Supervised learning for AI assistants means councils can better serve people affected by the recession
Artificial intelligence (AI) has the potential to revolutionise the way local councils and other public sector organisations serve their residents, particularly during times of crisis. One type of AI technology that has garnered significant attention in recent years is Conversational AI, or AI assistants that are able to communicate with humans through natural language.
These AI assistants can handle routine queries at high accuracy rates, freeing up staff time and allowing councils to respond to increased demand for assistance. However, the rapidly changing nature of crises means that AI assistants must be able to continuously update their understanding of new queries and requests in real time in order to be effective.
In this article, we will explore how supervised learning techniques, such as SMART Mesh, can be used to ensure that Conversational AI is agile enough to respond to emerging user needs and help local councils reshape their assistance during times of crisis.
Conversational AI and the cost-of-living crisis
The current economic recession has put significant pressure on local councils to do more with less, as household costs have risen for many residents. In July 2022, 91% of UK households reported that their cost of living had increased in the previous month, up from 62% in November 2021. This increase has put pressure on councils to quickly answer new questions, efficiently direct users to helpful resources, and offer automations that can help residents self-serve urgent transactions instantly.
AI assistants with human parity performance, or those that can perform tasks at a level indistinguishable from that of a human, have the potential to help local councils better serve residents during times of economic crisis. These AI assistants can handle routine queries at high accuracy rates, freeing up staff time and allowing councils to respond to increased demand for assistance.
For example, during the COVID-19 pandemic, a pre-trained AI assistant was able to help agents at Cheshire West and Chester Council respond to a 500% spike in COVID-19-related queries, scaling up their customer service capabilities.
However, the rapidly changing nature of crises means that AI assistants must be able to continuously update their understanding of new queries and requests in real time in order to be effective. This requires the use of supervised learning techniques, such as SMART Mesh, which allow AI assistants to learn from each other and continuously update their language models.
SMART Mesh and supervised learning
For conversational AI to be effective, it must be underpinned by a robust language model, or a network of topics and questions that users will be enquiring about. This language model typically has three layers: intents, or the task or action that the user wants to perform; utterances, or the different ways that the user might ask for this task or action; and features, or any items that are relevant to the user's intent.
A language model, like language itself, is not static. It must evolve and be updated regularly in order for an AI assistant to understand new queries. One way to ensure that conversational AI is continuously updating its language model is through the use of supervised learning techniques, such as SMART Mesh.
SMART Mesh is a supervised learning approach that enables AI assistants to learn from each other and update their language models without requiring the customer to use valuable resources to do so directly. All of the AI assistants on a particular SMART Mesh (e.g. local government, higher education, mental health) share data from user inputs, allowing new queries and requests to be added to the existing language models of all the AI assistants on that Mesh. This helps to ensure that the language models stay at or above human parity level, preparing organisations to quickly respond to unpredictable scenarios.
The SMART Mesh update process involves several steps:
• Collects data from multiple data sources and performs pre-processing on it
• Mines the data to discover patterns, terms and frequencies
• Classifies and compares data based on the intents in the language model
• Refines and selects the list of improvements to be made
• Creates a proposal to be approved and reviewed
This is a fully supervised process, and the customer has a final say over what is included or excluded from the model. Sensitive data is protected while we work on the updates across the Mesh, so customers don’t have to worry about privacy issues.
Watch this short video clip from a recent webinar, where one of our data analysts, Shubhangi talks about the SMART Mesh process in more detail:
Conversational AI with human parity performance has a real impact
The current challenging economic circumstance means that the SMART Mesh could make a real difference in how councils respond to increasing requests of vulnerable people. The most recent instance of SMART Mesh’s impact on customer experience has been how it discovered the cost-of-living crisis queries appearing across different councils and enabled all our local government customers to quickly update their AI assistants with the latest queries and resources. Based on the supervised analysis, a list of questions was created and trained into the language model to facilitate the users to get quick access to additional information around the cost-of-living crisis – play the clip above to see the whole process on video.
As our society is coming out of Covid into the ensuant recession, the public sector needs the agility to assist increasingly anxious residents in a fast, cost-effective and efficient way. SMART Mesh and human parity conversational AI promise a way to help both the local authority and the people who depend on its services.
As an AI vendor, we are exclusively focused on the public sector. This means that our Conversational AI is built for higher education and further education with universities and colleges, local government councils, central government and the health sector institutions. We are very close to the specific challenges the sector is experiencing and we’re always exploring ways for Microsoft’s technology to help solve them.
Comments