When machines fail, the layer that brings them back to working has never existed. Aleya is building it, starting with EV charging.
The world is quietly filled with connected machines. Chargers, gates, kiosks, terminals, robots. Each one is online and reporting back to a server somewhere.
One in four charges fails and most of those failures aren't hardware. The machine knows what's happening, the driver doesn't, and nothing in between is built to connect them.
Aleya is the AI recovery layer for electronic infrastructure. We start with EV charging. This isn't a support tool. It's a new layer of the stack.
In December 2025, EVs outsold petrol cars in Europe for the first time (ACEA's "Vehicles on European Roads", January 2025). Charging stopped being a curiosity and became infrastructure that millions of people depend on every day.
The build-out is real. The recovery layer is not.
Only seven in ten charges succeed. The rest dissolve into failed sessions, abandoned attempts, and calls to outsourced helplines that were never built for the problem. The bill, across Europe and the US, runs into tens of billions a year in lost revenue.
We've spent the last year inside the operations of more than thirty Charge Point Operators (CPOs), listening to call recordings, mapping failure modes, and sitting with the people whose job is to keep a distributed physical network alive.
The pattern is consistent.
Roughly half of failed charges come from user confusion or error. Nearly the other half can be resolved remotely through software actions like a reboot or a cable unlock. Only a small fraction need a person on site.
Almost every failure could be recovered in the moment. Almost none of them are.
Instead, the path to resolution runs through outsourced contact centres that work across multiple industries, carry partial context about the charger in front of the driver, and lose ten or more minutes to hold times during peak hours. The driver leaves. The session is written off. The operator never learns what actually happened.
The result is a system where the machine, the operator, and the driver are all in the same loop, and none of them can close it.
Aleya embeds an intelligent agent into connected infrastructure. For EV charging, that takes two forms.
A driver calls the white-labeled Aleya number on the charger. The agent answers immediately, with no queue and no menu. It reads live data from the charger backend, identifies the most likely cause, and walks the driver through recovery in plain language. It listens. It follows the thread. It behaves like someone who actually understands the machine, because, in a real sense, it is the machine.
See a conversation with the agent↗Every conversation, every failure mode, every resolution becomes structured operational data. The Control Room shows operators where failure is concentrating across their network, why it's happening, how drivers are responding, and what to do next. Not transcripts. Not tickets. A live view of recovery itself.
See the Control Room↗There's a familiar version of this idea that lives in the customer service category. Smarter call deflection, faster routing, better transcripts. That isn't what Aleya is.
The bet is that recovery belongs inside the machine, not around it. The machine knows what failed. The driver knows what they tried. Closing the gap between those two pieces of information is the actual work, and once it's done, the operator inherits an operational system they didn't have before.
The interface happens to be a voice. The product is the layer.
Aleya went to market in the UK in January 2026.
We're live in deployment with Y, the largest operator of charge points in Northern Ireland. Z, one of the UK's largest CPO networks, has given us full access to their charge management system as we finalise a pilot. We're in active conversations with one of the largest CPOs in the UK by capital raised.
The work is operational, not theoretical. Every conversation that happens on Aleya teaches the system how recovery works on a specific network, with specific hardware, in front of specific drivers.
The pattern repeats outside charging.
A parking gate that won't lift and traps a queue of cars.
A hospital machine that goes down at the wrong moment.
An automated carwash that locks a driver inside.
An access terminal in a logistics yard that quietly stops authorising deliveries.
In each case, the machine is online. In each case, the person in front of it isn't an engineer. In each case, the path to recovery is hidden behind a support queue that was never designed for the problem.
We think that changes. Connected machines should be able to explain themselves, guide a recovery, and turn every failure into something their operator can learn from.
Aleya is the layer that makes that possible.