Logirus · Expert commentary

The algorithm and the unpredictable human in logistics

Can artificial intelligence replace an experienced logistics professional? Why do algorithms help newcomers work more efficiently while experienced couriers can still find better routes? We discussed these questions at an expert session at E-Retail Week 2025, which Logirus covered in an extensive article.


A tired mechanic and a robot beside a truck engine

Where algorithms are already delivering results

At the session, I shared several examples of AI applications in logistics and retail.

In inventory management at X5 Group, algorithms analyze numerous factors affecting demand, from seasonality and weather to consumer trends. According to the data I presented during the discussion, sustained development of these systems had enabled optimization of 15–20%.

Another example is route planning at CDEK. Algorithms help couriers plan up to 60 stops a day while reducing mileage. In the examples I presented, mileage fell by 12%, and the efficiency of new employees who did not know the city well could increase two- to threefold.

These are no longer experiments for the sake of technology — they address real operational efficiency challenges.

Why an experienced courier sometimes outperforms AI

Algorithms nevertheless have their limitations.

I gave the example of couriers who have worked in the same area for years. They know courtyards, passages, access points to buildings and alternative routes that are not represented on digital maps.

Thanks to that knowledge, an experienced employee can plan a route more efficiently than an automated system.

In my view, there is no contradiction here. An algorithm processes large volumes of data quickly, while a person supplements that data with knowledge that has not yet been formalized.

The most promising model is therefore collaboration between people and AI, rather than competition.

Why many companies are not yet ready for AI

Another topic I raised during the discussion was the technological maturity of businesses.

Many companies still manage a significant share of their operations in spreadsheets, exchange data through messaging apps and lack a unified information environment.

In these circumstances, a sophisticated AI model will not solve fundamental management problems.

I proposed that companies approach their development in stages: from digital recordkeeping, through process automation and integration, to the use of AI.

At each stage, they should select tools appropriate to their current business maturity rather than following the latest technology trend.

Where people must remain responsible

We also discussed accountability for decisions.

An algorithm may suggest an optimal route, a demand forecast or a way to allocate resources. But the company and its executives are accountable for the consequences of an incorrect decision.

I therefore believe strategic decisions, risk management, long-term partner relationships and team leadership must remain within the scope of human responsibility.

My main conclusion: efficiency comes not from AI itself, but from intelligently distributing tasks between technology and people.

The original Logirus publication also brings together the views of other participants, examples of warehouse robotics and document workflow automation, and new management questions arising from AI implementation.