Technology

Right now, the Product team are nearing the end of complete app rebuilds. While working on NexDelivery, the design process lit everyone up as the team went straight into innovation mode.

Here’s Adrian, Co-founder and CTO

We were looking for a solution to help people on the frontline of operations manage deliveries more easily. From speaking with customers, we knew that easing parcel and mail processing would be helpful.

We were also aware there was a general increase in demand for delivery services as many coworking spaces are exploring offers such as virtual offices, and managing more and more mail.

I’ve known someone in the warehousing industry for some time, in fact they’re a customer of ours, so I called him to pick his brain.

She referred me to their software company. This organisation handle vast and very complicated global logistics, and quoted over £200k to use their services. It wasn’t an unfair number considering the sophistication of the package, but we didn’t need to use most of the features or workflows  – they’re built to cater for a very different set of logistics.

That’s when we turned to the data we already had and started to play with it.

We wanted to know if we could train a machine learning model to recognise and match data from an image of a mail label. We fed our model 10,000 a very diverse mix of photos of address labels, and waited.

The results surprised us all. Not only could it extract typed data, our in-house AI was able to recognise handwriting and match it to a name in the database. I will say, the handwriting does need to be legible, but that’s the same for any other scenario.

Adding AI functions to NexDelivery

We deployed this immediately and it’s now part of the NexDelivery app. When deliveries come in, users can upload a photo of the address label and NexDelivery will match it to known members and auto populate the ‘Register a delivery’ field. This, alongside an update that allows users to manage multiple deliveries at once, is incredibly useful for our users.

It’s a great example of best-practice design: start with a real customer problem, explore every avenue, and never be afraid to build the solution yourself when the perfect one doesn’t exist off the shelf.

I’m lucky enough to work with exceptionally bright and motivated people and am very grateful for the hard work the team put into rebuilding the app, and for enthusiastically embracing new experiments like this together.