IN Brief:
- Fifty-five per cent of chief supply chain officers are unclear about the return generated by AI investment.
- AI accounts for 67% of supply-chain digital spending among the organisations surveyed.
- Gartner says change resources should be allocated according to project scale, risk, and intended business outcomes.
Gartner has found that 55% of chief supply chain officers remain unclear about the return on their artificial-intelligence investments, even though AI now accounts for 67% of supply-chain digital spending.
The research company argues that the difficulty is increasingly connected with organisational change rather than technology selection alone. AI use cases are multiplying faster than many businesses can support them with coherent adoption plans, leaving leaders to fund several initiatives without a consistent method for deciding where implementation resources should be concentrated.
Gartner surveyed 394 supply-chain professionals from organisations with annual revenue of at least $250 million between November 2025 and February 2026 to examine digital-investment allocation. A separate survey of 135 senior supply-chain leaders, conducted between January and April 2026, assessed AI use cases and return on investment.
The figures expose an uncomfortable gap between spending and confidence. AI has become the dominant destination for digital budgets, yet more than half of the senior leaders responsible for those investments cannot state clearly what return the technology is producing.
Lorraine Gavin, senior principal analyst in Gartner’s Supply Chain practice, said organisations had improved at executing change for individual projects, but the larger challenge was deciding where limited change-management resources should be invested to support the most important business outcomes.
Gartner distinguishes between a change methodology and a change strategy. A methodology sets out the actions used to implement an initiative, including communications, training, journey mapping, and learning sessions.
A strategy determines how those activities should be prioritised across a portfolio, taking account of trade-offs, risk, constraints, and wider organisational objectives. That distinction becomes more important when AI programmes spread across planning, procurement, warehousing, transport, customer service, and manufacturing interfaces.
Each project may have a credible technical case, but the organisation still has finite process experts, data engineers, trainers, operational managers, and executive attention. Treating every use case as equally urgent can spread those resources too thinly for any deployment to mature.
Gartner predicts that, by 2030, organisations which tailor change-management effort to the scale and context of each AI initiative will achieve twice the long-term return of those continuing to use standardised legacy approaches.
The forecast is not a measured result, but it provides a clear test. Benefits should become visible through adoption, operating performance, and sustained use rather than the number of pilots announced or licences purchased.
The findings sit beside earlier warnings that supply-chain AI is moving faster than operating-model redesign. Gartner has separately reported that many organisations are applying AI incrementally to existing tasks instead of rebuilding workflows, while autonomous supply-chain models will require tighter links between technology, governance, and human decision-making.
Return measurement is difficult partly because AI value rarely appears in one financial line. Forecasting tools may reduce inventory, expedite costs, or lost sales; procurement applications may shorten sourcing cycles or identify price leakage; warehouse systems may increase throughput or reduce errors.
Those outcomes occur in different functions and over different periods, making a single project-level savings figure attractive but often misleading. Benefits may also shift between departments, with one function carrying the implementation cost while another receives the financial gain.
Baseline quality is another constraint. An organisation cannot demonstrate improvement reliably when it has not defined process cost, service level, inventory exposure, decision time, or error rates before deployment.
AI programmes frequently begin with a technology demonstration and search for a metric afterwards. That sequence makes it easier to report activity than value and encourages teams to select whichever measure appears most favourable once the project is running.
Gartner recommends establishing a formal AI change strategy, allocating resources towards high-value initiatives, using different execution approaches according to project scale and context, and developing leaders with business, workforce, and risk-management skills.
A proportionate approach should prevent modest tools from being burdened with enterprise-scale governance. An assistant used by a small planning team does not require the same implementation structure as an AI system authorised to alter production plans, supplier orders, or transport bookings.
Rightsizing must work in both directions. Important systems need deeper controls, while limited applications should not be delayed by processes designed for autonomous decision-making.
Organisations also need to distinguish between productivity tools and operational systems. An application that drafts a supplier summary carries different consequences from one that recommends an order quantity or changes a transport allocation.
The second category requires stronger data controls, approval rights, exception handling, audit trails, and accountability because poor output can affect inventory, service, and cost before an employee identifies the problem.
Budget allocation will remain a crude measure until companies connect spending with adoption and operating results. Directing 67% of the digital budget towards AI says little about whether employees trust the outputs, whether the data is accurate, whether recommendations enter operational workflows, or whether benefits survive after the implementation team leaves.
The 55% uncertainty figure suggests that supply-chain leaders are no longer struggling to start AI projects. They are struggling to determine which projects deserve continued investment and which should be redesigned, paused, or stopped.
The portfolio now needs sequencing, ownership, and occasional cancellation — less impressive than announcing another pilot, but considerably closer to how a measurable return is eventually produced.
