By Simon Duddy, Editor, Logistics Matters

THE NEXT phase of warehouse productivity will not be defined solely by mechanical automation. It will be driven by the ability to use operational data intelligently, at scale and with human oversight. So says global, tech-led 3PL GXO.

GXO has long invested in automation, particularly where processes are repeatable and where technology can reduce pressure on frontline teams. Nizar Trigui, chief technology officer, GXO Logistics sees that work continuing, but he is clear that AI represents a broader shift. It is not another tool to improve warehouse performance. It is a different operating model.

He says: “AI is definitely the next frontier. Historically, we have excelled in automation, bringing solutions to the market, but a lot of that was more towards processes that are repeatable, relieving the pressure on our teammates in the facilities and replacing repeatable action with automation.”

GXO’s view is AI can help customers manage complexity by improving productivity and giving operators better information at the point of decision.

Nizar adds GXO has been piloting AI capabilities for almost two years, drawing on large volumes of historical and real-time operational data from sites around the world. Some of those facilities have been running for close to two decades, giving the company a substantial base of practical warehouse knowledge. The aim has been to turn that information into models capable of identifying patterns that would be difficult for people to spot manually.

That work has been consolidated into GXO IQ, which Nizar describes as a smart operational platform. The company has committed to deploying the model across at least 50 sites globally this year.

Productivity improvements

The early gains have been measurable. Nizar says GXO has seen double-digit productivity improvements in operations where AI modules have been applied, even at sites already supported by strong operational leadership.

One example is GXO’s pick optimiser, recently deployed in the UK and Ireland. Traditional picking optimisation has focused on reducing the distance travelled by a picker. GXO’s AI adds another layer by considering order density. Instead of processing orders simply on a first-in, first-out basis, the system analyses the order pool to identify combinations that allow a picker to fill a cart more efficiently.

Nizar draws a clear distinction between machine learning optimisation and agentic AI. In GXO’s direct warehouse operations, the focus is primarily on prediction and recommendation. The system can propose decisions, but people remain responsible for approval and intervention. This human-in-the-loop approach reflects both operational caution and the importance of contextual judgement in edge cases.

Nizar clarifies: “We believe human in the loop will continue to be how we manage this because there are subtleties that these machine learning systems still do not understand in edge cases, so we are not ready to give them full control.”

The role of warehouse teams is therefore expected to evolve. Rather than only processing tasks, employees will increasingly validate AI recommendations, check assumptions and help improve the models through feedback. If an operator rejects a recommendation, GXO captures the reason and uses that information to refine the logic.

GXO IQ also includes a natural language interface, giving managers the ability to query operational performance directly. Users can ask whether a site is likely to meet its service-level agreement that day or whether labour should be moved from one activity to another. In indirect processes, such as payment processing or vehicle coordination, GXO is applying agentic AI more directly, while still maintaining oversight.

Robotics

Robotics is another area where AI is beginning to reshape expectations. Nizar says the development of physical AI is progressing quickly. Historically, robotic training required extensive software development and algorithm design. Newer systems are moving towards low-code and no-code approaches, where robots learn by observing and mimicking human actions through visual inputs.

“The pilots that we are in right now are almost no code. The robots are learning from actions, from mimicking actions from teammates through visual means. You can see the algorithm in real life improving how it picks,” adds Nizar.

While Nizar does not expect widespread deployment immediately, he believes useful applications of physical AI could become realistic within 18 months to two years. The key distinction is between pilots that demonstrate possibility and systems that can perform productive work in live operations.

Culture

For operations managers, the change is as much cultural as technical. Nizar argues AI must be treated as business transformation rather than a standalone technology project. Adoption requires education, communication and structured change management. Teams need to understand the value of the tools and see them as enabling rather than imposed from above.

Nizar says: “AI is really not a technology change. AI is true business transformation and operation transformation. It requires proper change management, lots of communication, a lot of education, a lot of training, and taking people through the emotional transition that they need.”

This is particularly important in third-party logistics, where customer requirements vary significantly. GXO has historically built highly bespoke solutions, but Nizar said the company is moving towards modular and configurable models. The goal is to preserve customer-specific functionality while creating platforms that can be scaled quickly.

AI is central to that balance. Unlike traditional automation, which can struggle when conditions vary, AI can be trained on site-specific data. Nizar said a globally developed model can be deployed initially and then trained until it learns the characteristics of a particular site.

Scale is one of GXO’s major advantages. Nizar says the company processes around 970 million events every day, including edge cases and operator trade-off decisions. This volume of data allows models to be retrained regularly and adapted across a large operational footprint. Around 70% of GXO’s sites are now connected to the infrastructure that enables nightly data collection.

GXO is not presenting AI as a replacement for people or as a short-term productivity fix. It is positioning it as an operating layer that can enhance decisions, improve consistency and help logistics networks adapt to rising complexity.