Australia Post is piloting a series of bots to help office staff scrutinise delivery data, and delivery centre workers get sorting machines up and running more quickly.
The ‘Dexter’ bots have three main functions – “notify me, tell me and help me” – the postal service’s general manager data science and strategy, Silvio Giorgio, revealed last week.
‘Notify me’, which is already in production, “monitors data in realtime” and sends alerts to relevant field staff to search for parcels due for delivery that day that haven’t been scanned onto a van.
“It’s very unlikely they’ll be delivery that day if there’s no on-boarding scan. [So the bot] sends people out into the network to look for it,” Giorgio told the AI NSW Summit in Sydney last week.
“It’s technology like this which is allowing us to get better,” he added.
The ‘tell me’ function, which is being piloted, gives back office staff the ability to send questions to a chatbot about various company data. For example – “How many parcels were delivered in Woollahra last month?”
The user is able to ask follow up questions; for example, “How does that compare to the same month last year?”
Giorgio demonstrated an experiment his team had done where questions were posed via a Google Home device. “You wouldn’t use a Google Home, it’s not practical or secure. But we’re looking at how people would want to engage with the information bot. It’s the concept we’re looking at,” he said.
The third function – ‘help me’ – is currently being trialled at Australia Post’s Melbourne Parcel Facility in Sunshine West, through which around a third of all parcels sent in Australia are processed.
“In any 24-hour shift period, four shift managers control the facility from a control tower. They continuously monitor machines and when there’s a breakdown, try to get them back up and running,” Giorgio said.
The biggest impacts on parcel processing time in the facility is the number of times the sorting machines go down, and the time it takes to get them up and running again (the rebound rate).
The Dexter trial uses the bot to identify the problem with the machine, alert ground staff “before the shift manager knows about it” and tell them what to do to fix the issue.
“If Dexter doesn’t know it will deviate to the shift manager who will tell it what to do. And then it will monitor the rebound rate – if the rebound is within a certain tolerance it will bank that response for the next time,” Giorgio explained.
“The idea is the shift manager is going to be working with Dexter and that’s Dexter learning. And at any point in time, we get the best shift manager, regardless of who is on shift,” he added.
Giorgio also hinted at a separate experiment in which less experienced postal workers complete difficult delivery rounds in virtual reality.
“We could use virtual reality to train some of the younger drivers on the routes they are unfamiliar with, and use data as they’re going through that VR to highlight these are the hazards and these are where they’re happening,” he said.