Artificial intelligence (AI) is increasingly being used to tackle some of society’s most complex challenges—and homelessness is no exception. Across Canada, emerging AI tools such as neural networks, large language models (LLMs), and federated learning are helping cities and nonprofits better understand, predict, and intervene in homelessness with greater precision and humanity.
One of the most impactful uses of AI in this space is predictive modeling. In London, Ontario, a system known as the Chronic Homelessness Artificial Intelligence model (CHAI) was developed to identify individuals at risk of experiencing long-term homelessness. During COVID-19, CHAI achieved 93% accuracy in forecasting chronic shelter use, allowing city workers to prioritize early intervention and supportive housing placement before individuals slipped into crisis (Reuters).
Similarly, researchers at Carleton University partnered with the City of Ottawa to develop a machine learning tool that flags individuals likely to enter or remain in chronic homelessness based on intake data and social risk factors. This initiative empowers social workers to deliver proactive, rather than reactive, care (Carleton Newsroom).
Accuracy alone isn’t enough—interpretability matters. In 2020, researchers VanBerlo, Ross, and Booker created a deep learning system that not only predicts shelter use across Canada, but also explains why it made certain predictions. Using techniques like LIME, their HIFIS-RNN-MLP model maintains transparency and helps build trust between AI systems and frontline shelter staff (arXiv).
One major barrier to AI collaboration among housing organizations is data sharing. That’s why new projects in Calgary are leveraging federated learning, a privacy-preserving method where agencies train shared AI models without directly exchanging sensitive client data. This approach promotes equity by allowing even smaller nonprofits to benefit from AI insights without compromising client confidentiality (arXiv).
AI is also being used to improve decision-making across entire cities and provinces. The InnSoTech platform, developed by HelpSeeker, AltaML, and A Way Home Canada, combines real-time data from over 134,000 social services to forecast trends in homelessness, domestic violence, and suicide. By integrating housing, health, and economic data, the tool helps governments allocate emergency funds where they are most needed (Digital Supercluster).
While predictive models handle the when and who, LLMs like ChatGPT can assist with the how. AI-powered chatbots could help individuals:
Navigate complicated housing applications
Understand their rights and available services
Translate information into accessible language
Offer non-judgmental emotional support before professional help is available
These tools can be embedded in outreach apps, public kiosks, or community center websites to enhance access.
All AI deployments in homelessness must be grounded in ethics, privacy, and human-centered design. As noted in a 2025 report by the Centre for International Governance Innovation, the misuse of AI—whether through bias, surveillance, or techno-solutionism—can harm vulnerable populations unless strong safeguards and regulatory oversight are in place (CIGI).
AI will not end homelessness alone—but it can make our responses smarter, faster, and more compassionate. By investing in responsible AI applications and governance frameworks, Canada has the opportunity to lead globally in using technology for good. Our nonprofit remains committed to supporting innovation that respects privacy, promotes equity, and upholds the dignity of every individual.
This post was brought to you by AI By AI Canada Center, a nonprofit advocating for responsible AI use across Canada.