Information overload
AI’s potential lies in its ability to handle very large data sets, improving both internal and external interactions, especially if organisations work together
While there may be understandable issues, both ethical and technical, around creating AI chatbots that interact directly with customers, what if the AI was used internally by customer service agents to better guide their interactions instead?
That was the innovation that Julia Mixter (now at Anchor) imagined when considering how AI could improve services when she was director of transformation at Raven Housing Trust.
Too much information
Explaining the genesis of the idea, Mixter says: “I think as human beings, we are expected to consume, understand and read vast amounts of data and information way beyond our capacity. A great example is people who get 100 to 300 emails a day. In those emails, you might have five attachments. If you get board packs as a member of an exec or a board, you’re going to be reading 500 to 800 pages of information. And if we stand back and ask if that’s a good and robust way of managing businesses, it’s really not.
“So if you take that to a more frontline level and you think about customer service teams, the idea that someone can just hold and retain 250 policies that are relevant to customers and know the answer quickly without reference to materials is unrealistic. That’s where AI has so much potential.”
Mixter says she turned to LinkedIn to find out if anyone had already built something that would create huge time savings for frontline customer-facing staff: “The idea was to try and build something, not for external customers, but for the internal customer service agents, so that the chatbot could go and interrogate your SharePoint site – a set of curated documents – and come back with an answer.”
Mixter wanted to find a model that was working effectively in the real world that would prevent her from having to spend months carrying out the build work for something that may never even get deployed at a meaningful scale. She found data scientist, Andrew Burgess, who had built something similar for a different client.
“So we gave him a curated data set and he went on and built a proof of concept chat bot model,” she explains. “We tested it and it came back with really good, solid answers, and those answers were referenced back to the original source material. So we were really starting to see the potential gains of using it internally.”
Julia Mixter
Executive Director of Business Services, Anchor
“The idea that someone can just hold and retain 250 policies that are relevant to customers and know the answer quickly without reference to materials is unrealistic. That’s where AI has so much potential.”
Stronger together
Although Mixter has since left Raven, the landlord is now working on the business case to scale the AI model up and make it operational throughout the business.
For Mixter, the important part of this story isn’t so much the AI model itself, but how important collaboration is for the housing sector if it is to unleash the full potential that lies latent in the technology.
“I think where there is huge potential for value is at a national level,” Mixter explains. “Housing associations are like behemoths that merge and grow, but at the same time, we put these artificial boundaries around our stock, and we say ‘this is mine’ even though we know we are operating side-by-side in the same counties.
“We talk about what’s best for our customers, but then we talk about our customers within our housing association, as opposed to saying, ‘actually, AI works at its best on very, very large data sets.’ So if we were to be able to finally reach a point where we could have a common data model around our property, then you could look at how AI can be applied to that data model.”
“I think where there is huge potential for value is at a national level.”
Common platform
The collaborative model could then extend outside the sector to drive even greater understanding of customers. Mixter elaborates: “You look at big data and how you can feed in information from local authorities, water and energy companies, and there is so much more value we could give to our customers.
“There is a great idea being considered in the sector at the moment by ODX to create a common platform – a customer portal – so any individual, who is looking for social housing, could go on to a platform and look up a property and look at the condition of the property and the EPC performance and everything else at a national level. Because maybe they don’t want to just stay in, say, Surrey and only look at this particular housing association’s information. Actually, maybe they need to move somewhere else and they’d like to know what their options are elsewhere. That would be when we are truly collaborating and driving sector-level change.”
Larger data sets would also drive improved performance within individual businesses, for example, when making predictive assumptions in relation to voids or rent arrears.
“Because the data sets are relatively quite small, particularly when you are looking at a small group of properties, the signals in the data are not strong, and so it’s difficult to make future predictions,” Mixter adds. “I believe that to get much better at some of the predictive analytics we’re all talking about, you definitely need larger data sets.
“If we looked at arrears at a national level, using AI, we may find trends and themes we aren’t even aware of at this stage. But I think we need some collaboration around this at a sector level if we really want to make good progress.”
SHARE THIS PAGE
