AI in logistics: a six-step roadmap from pilot to practice

Logistics and AI in numbers

Fewer than you might think: only 6.08% of EU companies using AI applied it to logistics in 2025.

Across the EU, 19.95% of enterprises with 10 or more employees used AI in 2025, and 55.03% of large enterprises did, according to Eurostat. But of all companies using AI, only 6.08% applied it to logistics, the least common purpose Eurostat measures.

The Netherlands, a logistics hub for much of Europe, is a telling example. According to Statistics Netherlands (CBS), one in six Dutch companies used AI in 2025, but in transport and storage the figure was just 7%. Together with hospitality and construction, the sector uses AI the least. And of all Dutch companies that use AI, only 4% use it for logistics. The 2025 sector study on transport and logistics by insurer TVM, bank ING and research firm Panteia sees the same in practice. Most Dutch transport companies use AI mainly for standard applications, not for core processes such as automated route planning. It is mostly the largest companies that have a strategy for it.

A lot of money and material is lost because the links in the chain do not connect well; that is where the biggest opportunity for AI lies. The factories and warehouses of the future, including their planning and scenarios, are designed to run without people, just like self-driving vehicles and drones.

What are the six steps?

Explore and choose first, then get your organisation and data ready, and only then buy, test and deploy.

Step 1. Explore what AI can do

Know what AI is and what it can do for your business model. The sector study recommends reading up, attending seminars and talking to peers, customers and staff. Ask your clients how they already work with AI. A student from a university or college looking for a graduation project can also help with this exploration.

Step 2. Decide where AI adds value

Go back to the core of your business. Where are the bottlenecks? Which processes take a lot of time? For each process, from taking an order to planning a trip and invoicing, list which tasks AI can do and where the planner or the driver decides. The study names route optimisation, fuel savings and better use of vehicles and staff as obvious places to look. A good strategy contains clear goals, the choice between building and buying, and a plan for roll-out. Without such a plan, the study says, AI projects often remain isolated experiments.

Step 3. Prepare your organisation and data

The rest of the plan rests on this step. The sector study lists six building blocks:

• Data quality.
  “Garbage in is garbage out.” A transport management system (TMS) needs clean, well-structured data, including from partners in the chain.
• An open mindset.
  Dare to experiment and learn from mistakes, but build in checks.
• Knowledge.
  Hire or contract specialists. AI literacy matters on the shop floor and in management.
• Support.
  Let staff practise with AI tools themselves, in a controlled way.
• Standardisation.
  What you want to automate has to be standardised first.
• An owner who backs it.
  In family businesses especially, this is decisive.

AI literacy is no longer optional in Europe either. Article 4 of the EU AI Act asks organisations that deploy AI to support the AI knowledge of their people. That duty has applied since 2 February 2025, and national authorities have supervised it since 2 August 2026, according to the European Commission. No certificate is needed; organisations can keep an internal record of training or other guidance.

Step 4. Buy or build?

For most companies, buying is the logical choice. According to the sector study, transport companies usually buy off-the-shelf software, and many are too small to develop AI themselves. Only large players have enough data for their own systems. Building in-house requires an IT department with AI expertise and can take months or even years. Buying has a downside. You become dependent on the supplier, and the question of who owns the data becomes important: the transport data belongs to you, while the AI tool often belongs to the supplier. Put that in writing before you sign. The study’s advice: start small and show your successes.

Step 5. Test and train in practice

A tool that works in a demo does not automatically work in your planning. The study gives a familiar example: an AI tool fails to recognise the input because someone entered information in the wrong field. You only find errors like that by testing with real orders and real trips.

Step 6. Deploy AI under supervision

AI makes predictions. It is not an exact science. According to the sector study, companies first use AI agents for administrative work, such as processing documents and invoices and answering standard questions. Later, an agent can draw up a standard plan that the planner checks. Some form of supervision is always needed to check the results. What happens when that supervision is missing, you can read in AI bloopers: artificial intelligence’s most memorable mistakes.

Where to start in logistics?

Start with one process with good data, and take extra care with tools that make decisions about people.

Pick one process that recurs often and for which you have good data. The sector study mentions route planning, predictive maintenance, load optimisation, inventory management and order picking in the warehouse, and customer questions about track and trace. In the warehouse, AI often goes hand in hand with robots. How far those robots have come, we described in Will logistics robots soon be advanced enough to replace human workers?

Be careful with applications that make decisions about people. The EU AI Act classes AI used to recruit and manage workers as high-risk, just like AI used as a safety component of critical infrastructure such as transport, according to the European Commission. For most of those applications, the strict rules apply from 2 December 2027. If a tool rates drivers or builds rosters, for example, ask the supplier how it is preparing for that.

What does AI mean for people?

Drivers and planners are expected to take on a different role, and knowledge grows fastest by doing.

Employees worry, and that is understandable. The sector study expects drivers and planners to take on a different role, with more emphasis on checking and deciding. The study also expects administrative departments to shrink in companies that introduce AI agents. At the same time, the study calls people indispensable in complex situations and moral trade-offs. The bigger question of what work means when machines take over tasks is one we explored in How do we find meaning, purpose, and identity after robots take our jobs? The most commonly cited obstacle is knowledge. Among EU enterprises that considered AI but did not use it, 70.89% cited a lack of relevant expertise, followed by unclear legal consequences (52.52%) and data protection and privacy concerns (48.83%), according to Eurostat. Dutch companies give the same top reason: 73% cite lack of experience, according to CBS.

I think knowledge grows fastest by doing. That is why I advise managers to give planners, drivers and warehouse staff room to try and experiment with AI, on one process from their own work. With every new technology, from AI agents to robots in the warehouse, I believe enthusiasm and a critical eye on what it delivers belong together. People who help shape their own new role make the transition easier, and that side of change is also the subject of the keynote on the organisation of the future.

Frequently asked questions

Is AI only for large companies?

No, but the approach differs. According to the sector study, large companies invest in their own models and data, while smaller companies are better off starting with the low-hanging fruit, using off-the-shelf tools that fit their existing systems.

Should I build or buy AI?

For most transport and logistics companies, buying makes more sense. Building requires in-house AI expertise and can take years. If you buy, agree in writing who owns the data.

Which rules apply to AI in my company?

In the EU, you have had to support the AI literacy of your staff since 2 February 2025, with supervision from 2 August 2026. Stricter rules apply from 2 December 2027 to high-risk applications, such as AI used to manage workers. The GDPR continues to apply as well.

Will AI take over the work of planners?

Not entirely. The sector study expects an agent to draw up the standard plan, while the planner checks it and gains time for other work. Supervision of the results remains necessary.