Milo Explore
Analysis and root causes. Ask a question about performance, cost or failures and get the answer with the data behind it.
Which carriers caused most failed deliveries in September?
Urbantz / Site map
Milo / The AI layer of Urbantz
Milo works inside the Urbantz platform. It calls the same APIs, the same optimiser and the same permissions as your teams, so it inherits their controls instead of inventing new ones.
01/The three tiers
Analysis and root causes. Ask a question about performance, cost or failures and get the answer with the data behind it.
Which carriers caused most failed deliveries in September?
Real-time replanning. Describe the change, Milo finds and counts the stops, states the consequences and hands placement to the optimiser.
Inject the 14 late orders and remove cancelled stops before departure.
Builds workflows from plain language. Describe the process, review the steps Milo drafts, and switch it on.
When a premium order is late by 30 minutes, alert customer service.
Example prompts, to show the shape of a request rather than a fixed list.
02/Agentic workflows
Build a workflow in the visual editor, or describe it to Milo. Either way it is a chain of steps of three kinds.
Deterministic rules and actions. Find all failed tasks, update the delivery, re-plan the round.
A decision Milo takes from its inputs — attempts, customer profile, available slots — and the business rules your platform manager writes as a prompt.
Approvals placed only where you want them. The approver can be a planner, customer service or the recipient. A run can wait hours or days.
03/How a run goes
04/Workflow templates
Workflows taken from everyday last-mile operations. Each one starts from a trigger, runs a few steps, and ends with an action or an approval.
Milo finds the rounds likely to miss their windows — overloaded, late start, driver absent — and proposes fixes. The dispatcher confirms in one tap.
Sensitive · approval
On-time rate, failed deliveries and driver performance, compared with your usual range. Milo explains the anomalies in a short note to ops managers.
Read-only · no approval
Milo picks retry today, reschedule or pickup point, and the recipient chooses a new slot.
Reversible · approval optional
When an ETA slips past your threshold, Milo writes to the recipient in their language, unless they were already told. Key accounts go to a manager first.
Reversible · sensitive for key accounts
Upload a spreadsheet. Milo checks every row, fixes obvious errors, asks about the unclear ones in one go, then creates the tasks and reports what it rejected.
Reversible
Tell Milo Workflows what should happen, review the steps it drafts, and switch the workflow on.
05/Control & trust
You set the rules, choose where an approval is needed, and see every action Milo takes, with its reasoning.
Nothing changes in live data. Reports, analysis, root causes.
Can be undone, like a reschedule or a created task. Switch the approval off once you trust the workflow.
Touches live rounds, drivers or key accounts. A person confirms before it happens.
Milo acts with the rights of the person or workflow that runs it. Nothing more.
Ranking, feasibility and cost come from tested tools and the optimiser, not from the model.
Milo reports what the platform returned. If it cannot see a result, it says so.
Every action is logged with its inputs and reasoning, next to manual edits.
06/Why Milo
Milo is one agent your team works with every day, on screens they already use. It is not a set of autonomous bots running on the side.
Routes, driver app, delivery slots, recipient tracking and a Delivery Core per sector. We go deep on one job.
Milo applies your rules and works next to the dispatcher. You choose where a person approves, and every action is logged.
Milo Explore, Agents and Workflows act on the operational plan itself, with the optimiser doing the placement.
Shippers, carriers and drivers on one platform since 2015. Together, we deliver.
07/Where this goes
Urbantz is becoming the agent platform for the last mile. Milo is where it starts.
Milo Explore, Milo Agents and Milo Workflows, with agent steps inside your workflows, within your rules.
Your rules, data and workflows, described in plain language or in the designer. The platform runs them.
Retailers', carriers' and Urbantz agents coordinating every delivery, with people setting the intent.
We will run the agents against a plan that looks like yours, and show you where you would put the approvals.