In logistics, artificial intelligence is gradually establishing itself as a lever to better forecast demand, optimise transport and manage warehouses. But in practical terms, how do you go about it? Supply chain expert Yan Dupeyré explains.
Artificial intelligence is gradually moving out of the realm of buzzwords and into your warehouses, transport plans and demand forecasts. Companies are already seeing double-digit gains on their costs, stocks and service levels. So how can AI help you to better forecast, better plan and better manage your logistics operations, in France, Europe and internationally?
AI in logistics: key figures
AI brings together methods capable of learning from your data to provide forecasts, alerts and recommendations for action. In the field of logistics and the supply chain, it is based on your existing systems: WMS, TMS, ERP, IoT sensors, sales or stock files.
According to research by strategy consultancy McKinsey, which has been widely reported on the web and social networks, companies that have deployed AI in their supply chain have :
- improved their logistics costs by around 15 %
- reduced their 35 %
- increased their service levels by 65 %
In France, the market for AI applied to industry and the supply chain is expected to grow by around 19 % per year by 2029 (Technavio, Dec. 2025).
Forecasting demand and flows
For many logistics and supply chain managers, forecasting remains a mixture of Excel, history and intuition. AI does not replace this expertise, but it adds a layer of calculation capable of’aggregate dozens of signals :
- promotions
- weather
- calendars
- social networks
- web data
- etc.
In addition case studies show that companies that have introduced AI models to forecast demand have :
- reduced their stock levels by 10 to 20 %
- cut their supply chain costs by 5 to 10 %
- while improving their sales thanks to a better availability rate
For example: a consumer goods supplier supported by DataRobot has improved its forecasting accuracy by almost 9.5 points, with a direct impact on reducing out-of-stock situations and optimising stocks.
Testimony of a forecast manager
«Before, meetings S&OP was above all a battle of numbers between commerce and operations.
Since the introduction of an AI forecasting model, everyone starts from the same basic scenario. The tool gives us a more stable forecast, with around 15 % less error on our highly variable families, which has reduced our shortages and overstocks.
My role has changed: I spend less time on Excel macros and more on discussions with sales and production to adjust the scenarios».»
[Training] Manage and optimize supplies and inventory
Key points:
- optimising and streamlining supplies
- reduce the value of average stock
- improve the service rate
Practical example: calculating an optimum average stock and a safety stock.
Participants speak out:
"Concrete examples help us to understand things better.
"I can't wait to apply these tried and tested methods.
"I think I'll be able to handle my job better afterwards.
AI in logistics: optimising transport and routes
In the transport sector, AI complements the optimisation engines already present in TMSs and route planning tools. It takes better account of real-time data traffic, delays, actual loading times, performance histories by zone or customer. From feedback show that networks that have deployed AI solutions for transport planning have reduced their logistics costs by 15 to 25 % and their circulating capital requirements by 5 to 15 %. All thanks to better fillings, fewer empty kilometres and more reliable plans. AI also helps to provide more accurate ETAs (Estimated Time of Arrival), to reallocate routes in the event of unforeseen events and to anticipate the risk of delays on certain routes.
[Case study]
A French third-party logistics provider (3PL) managing multi-client warehouses was experiencing high transport costs and irregular deliveries.
- AI has been used to analyse route history, loading times, customer profiles, traffic data and delays.
- The system now offers optimised daily routes and Dynamic ETAs, by prioritising the most sensitive customers.
This type of approach, observed in several European networks, leads to reductions in transport costs of the order of 10 to 20 % and a significant improvement in the rate of on-time deliveries.
Managing the warehouse with AI: slotting, staffing, quality
In the warehouse, AI uses your WMS data: order history, picking times, errors, returns, picking profiles. It is used to :
- optimise slotting (product location) according to output frequencies and order associations
- propose staffing schedules based on expected volumes
- targeting quality controls on orders with the highest risk of error
According to a Technavio analysis, a market research and consultancy firm, the first deployments of AI in logistics environments in France are generating gains of up to 30 % in picking efficiency and 25 % in delivery rate. In addition, analyses McKinsey have shown that AI can reduce stocks in the retail sector by 20 to 30 %, thanks in particular to improved forecasting and flow management.
[Case study]
A European distributor with a central warehouse was suffering from a high rate of picking errors.
- A AI slotting engine has reorganised the locations according to picking frequencies and typical baskets.
- A staffing algorithm proposed schedules that were adjusted on a day-to-day basis, based on order forecasts.
The results observed in this type of project are consistent with the studies: a 20 to 30 % reduction in preparation errors, a 15 to 25 % improvement in picking productivity and a reduction in overtime.
Strengthening the resilience of industrial logistics
In industry, AI plays a key role in both production and the associated logistics. It can detect anomalies on lines, predict breakdowns and anticipate the impact on the downstream supply chain.
According to Technavio, French companies using predictive maintenance algorithms have reduced their unplanned stoppages by around 40 %. These gains enhance resilience: fewer disruptions, less emergency transport and more reliable planning.
Maturity checklist: is your logistics ready for AI?
Before launching an artificial intelligence project, it is useful to assess the maturity of your logistics organisation. To do this, answer the following questions:
| Yes | No | |
| 1/ Do you have reliable historical data on your sales, stocks, transport and warehouse operations? | ||
| 2/ Are your logistics systems (WMS, TMS, ERP) already being used in a structured and integrated way? | ||
| 3/ Do you regularly monitor key indicators such as service rates, stock-outs, average stock levels and transport costs? | ||
| 4/ Is your data accessible and usable, without complex manual reprocessing in Excel? | ||
| 5/ Do you have a team or a data manager capable of analysing and adding value to logistics data? | ||
| 6/ Do the logistics, purchasing, production and sales teams already work together on forecasting and planning (S&OP process or equivalent)? | ||
| 7/ Have you identified a priority logistics problem where AI could bring measurable benefits (forecasting, transport, warehousing, maintenance)? | ||
| 8/ Is a business manager (supply chain director, logistics director, industrial director) ready to take on the project? | ||
| 9/ Are your field teams open to the use of data-driven decision-support tools? | ||
| 10/ Are you ready to launch a pilot project (POC) on a limited perimeter to test the value of AI? |
If you answered «Yes to at least 6 or 7 questions, your organisation already has the necessary foundations to experiment with AI in logistics on an initial concrete use case.
AI and logistics: summary table of uses and conditions for success
| Main use | What AI does in practice | Typical benefits | Conditions for success |
| Demand and flow forecasting | Analyses history, seasonality and numerous external signals to produce a more detailed forecast by product, customer or channel. | Reduction in stocks by 10 to 20 %, fewer shortages, greater stability in the production and transport plan. | Reliable historical data, commercial/supply chain collaboration, integration with your S&OP processes. |
| Transport and route planning | Propose optimised routes, adjusting plans according to traffic, delays and customer constraints. | Transport costs cut by 10 to 25 %, better on-time delivery rate, fewer empty kilometres. | Up-to-date cost and time data, integrated TMS or tour tool, close dialogue with carriers. |
| Warehouse management (slotting, staffing) | Recommends the best stock layout, anticipates volumes to size teams and quality control functions. | +15 to 30 % increase in productivity, 20 to 30 % reduction in preparation errors, improved service rate. | Usable WMS, performance measurement (time, errors), acceptance by field teams. |
| Predictive maintenance and quality (industry) | Monitors equipment, detects signs of breakdown and identifies product defects in real time. | Up to 40 % reduction in unplanned downtime, less scrap, more stable supply chain. | Reliable sensors, maintenance data collection, production/maintenance/supply chain cooperation. |
How to get started : Targeted POC (Proof Of Concept) or global approach?
There are two possible approaches to moving from theory to practice.
1/ Start with a targeted POC
You choose a practical problem A product family with high variability, a pilot region for route optimisation, a warehouse under strain.
- Timeframe: 3 to 6 months.
- Objective: to prove a measurable gain (for example - 5 points in forecast error or - 10 % in transport costs) over a limited perimeter.
2/ Aiming for a more global approach
You are making AI part of a supply chain roadmap over 18 to 24 months, combining data, tools and organisation.
- Timeframe: 18 to 24 months.
Objective: to deploy several coherent use cases (forecasting, transport, warehouses, maintenance) with data governance and reinforced internal skills.
[Points to watch]
In both cases, the key success factors remain the same:
- data quality
- a clearly identified business line manager
- integration with your existing systems
- support for teams in the field
The success of an AI project in logistics also depends on the support of teams in the field. Involving them from the outset, training them in the new tools and sharing the results obtained are essential conditions for transforming technological innovation into a genuine lever for operational performance.
In conclusion, AI is not a magic wand. It is a powerful lever for improving your forecasts, your transport plans and the management of your warehouses, but only if you start from concrete problems and usable data. As a logistics or supply chain director or manager, you can start today by identifying an initial priority use, setting a realistic target figure and launching a POC or a structured approach that puts AI to work for your logistics, both in France and internationally. So, where are you going to start?
