Why does CHV have an artificial intelligence (AI) chat tool? 

Our new AI-supported chat solution expands our reach while providing timely, accessible information to young people who prefer not to talk on the phone or don’t have a Covenant House in their community.

This new tool reflects CHV’s commitment to innovation and accessibility, using the technology youth use to meet them where they are and to offer support that’s safe and responsive.

How does it work?

Embedded on CHV’s website, the chat tool provides quick answers to common questions about CHV services so youth can identify programs that are right for them.

It works similarly to a web search tool, using generative AI technology to provide search results in a conversational way so that is intuitive to use, interactive, and responsive. CHV monitors the chat tool regularly and reviews interactions with users in order to implement upgrades or temporarily shut down the service as needed.

What happens when a young person using the chat is in crisis?

This tool has been designed to protect the safety of young people. At the start of each chat, youth are explicitly asked if they are in crisis. If so, they are handed off to a CHV youth worker trained to support a young person in crisis.

The tool is able to identify when a youth may be in distress and connects them directly to a youth worker. A youth may also ask to speak to staff at any point.

Does this AI chat tool replace a person?

No. At the heart of CHV’s work is human connection. This tool aims to increase our capacity to serve young people while at the same time facilitating connections to staff when needed or requested.

Because it is able to answer routine questions, the AI chat tool allows staff to focus on and be responsive to the needs of young people currently using our programs and services. Our team is there to intervene when a young person using the tool expresses distress or crisis, or requests to talk to a person.

Why is it called Lumi?

The AI chat tool is meant to be approachable, easy to use, responsive, accurate, helpful, safe, and knowledgeable. The name Lumi was selected to reflect that core purpose, as well as reflecting traits of the young people CHV serves: resilience, strength, dignity, and independence. The name was chosen through a process that included consulting with youth, staff, and volunteers at CHV.

Lumi is a play on luminate or illuminate, providing a light for the path forward. To illuminate something is to make it understandable. Lumi aims to do the same for youth seeking information about services and supports available to them.

How was it built?

The AI chat tool was developed by CHV in consultation with youth and staff, and in partnership with the technology experts at Slalom.

Slalom is a global consulting company with expertise in AI services and teams in Vancouver, Toronto, Calgary, and Montreal. In addition to partnering with international organizations, Slalom has also provided technology solutions to non-profits like Make-A-Wish Canada and National Kidney Foundation.

The tool was built using Microsoft’s Azure Services and Open AI models that enabled CHV and Slalom to access the same underlying models as ChatGPT but customise them for our own use.

How does CHV protect users’ privacy?

Users of the AI chat tool are first asked whether they are under the age of 19. If they answer yes, they are automatically referred to a youth worker at CHV. Aside from this additional layer of protection, no personal information is required when using the chat tool.

CHV’s chat tool is built on what’s called retrieval augmented generation (RAG) architecture to generate responses. This means it fetches information from documents we’ve provided to ensure accuracy. The application and all information are hosted on CHV’s Microsoft environment, which means no information that is shared during the chat is provided to train the underlying AI model as other generative AI tools might. The software is secure and privacy protected.

Logs are stored for one year, after which CHV will remove identifying information unless required for legal purposes.

How does CHV ensure the model is safe for users and implemented responsibly?

CHV has completed a Privacy Impact Assessment, a Human Rights Impact Assessment, and a legal review to ensure responsible operation of the AI chat tool.

Additionally, CHV staff will review chat logs on a regular basis to ensure that it is providing appropriate and/or expected responses. CHV will disable the tool if there are any concerns about its operation.

What about the environmental impacts of AI?

Before building the AI chat tool, CHV undertook an assessment of broader environmental, societal, and ethical considerations of AI. This assessment led to the decision to use an AI platform provided by Microsoft, which has stated its commitment to be a carbon negative, water positive, and zero waste company by 2030 and joined the Government of Canada’s Net Zero Challenge to develop facilities and operations that have net-zero emissions by 2050.

Ultimately, CHV aimed to balance AI’s potential for supporting youth with the tool’s known environmental footprint.