Evri deploys Microsoft AI agents across parcel network

TL;DR:

• Evri chief technology officer Marcus Hunter is building five AI “super agents” to automate routine operations and surface accurate information for employees, couriers and customers.
• The agents interrogate cloud-based data to improve visibility of the parcel journey, optimise routes and tailor customer communications.
• Microsoft’s Agent365, part of its new E7 suite, provides observability and control over the agents themselves.

Evri handles parcels for around 25 million UK customers, which makes this one of the larger agentic AI deployments in British logistics — and one where the results will be visible to the public rather than confined to an internal dashboard.

Hunter’s stated ambition is personalisation at individual level: “Everybody talks about personalisation. But we’ve never had the tools to do personalisation properly before. Now we do. So over the next few years, we will eventually be able to give individually tailored, personalised deliveries to our 25 million customers.” The immediate mechanics are more prosaic — route optimisation, better parcel-journey visibility, and giving couriers and call handlers faster access to accurate answers.

Managing the agents, not just deploying them

The Agent365 component is worth separating out. Microsoft is selling not only agents but a control plane for them — observability and governance across an organisation’s whole agent estate. That an enterprise of Evri’s size is buying oversight tooling alongside the agents themselves suggests the operational lesson about agent sprawl has been learned before deployment rather than after it.

A testable claim

Evri’s customer service reputation makes this deployment unusually falsifiable. Where most enterprise AI announcements rest on internal efficiency metrics no outsider can check, this one will be measured by whether parcels arrive when the company said they would. That matters given research this month showing AI chatbots gave false information about 64% of UK retailers — accuracy at the customer-facing edge remains the weak point in retail AI.

Looking forward

No timeline or spend has been disclosed, and Hunter frames full personalisation as a multi-year goal rather than a shipped feature. For UK logistics operators watching, the useful signal is architectural: a named enterprise committing to a small number of broad agents with a governance layer over them, rather than dozens of narrow point solutions.