Virtual agents, the next efficiency measure for government?
Virtual agents, the next efficiency measure
Governments are experimenting with virtual agents with human-like sensory capabilities to offer help to residents | 🕒 3-min read.
Lack of funding is a matter of fact for many local governments. It creates a need to be efficient, but at the same time councils are under pressure to provide the digital capabilities expected by today’s citizens.
This is leading to the rise of the “virtual” or “cognitive” agent. These intelligent systems, based on artificial neural networks similar to the brain, are able to – in theory – mimic a human being.
They are being enabled by constantly improving computing power, which is giving experts the ability to “train” systems to analyze large amounts of data to develop machine learning – or artificial intelligence (AI) – algorithms.
The resulting virtual agents have human-like sensory capabilities. They can understand language, read text and recognize human emotions, giving them vast potential to improve service delivery for local and national governments.
Cognitive systems are already emerging in the UK, where Enfield Council in London is using a virtual agent called Amelia, based on US firm IPsoft’s AI technology platform. The agent can help residents locate information and complete standard applications, simplifying some of the council’s internal processes.
Eight municipal governments in Poland are using cloud-based virtual agents to deliver information on government services to citizens. The virtual web chat program developed by InteliWISE provides information on a wide range of services from identity cards to company registration and utilities.
Central government bodies are also exploring the benefits of virtual agents. In Singapore, chatbot Joanne1 at the Inland Revenue Authority is able to assist with tax-related queries. The agent, the first “V-Person” to be deployed by Creative Virtual in Singapore, has been implemented through partnership with Virtuariod.
“Artificial neural network algorithms based on how the human brain works are behind some of the most exciting advances we have seen in cognitive computing,” says Michael Rochanakij, director, solution 49x management consulting at KPMG in Australia. “They can read text, understand voice and see and identify objects. They can also process and respond to deviations from normal patterns. You can train them to be used in various ways – and because they can continuously learn, they improve all the time.”
Cognitive systems are developing rapidly, championed by software giants IBM, Apple, Google, Facebook and Microsoft. Among their advantages, they are able to detect tone of voice, so they can tell when people become frustrated or angry and act accordingly. In addition, the agents can man the phones around the clock and unlike a human, they do not get tired or grumpy.
“Governments have a lot of data, particularly text contained in documents,” says Rochanakij. “Because cognitive systems have natural processing capabilities and can understand language and text, there is big area of opportunity. For example, they can be used to go through years of child protection services case notes and understand systemic issues that may
There will be more opportunities as technology becomes cheaper. According to Rochanakij, processing power equal to a human brain will be available for roughly $1,000 (£806) within the next 20 years.
However, despite excitement around virtual agents, there are several issues preventing their wider adoption. Data quality is important to ensure the systems are trained optimally and this is difficult to manage. There are also cyber security implications: because many virtual agents are cloud-based, data sovereignty – rules governing the location in which information can reside – adds to complexity.
In addition, cognitive agents have not yet reached the point where they can mimic the human brain with complete accuracy. This will happen over time, says Rochanakij. “Language understanding has improved by 20% over the last two years, but at the moment we are only seeing basic interaction.
Currently, KPMG firms are working with clients to design systems that meld the intelligence of machines with humans, says Rochanakij. ”In other words, we are augmenting the human intelligence, not necessarily completely replacing it.”
Therefore, the best is yet to come. “There will be many exciting advances to look forward to over the coming 12 to 18 months,” says Rochanakij. “And the next few years will see the emergence of life-like digital avatars able to select the most appropriate persona to engage with a citizen.”
This content was originally published on the KPMG partner zone on the Guardian public leaders network on December 5, 2016.
1Chatbot Joanne at the Inland Revenue Authority.