Urbantz launches Milo, AI agents to manage last-mile deliveries
IA en un minuto newsroom · Editor: Jon Elgezabal
In 30 seconds
Urbantz is rolling out Milo, an AI assistant that sets agents to work on the delivery schedule when someone on the team asks in writing. Explore clears up doubts from the customer's information, Agents applies changes to routes and Workflows prepares complete processes for review. Milo always acts under each company's permissions and policies, and building custom agents will come later, with no date set.
The Belgian delivery software company puts AI agents to work changing the delivery plan from requests written in everyday language.
Urbantz, a Belgian maker of software for managing deliveries, has launched Milo, an artificial intelligence assistant through which planners and dispatchers ask ready-made agents, in their own words, for work that is carried out on the delivery plan. The company announced it on October 13, and Milo and its Explore agent have been open to every customer since that same day.
Urbantz works in the last mile, the final leg of a delivery to the recipient. Founded in Brussels in 2015, it says its platform handles upwards of 50 million deliveries a year across 20-plus countries and counts over 100 large companies as customers, El Corte Inglés, Carrefour, Bosch, FM Logistic and Chronopost XL among them.
The starting point is that a route is not settled once it has been planned. The company puts the number of changes made each week to already planned routes on its platform at roughly 52,000, and says most of them are still applied by hand, one at a time.
Milo is organised in three levels. Explore answers questions using the customer's own data: the company's example is a manager who wants to know why a depot failed to meet its delivery slots the previous week and is given the cause with no report to put together. Agents carries out instructions on the plan, such as adding orders that are running late to the afternoon routes, with the change applied on the spot. Workflows is meant for whole processes: the team describes one, for instance proposing another slot when a delivery fails, and Milo leaves a draft that people check and turn on.
The agents operate within the platform and use its own optimiser, so they take on the customer's rules: each action is kept inside the permissions, rules and approvals that customer has defined, and is recorded and explained. Urbantz argues that this lets teams get ahead of problems instead of reacting to them, with fewer failed deliveries and a lower cost for each one.
Thomas Paré, chief technology officer at Urbantz, calls Milo "only the first step".
That next step, letting customers build agents of their own, has no date. The release does not give prices either, or say whether Milo is charged separately.
Why it matters · analysis and opinion
Planning a route was already the job of an optimiser; what stayed in human hands was everything that goes wrong afterwards. Milo aims right there: an agent that receives a written instruction and alters the plan while it is running, without anyone moving each stop. For whoever runs a dispatch team, the difference lies in where the time goes, shifting from correcting to supervising. The condition that makes the product workable is the least showy one: permissions, approvals and a log that explains every change. Without that, no company would let a program reorder its deliveries. Two unknowns remain that the release does not settle: how much it costs and when customers will be able to build agents to fit their needs. A delivery company already using a routing platform would do well to ask its supplier what it offers in this area.
Official source: Urbantz · Written with the help of AI: how we make the news
