Go back to the homepage

Delivering accessible family services with AI

Rebuilding Nia, an AI chatbot for BCP Council, to improve access to family services through clear UX, accessibility, and scalable design.

Overview

Nia is an AI chatbot developed for BCP Council's Online Family Hub, designed to support residents across Bournemouth, Christchurch and Poole.

It provides quick, conversational access to information already available on the website, helping users find answers without navigating multiple pages.

Key purpose

  • Reduce friction in accessing council services and information
  • Support users who prefer direct, immediate responses
  • Complement existing website content

Core focus

  • Retrieving relevant content from the Family Hub
  • Presenting it in a clear, user-friendly format
  • Guiding users to the right service or next step
The interface of Nia on the Online Family Hub website

The problem

The Family Information Directory (FID) contains a large volume of content across many pages and services.

While comprehensive, users often struggled to find the right information quickly.

User challenges

  • Faced an overwhelming number of pages
  • Struggled to identify relevant services
  • Needed multiple clicks to reach a complete answer

Barriers to access

  • Uncertainty around terminology
  • Unfamiliar service structures
  • Preference for direct answers over browsing

Service impact

  • Underutilisation of available content
  • High reliance on search and navigation
  • Inconsistent user journeys

The issue was not content availability, but the effort required to access it.

Initial delivery

Nia was initially developed with a third-party consultancy using an existing chatbot framework.

This delivered a working proof of concept and validated the use of conversational access to the Family Hub.

What was delivered

  • An embedded chatbot within the website
  • Retrieval of Family Information Directory content
  • A basic conversational interface for users

Key constraints

  • Limited control over UI and interaction design
  • Restricted ability to customise behaviour and flows
  • Dependency on external framework updates
  • Difficult to extend or reuse for other use cases

The solution demonstrated value but was not suitable for long-term flexibility or reuse.

Identifying the gaps

User experience

  • Rigid interaction patterns
  • Inconsistent response structure
  • Limited control over presentation

Accessibility

  • Inconsistent keyboard navigation
  • Unclear assistive technology behaviour
  • Limited control over contrast and focus states

Design

  • Framework-constrained UI
  • Inconsistent alignment with Family Hub
  • Limited iteration capability

Technical

  • Tight coupling to third-party system
  • Limited extensibility
  • Poor reusability

Strategic decision: rebuild vs iterate

The focus shifted from incremental improvement to long-term viability.

Iterating within the existing framework would have meant working around constraints rather than resolving them.

Iterate on existing solution

  • Faster in the short term
  • Lower immediate effort
  • Continued dependency on framework
  • Key functionality locked behind libraries
  • Required learning an unfamiliar codebase

Rebuild internally

  • Higher long-term return on investment
  • Full control over UX and accessibility
  • End-to-end system understanding
  • Removal of framework dependency
  • Foundation for reuse across services

The decision was to rebuild.

Drivers

  • Control over user experience
  • Flexibility for future development
  • System transparency and maintainability

Design principles

Research included public sector chat approaches such as GOV.UK patterns and other local authority implementations.

The GOV.UK Design System was used as a foundation, adapted for an embedded conversational interface.

Examples of the button variants used in the Nia interface, showing primary and secondary visuals.

Principles

Simplicity over complexity

  • Focus on clear interactions
  • Reduce visual noise
  • Prioritise clarity

Accessibility by design

  • Built in from the start
  • Not treated as retrofitting
  • Broad usability considerations

Familiarity and consistency

  • Aligned to public sector patterns
  • Reduced learning effort
  • Consistency with Family Hub

Reusability

  • Shared components across chatbots
  • Avoid one-off implementations
  • Support scalable delivery

User experience (UX)

The rebuild addressed core usability issues in the original system.

Key issues

  • Inconsistent colour contrast
  • Variable message structure
  • Limited interactivity clarity
  • Poor mobile consistency
  • Unclear keyboard focus states
  • Inconsistent response formatting

Improvements

  • Standardised message structure
  • Consistent interaction patterns
  • Mobile-first design approach
  • Refined system prompt behaviour

Accessibility considerations

  • Consistent colour contrast
  • Keyboard navigation support
  • Clear focus states
  • Screen reader compatibility improvements
  • Reduced cognitive load

The result is a predictable and accessible experience across devices.

Interface and visual design

The interface was redesigned using GOV.UK Design System principles adapted for chat-based interaction.

Chat interface showing the end chat pattern

Design approach

  • Translate design system into conversational UI
  • Prioritise readability and hierarchy
  • Minimise visual complexity

Key decisions

Message layout

  • Clear separation of speakers
  • Consistent structure
Examples of the chat buttons in the two contexts - user and system

Typography and spacing

  • Optimised for readability
  • Reduced cognitive load

Colour and contrast

  • Accessibility aligned
  • Functional use of colour only

Interaction states

  • Loading and response states defined
  • Clear system feedback

Mobile responsiveness

  • Mobile-first design
  • Consistent cross-device behaviour

The interface supports conversation rather than competing with it.

Architecture and development

The rebuild focused on a lightweight and maintainable architecture.

The chatbot was built using Next.js with the Vercel AI SDK and Azure services for search integration.

Stack

  • Next.js for frontend and API routes
  • Vercel AI SDK for chat handling and streaming
  • Azure AI Search for retrieval

Architecture decisions

Component-based structure

  • Reusable UI components
  • Separation of layout and logic

Embed-first approach

  • Single embed script for deployment
  • No dependency on web team changes
  • Simplified rollout process

Prompt control

  • Standardised system prompts
  • Consistent output behaviour

The result is a simple, maintainable system with low deployment overhead.

Azure integration

The system uses Azure services for search and hosting.

Components

  • Azure App Service hosting
  • Azure AI Search index

Flow

  • User submits query
  • Query sent to search index
  • Relevant results returned
  • Response generated conversationally

Rationale

  • Uses existing structured content
  • Avoids duplication
  • Maintains trusted data source
  • Scales without complexity

Building for the future

The rebuild shifted the system from a single chatbot to a reusable capability.

Approach

  • Reusable component library
  • Embed-based deployment model
  • Config-driven behaviour
  • Separation of content and UI

Outcomes

  • Faster rollout of new chatbots
  • Consistent user experience
  • Reduced development overhead