While GDR has frequently discussed the growth of Artificial Intelligence in the retail and hospitality sectors in recent years, one of the most potentially significant rollouts of the technology is taking place in the public sector. Enfield is the first UK council body to introduce AI to its services in the form of IPSoft’s cognitive agent, Amelia. GDR Innovation Researcher Sophia Platts-Palmer met with the council’s interim Assistant IT Director Rocco Labellarte, formerly of Calvin Klein and Unilever, to discuss how Enfield will use the technology and what the biggest challenges have been so far.
SPP: Why is Enfield looking at implementing AI as part of their services?
RL: I think Enfield actually gets it when it comes to saying “You can’t just keep top slicing, you can’t keep driving efficiencies by decreasing the number of staff.” Whilst there is always an aspect of efficiency and optimisation involved in local government, Enfield and its partners have been looking at how they can move into a more commercial space and what they can lead with. I have been looking at how digital technology is being used in the private sector as an exemplar.
What specifically will you be using Amelia for?
There are three areas of local authority delivery, and one of these could be targeted easily by Amelia for maximum effect.
The first, we call high volume-low complexity transactions. These include tasks such as paying council tax and reporting potholes. Visitors answer five questions and they’re done. Should we use AI for this? Probably not, the same result can be achieved via an e-form at a much lower cost to the budget.
Then at the opposite end of the spectrum we have low-volume-high complexity requests, areas such as job seekers allowance, housing, asylum. Would you want to unleash an untested AI in these areas? No.
So the requests that we are targeting with Amelia are the mid-complexity-low-volume transactions.
E-forms don’t really cut it when it comes to these requests as a little bit more technical know how is needed to provide a satisfactory answer. Let’s look at planning permission as an example.
Currently we actually have a lot of experienced staff sitting on a phone answering lots of unnecessary queries rather than doing their actual job. But if the AI can provide the same answers then why not bring it in to free up the time of our staff who are not currently being allowed to work to their full potential.
It’s going to take a lot of pressure off those teams that are slightly more expert and free them up to focus on more pressing tasks.
This is how we will start.
Why have you opted to work with IPSoft?
What we wanted was two things. One, we wanted a solution that can help us interact with people 24/7 in the ways already mentioned. Equally important, in terms of looking forward to a slightly more social space, we want to have some kind of software that can recognise emotion.
We have booths where people sit down and talk to you about their problems. Now, we can’t be monitoring these staff members and the conversations, you don’t want it recorded. But actually you could have real-time audio listening software, so Amelia for example, can pick up words, can tell from intonation that the emotional construct is changing. When Amelia thinks, “this conversation sounds aggressive,” then it can alert senior members of staff. There’s a multitude of things you can do with emotional context there.
There was a lot of press over the 300 jobs at Enfield Council that were cut around the same time as Amelia was announced. And there is a public anxiety that automation will replace jobs, particularly service jobs. What are your thoughts on this?
In our case there is no correlation between the two. The cuts were already planned.
It’s a very fine balancing act. If you think about the last industrial revolution, we went from an industry-based economy to a service-based economy and exactly the same thing is likely to happen now, if history is to repeat itself.
In any workforce facing an economic shift, you have a certain number of people with skills that are no longer going to be required.
In the nine months that we have been working on Amelia, we’ve seen very clearly that the skills needed to feed and water any AI are a whole new set of skills that our current economy doesn’t have. So they have to be built.
How will you do that?
It’s a difficult thing to do; you’re asking people to interact with a machine in natural language. How do you get that information across to people when their language isn’t necessarily English?
It’s the human beings that are building the nuance into it and this requires identifying and developing a whole new set of skills and once you come up with the idea, you need the skills to connect the dots together.
So we’re going to shift from a service-economy, to an AI economy.
How do you think residents of Enfield are going to respond to this? And how is Amelia’s implementation going to be communicated to the residents?
That’s a really important question – probably one of the most important questions.
When the press announcement from IPSoft came out last year saying that Enfield is going to be the first council to implement AI, everyone thought we had it already.
We did have a live agent chat solution. We started getting a few complaints from people, saying that our robot wasn’t being very helpful. It was actually live agents, but of course when you’re texting you can’t tell the difference. So that was interesting in terms of, how perceptions feed behaviour.
However, if we don’t say anything and launch it quietly, the press will get wind of it and think we have snuck it in covertly. So we’re going to have to find a middle ground.
What we are planning to do is to launch Amelia within the same environment as our live agents. We will have live agents monitoring the flow of requests who will be working side by side with Amelia, keeping an eye on things. This will eventually be phased out as the system grows smarter but live agents will always be on hand in the event a query that Amelia cannot process.
She will only be working 9-5 to begin with, which is what we did when we introduced the live chat function. We work in increments of innovation.
Are the public going to be made aware they are talking to Amelia?
Yes. The way we have chosen to do it is to be very transparent.
How do you hope Amelia will serve your residents in the future?
We run cities, and we run everything in the city; from roads to social care to the bins. The tech to streamline and optimise these tasks is there, but in the past it’s just not been strung together in a way that’s cost effective and part of our challenge is to do that.
If local councils can come up with a way to create a digital ecosystem where they all invest together in building and implementing new, optimising technologies and pooling resources, we would be able to do things we could never do by ourselves, and do it faster.
We’re doing a big regeneration project called Meridian Water, which is anticipated to build 10,000 houses in the Enfield area.
You imagine having an AI like an Amazon Alexa in every home, all networked together, all running all the council services and everything else together by voice command and what else could you plug in to there then? Once you have that ecosystem in 10,000 houses, its not just about council services, what else can you do?
Do you have any contingencies in place if the Amelia project is not successful?
Failure is not an option. I think the important thing is to say we’re on a journey. If we were to say were not going to do this, we’re just going to wait until someone else has done it. We’d find ourselves two to three years from now having to start doing it, and having to learn how to do it, so now feels like a good time.
Enfield Council’s Amelia-powered services will be officially launched on 27 June.