Why smarter route-planning software still needs a better view of reality
Industrial logistics is gaining new development capacity from AI agents, but software sophistication alone is not making route plans more dependable. The central issue is whether a digital model can represent the conditions under which deliveries actually take place.
Research by Kardinal, conducted between 1 July and 8 September 2026, found that 35% of surveyed developers, CTOs and technical leads use an AI agent every day. A further 23% use one from time to time. Claude Code was the most frequently cited tool, ahead of GitHub Copilot, and respondents commonly reported using several systems together.
Automation does not remove operational complexity
The leading difficulty, identified by 56% of participants, is the mismatch between planned routes and real-world conditions. Computing performance and processing time were cited by 47%, while 43% pointed to changing business requirements. For 39%, even judging the quality of an “optimised” route is difficult.
This matters for Industry 5.0 ambitions, where digital systems must support resilient operations rather than produce isolated theoretical results. A route model may need to account for delivery windows, loading-bay access, refrigerated vehicles and customer-specific behaviour. These factors can determine whether a plan is executable, with consequences for service reliability, vehicle utilisation and wasted mileage.
Digital ownership brings trade-offs
Internal development is widespread: 56% of teams have built their own planning solution. Spreadsheets remain in use at 31%, matching the share using commercial software or an API. Hybrid architectures are common, and nearly half of the teams developing in-house also use another planning tool.
However, ownership does not by itself improve resilience. The gap between plans and field conditions affects 51% of in-house teams and 50% of others. Maintenance also consumes capacity: 35% of organisations running their own solution report regular maintenance, while 18% require substantial effort and dedicated resources. Only 12% describe maintenance as almost negligible.
Cost governance is another part of responsible digitalisation. Sixty-five percent of respondents are strongly put off by products without public pricing, and 26% will not assess them. The leading concern, cited by 41%, is an unexpectedly expanding bill.
The survey points to a practical priority for industrial adopters: evaluate how a planning system learns and represents operational constraints, not only how powerful its optimisation engine appears. Without that connection, the most efficient route in software remains an uncertain proposal for the people who must execute it.
