AI adoption grows, but route planning still misses reality
Route-planning teams are adopting AI agents at speed, but the technology has not removed the most persistent weakness in transport optimisation: plans often fail to reflect what happens on the road.
A survey conducted by Kardinal between 1 July and 8 September 2026 found that 35% of developers, chief technology officers and tech leads use an AI agent every day. A further 23% use one occasionally. Claude Code was the tool mentioned most frequently, followed by GitHub Copilot, with many respondents combining several platforms.
The operational gap remains the main obstacle
Despite rapid advances in software development, 56% of respondents identified the difference between planned routes and actual operating conditions as their leading challenge. Performance and computing time followed at 47%, while 43% pointed to changing business requirements. For 39%, it is difficult even to determine whether an algorithmically optimised route is genuinely good.
The finding has direct implications for industrial companies investing in automation. A route may look efficient in a model while overlooking delivery windows, loading-bay availability, vehicle refrigeration requirements or established customer practices. In those circumstances, better computation alone does not guarantee better execution.
Build-versus-buy does not close the gap
In-house development is used by 56% of the teams surveyed. Spreadsheets remain part of the technology mix for 31%, the same share that uses commercial software or an API. These methods are frequently combined, and almost half of the teams developing internally also rely on another planning tool.
However, internal ownership makes little difference to the central problem: 51% of in-house teams reported difficulty aligning plans with field conditions, compared with 50% of other respondents. Maintenance is another burden. Among organisations operating their own solution, 35% described maintenance as regular and 18% as heavy enough to require dedicated resources. Only 12% reported almost no maintenance effort.
Commercial transparency is also becoming a purchasing issue. Some 65% said they are strongly deterred by products without public pricing, while 26% would not assess such a tool at all. The most common concern, cited by 41%, is uncontrolled cost escalation.
Kardinal co-founder Cédric Hervet argues that the industry does not lack capable algorithms, including open-source solvers. The differentiator is whether software can encode the constraints that shape daily operations. Without that connection to reality, an apparently optimal route remains a theoretical result rather than a dependable industrial plan.
