 2026. The October 23 publication is coverage of that earlier study, not the date on which the research was first announced.
The findings raise a practical question for supply chain leaders: what does it mean for a planning organization to be ready for AI? The answer cannot be limited to software availability, executive enthusiasm or the number of employees who have completed training. Operational readiness becomes meaningful when people can evaluate recommendations, understand their limitations and make accountable decisions within the planning cycle.
What the ASCM and DSCI research reports
According to the survey findings presented by ASCM and DSCI, 83% of respondents said their employees were unprepared to use AI effectively. At the same time, 91% of organizations reported plans to use AI to augment human workers. The contrast illustrates a gap between investment intentions and the perceived ability of organizations to put AI to work.
The study also reports a difference in AI project outcomes between organizations classified by their workforce readiness. Organizations reporting an AI-ready workforce represented 17% of the surveyed group and reported a 71% project success rate. Organizations with unprepared workforces reported a 28% success rate, while 64% of that group remained at the pilot or planning stage.
These percentages are self-reported survey results. They indicate an association between organizational readiness and reported project success, but they do not establish that readiness alone caused the difference. Organizations that are further advanced may also have greater resources, stronger management practices, better-defined use cases or more favorable starting conditions. The reported success categories should not be treated as independently audited financial or operational results.
Planning leads adoption, but remains far from autonomy
The study assessed AI adoption across 13 core supply chain functions using a five-point scale, with 1 representing no AI use and 5 representing autonomous execution without human review. The reported overall average was 1.94 out of 5.
Planning functions scored higher than many execution functions, with demand planning at 2.28 and supply and inventory planning at 2.12. Reverse logistics was reported at 1.59. These scores suggest that the surveyed organizations were relatively more advanced in applying AI to planning activities, but they do not mean that demand planning had achieved broad autonomy or that its AI recommendations were consistently improving outcomes.
This distinction matters because planning offers many opportunities for analytical support without granting an AI system authority to execute business decisions. Forecast comparisons, exception identification and scenario explanations can be useful even when inventory adjustments, purchase commitments and service promises remain subject to established decision rights.
The relatively low adoption scores also caution against describing the market as having already moved to autonomous supply chain planning. The research points to a landscape in which analysis and recommendation are generally more mature than authorized execution.
Why workforce readiness is a decision capability
For Dataleo, the most important finding is not the percentage of respondents reporting inadequate skills. It is the implication for how organizations design planning work around AI recommendations. A planner is not operationally ready simply because they know how to use a conversational interface or have completed a training module.
A more useful readiness test asks whether planners can identify the assumptions behind a forecast or optimization recommendation, recognize missing evidence, understand the material consequences of accepting it and decide when to override or escalate the result. These abilities connect technical literacy with accountability for real inventory, capacity, purchasing and customer-service decisions.
In demand planning, for example, an AI system may detect an unusual forecast deviation. A trained user may know how to ask the system for an explanation. A decision-ready planner must also determine whether the deviation reflects an actual market change, a promotional event, a data correction, a product transition or an error in the model's interpretation. The distinction affects whether the forecast should be accepted, amended or challenged.
In supply and inventory planning, a recommendation to increase buffers may appear reasonable under a statistical service-level objective. The decision owner still needs to examine supplier reliability, product criticality, working-capital limits and the operational consequences of a shortage. The AI output is one input to that decision, not an automatic transfer of responsibility from the planner to the model.
The leadership gap between strategy and daily execution
The ASCM/DSCI findings also describe an organizational participation gap. According to the reported survey, 82% of C-suite leaders actively led, implemented or scaled AI initiatives, while the corresponding figure at director level was 12%. Many managers and frontline employees were aware of initiatives without being actively involved.
This matters because the people who define an AI investment are not necessarily the people who must change the underlying planning processes. Executive sponsorship can establish priorities and funding, but implementation requires decisions about exception thresholds, workflow ownership, escalation rules, process measures and the authority to accept or reject recommendations.
Readiness should therefore be assessed at the level where decisions actually occur. Operations leaders need evidence that planners, planning managers, sales representatives, procurement teams and finance partners understand their respective roles when AI changes the information available to them.
Learning by doing is valuable, but must be controlled
The research contrasts organizations with more mature AI use against those still preparing or piloting. This supports the interpretation that workforce capabilities develop through practical exposure to real workflows rather than classroom training alone. However, learning through deployment introduces its own trade-off: organizations need enough operational experience to build competence without exposing critical decisions to uncontrolled experimentation.
A sensible progression starts with clearly bounded tasks, known data sources and measurable acceptance criteria. AI can first assist with analysis or exception triage, then move toward recommendations that are independently validated, before any broader execution authority is considered. The appropriate progression depends on the consequences of errors and the reliability of the underlying process.
The goal is not to maximize autonomous action as quickly as possible. It is to establish where automation can be trusted, where human investigation creates value and where the organization must retain explicit approval.
What operations leaders should measure
Training completion and tool usage may be useful implementation indicators, but neither is sufficient to demonstrate operational readiness. Leaders should examine how the planning process changes after AI is introduced.
Relevant evidence may include whether planners correctly identify invalid recommendations, how long important exceptions remain unresolved, whether approved overrides include reasons, whether cross-functional assumptions become more consistent and whether planning decisions are executed within the expected cycle.
Outcome measures should be chosen according to the use case. Demand-planning initiatives may examine forecast bias and accuracy by segment, while inventory applications may focus on service levels, buffer costs and the quality of replenishment decisions. These measures must be interpreted alongside changes in market conditions, product mix and operating constraints.
It is also important to measure the cost of maintaining an AI-supported workflow. Additional review effort, model evaluation, process updates and exception handling can offset apparent productivity gains. The decision is whether the overall operating process performs better, not whether the new tool generates more recommendations.
Dataleo angle: readiness must be demonstrated at the decision point
The strongest lesson from the ASCM/DSCI research is that AI readiness should be defined in terms of decision capability rather than training activity. For supply chain planning, the affected operating model is the process through which forecasts, inventory recommendations and scenarios become authorized actions.
Value depends on planners being able to investigate recommendations, challenge assumptions, document justified overrides and escalate decisions when consequences exceed their authority. The main failure mode is declaring the workforce ready because employees have received training or adopted an AI interface, while the organization has not established who can rely on the output, correct it or accept the resulting risk.
Human accountability is not a temporary obstacle on the way to full automation. It is part of the mechanism that allows automation to operate reliably in a complex supply chain. The appropriate allocation of human and machine responsibilities should change only when evidence shows that the process can support the new level of decision authority.
Evidence, methodology and limitations
This insight interprets survey findings presented by ASCM and DSCI and reported by Supply & Demand Chain Executive on October 23, 2026. The underlying research was released on October 14, 2026. The percentages, success rates and adoption scores are attributed to the study and should not be interpreted as a representative measurement of every supply chain organization or as an audited assessment of project outcomes.
The comparison between AI-ready and unprepared organizations is observational. The report does not establish that training alone generates a 71% success rate, that low adoption scores are necessarily undesirable in every function or that autonomous execution is the correct objective for all planning decisions. Dataleo's proposed measures of decision readiness and operational accountability are editorial analysis, not survey findings.
Sources and editorial credit
Original October 23 coverage: Marina Mayer, Supply Chain AI Gap: 83% of Leaders Face Workforce Unpreparedness Challenge, Supply & Demand Chain Executive.
Underlying research: Association for Supply Chain Management (ASCM) and Digital Supply Chain Institute (DSCI), The People Paradox in Supply Chain AI Integration, released October 14, 2026. The reported research findings are credited to their originating organizations. The analysis of decision authority, planning workflows and operational readiness is Dataleo's independent editorial interpretation.
Illustration credit
The featured image is a thematic business-training workshop stock photograph sourced from Pexels and supplied under its free-use licensing terms. It does not depict ASCM or DSCI survey participants, the researchers, a specific supply chain organization or an independently verified AI implementation. Its use does not imply endorsement by the photographer, Pexels, ASCM, DSCI or Supply & Demand Chain Executive.