From Pilots to a Production Program
Unilever and Accenture have announced a partnership to scale AI-enabled digital twins across Unilever's global manufacturing network, with a plan to build more than 40 new twins over the next 18 months as a blueprint for wider rollout. The significance is in the verb. Plenty of manufacturers have built a digital twin or two as a showcase. Committing to 40 new builds on a fixed timeline, with an explicit intent to standardize the approach across the network, is the move from experiment to industrialized program. That transition is where most enterprise AI initiatives either prove their worth or quietly stall.
A digital twin, in this context, is a virtual model of a factory's equipment and production lines that ingests live shop-floor data and pairs it with AI agents for prediction and simulation. The model can forecast where a production line will bottleneck, simulate the effect of a change before it is made on the physical line, and increasingly take on automated adjustments with human oversight. Done well, it turns a factory from a system that reacts to problems into one that anticipates them, which is the long-standing promise of industrial AI that has so often outrun the reality.
The Plant-Level Numbers That Matter
What makes this announcement more than a partnership press release is the specificity of the results Unilever has already seen at individual plants. At its Raeford, North Carolina facility, which produces Dove, Degree, and Axe deodorants, the digital twin predicts 95 percent of process flow restrictions and has delivered a 20 percent reduction in waste and a 10 percent uplift in capacity. Predicting nearly all of the flow restrictions before they bite is the kind of operational visibility that separates a genuine capability from a dashboard, and the waste and capacity figures translate directly into cost and output.
The pattern repeats across geographies. A plant in Poznan, Poland cut stoppages by 20 percent and waste by roughly 30 percent. A facility in Gandhidham, India reduced quality defects by 30 percent over four years. A site in Cu Chi, Vietnam achieved savings of 1 to 2 percent in premium ingredients. None of these figures is individually spectacular, and that is precisely the point. Sustained double-digit improvements in waste, stoppages, and quality across multiple plants are the unglamorous, compounding gains that actually move a manufacturer's economics, as opposed to the one-time demonstrations that generate headlines and little else.
Why Scale Changes the Math
Unilever's sheer scale is what turns these percentages into a strategy worth pursuing. The company's products are used by 3.7 billion consumers daily, it employs 96,000 people, and it reported 50.5 billion euros in sales in 2025. Against numbers that large, a 20 percent waste reduction or a 10 percent capacity uplift on a single line, multiplied across dozens of plants, becomes an enormous absolute figure. The economics of digital twins improve dramatically with scale, because the fixed cost of building the capability is amortized across vast production volumes.
This is the structural reason large incumbents, rather than nimble startups, may capture the most value from industrial AI. The investment required to build, validate, and operate a fleet of digital twins is substantial, and it pays back fastest for organizations with the production volume to absorb it. Unilever's commitment to 40 new builds is a bet that standardizing the approach across its network will let it reuse engineering, share learnings between plants, and drive down the per-twin cost over time. Smaller manufacturers will struggle to justify the same investment, which could widen the operational gap between the largest players and everyone else.
The Human Oversight Question
Unilever has been careful to frame the digital twins as augmenting rather than replacing human operators, describing a model in which industrial AI handles predictive maintenance and can progressively take on automated adjustments while keeping humans in the loop. That phrasing, progressively and under oversight, is doing important work. It signals a gradual transfer of control to the AI as trust is established, rather than a wholesale handover. In manufacturing, where a bad automated decision can damage equipment or injure workers, that caution is appropriate and probably non-negotiable.
The harder question, which this stage of the rollout does not yet answer, is how far the automation eventually goes. Each increment of automated adjustment that proves reliable creates pressure to grant the next one, and the boundary between human-supervised and human-absent operation tends to drift over time. Unilever's challenge will be to keep meaningful human judgment in the loop as the twins become more capable, resisting the efficiency logic that would push toward removing it. How the company manages that drift will be instructive for every manufacturer following the same path.
A Template for Industrial AI
Adam Raeburn-James, Unilever's global vice president of digital business operations, framed the effort in terms that reach beyond cost. "Scaling AI across our operations isn't just a technological shift, it's a commitment to superior products," he said. Nicole van Det, who leads Accenture in the Netherlands and Nordics, noted that "Unilever has long been recognized for its supply chain excellence," positioning the digital twin program as an extension of an existing operational strength rather than a departure from it. The messaging ties AI investment to product quality and operational reputation, not merely to efficiency.
For enterprise technology leaders in manufacturing and adjacent industries, the Unilever program offers a usable template. Start with individual plants, prove concrete operational gains with hard numbers, then commit to scaling the validated approach across the network on a defined timeline. The discipline lies in demanding real metrics, waste, stoppages, defects, capacity, before scaling, and in resisting the temptation to declare victory on the strength of a single showcase. Unilever's willingness to publish plant-level results, rather than vague claims of transformation, is the part of this story most worth emulating.
