Transforming Public Service Delivery with AI

South Waikato District Council provides municipal services to homeowners and businesses in New Zealand’s Central North Island. The area is rich in cultural heritage, with the Māori iwi (tribe) Ngāti Raukawa having a significant presence in the rohe (area). Following COVID, the team at Council faced challenges common to many; high staff turnover, increased pressure on service delivery, tight budgets and a need for organisational change. With many organisations reducing services and laying off staff, the executive and elected members sought technology to streamline their operations and rapidly reduce time-to-effectiveness for new employees.

They took aim at:

·    Inconsistent information taken from customers (delaying service delivery)

·    Knowledge ‘inertia’ with 6 months+ for new Customer Service team members to become effective

·    Customer queries passed around the organisation

·    Siloed knowledge, effort and a lack of a cross-Council view of delivery

An extensive investigation into Knowledge Management and automation was commissioned. Trials of 3 major Knowledge Base (KB) solutions served to demonstrate aging products that lacked innovation. At the same time, AI toolsets were becoming available, with the potential to eclipse traditional platforms. With partner Bluefern, Council took its drive for transformation through advanced technology to heart and opted for the IBM Cloud platform – and WatsonX in particular – to produce a solution. The Chief Executive was clear in what she wanted:

“I want anyone at Council to be able to answer any question from stakeholders and to trigger the delivery of any service at the point of interaction with the customer”

Development and Delivery

 It became quickly apparent that WatsonX Assistant not only provides AI-driven answers but can deliver workflows – business automation. It could integrate with existing Council systems to enable service delivery. Agile collection of business process information from customer-facing teams was quickly translated into Actions. Any Council employee can complete these Actions and turn them into services, without seeking permission or having an in-depth understanding of a wide variety of topics.

Interconnection with Microsoft (Teams and Graph API) and 3rd party systems including Council’s ERP and geospatial services was straightforward. With an orchestration and integration layer, automation of key processes such as new building work, animal registrations, planning, building standards, event management, rates and other financial queries could be quickly added. The platform has sped-up responses to roading, water supply, drainage, animal control, parks and facilities issues.

With Watson Discovery, the project has an effective knowledge repository – loaded with Council’s own bylaws, policies, FAQ’s and web content – processed using natural language in the cloud privately. Large Language Models are used within the IBM platform without exposing information for 3rd party use such as training. The team worked with IBM Customer Engineering to interface with WatsonX.ai to process employee questions into accessible, understandable answers. Using the platform API, the Council can also manage the knowledge base content in Discovery from their own Microsoft SharePoint Online tenancy.

The result is Moko – Council’s own knowledge and automated service delivery agent.

Impact

Moko has over 100 Actions, carrying 105 different service variables. She is connected through 6 external custom integrations to Watson Discovery, WatsonX.ai, Microsoft Teams, Microsoft Graph API, GIS and Council systems. Launched from their own infrastructure, Moko has been adopted by Customer Services, with rollout continuing across Council.

The pace of automation started modestly. As the integrations were refined and quality improved then the flow of requests-for-action has accelerated dramatically. Moko can now handle the majority of all request types, with more being added daily.

Take-up was initially slow, particularly amongst more experienced staff. New team members have been most enthusiastic. The technology has proved a significant shortcut to training; as well as a boost to individual staff confidence in dealing with the public. In continuous use across 3 teams, the programme is starting to promote use outside of Customer Services.

The adoption of Moko as a reliable source of answers has been remarkable. With the Llama3 model at its heart, the Assistant now provides answers against 50 key Council documents including bylaws, FAQs, standards, guidelines and news items as well as content from its website. 3rd party information on non-Council services is also helping teams redirect customers to agencies who can help. Content is expected to double in the next 3 months.

Coming up

The pace of development and delivery continues with exciting new capabilities on the roadmap.

• Migration to the latest Llama 3.1 large language model

• Appointments capability - booking of inspections, advisory sessions and meetings within Moko

• Release to the general public; starting with a customer focus group

• Connection of the voice channel through Genesys call centre integration

• Connection of social media channels

• Semantic search enhancements