Artificial intelligence is moving deeper into logistics, from warehouse planning to transport visibility. Yet Matthew Sterner, Global Customer Marketing Leader for Transportation, Logistics and Warehousing at Honeywell, argues that the technology should follow operational understanding rather than lead it.
The starting point is data. AI can analyse structured business records alongside contextual or unstructured information, but unreliable inputs produce unreliable recommendations. Logistics companies therefore need to identify duplicated records, hidden delays and exceptions that are not reaching the right decision-makers.

Process mapping before automation
One shipping workflow reviewed by Sterner’s team contained roughly 30 steps and repeated manual data entry. Shipments could progress despite mismatched information or a missing service requirement, such as Saturday handling for a Friday dispatch. After mapping the workflow, the organisation separated value-adding work from checks that could be automated. Human handling time fell from approximately six minutes to 30–45 seconds.
The lesson is broader than one efficiency gain. Automation is most useful when it removes repetitive activity while directing unusual cases to people with the authority and experience to resolve them. Companies must define which recommendations a system can execute, which require approval and when an issue must be escalated.
From prediction to action
AI may help warehouses anticipate demand changes, including sudden interest in a product, and adjust inventory positioning before orders arrive. In transport, electronic documentation, digital signatures, route optimisation and carrier matching can reduce manual friction. Weather data combined with vehicle location could identify a likely delay, propose an alternative and update an arrival estimate.

These applications improve resilience only if the underlying signal is accurate and teams can act quickly. The same principle applies to warehouse workers. Tools should reflect the task, the environment and the user’s experience, whether that means hands-free guidance or fewer routine decisions.
Sterner recommends beginning with a contained problem and measuring the result. A successful pilot can build trust, expose data weaknesses and provide practical lessons for wider deployment. The strategic priority is not acquiring AI for its own sake, but creating processes in which technology and people share responsibility clearly.







