Chalmers, an AI chatbot helping people facing homelessness

Ample Labs is committed to solving invisible and transitional homelessness. At Ample Labs, we harness the power of technology and deploy it where it fits best to address and prevent homelessness. Since last November, our chatbot Chalmers (try it here) has helped thousands facing homelessness, by letting them find resources and services in Toronto much more efficiently. It will positively impact people around the world by connecting those with needs to services and programs faster, and those facing homelessness will have the potential to prosper in our community in the future.

We aim to help individuals who are or at-risk of becoming homeless to find social programs and services. The direct impacts from using Chalmers are tremendous time savings, reduced stress in dealing with legacy directories, and being able to re-allocate their time to secure housing and employment.

In Toronto, homelessness is growing rapidly in the streets and beyond, especially more challenging for youth, with much higher housing prices. People who face homelessness may be couch-surfing, may have been kicked out for coming out to their family, may be facing eviction, or could be new refugees. Based on our estimates, as many as 180,000 Canadians can be facing homelessness at any given time today.

Over 90% of people experiencing homelessness have access to devices (smartphone/computers), and use search to locate resources. However, on average, people we talked to revealed that the journey can take 20+ minutes, if it does not end up completely fruitless. It also adds to the stress of being in a tough situation and aggravates mental health issues. This is why we built Chalmers, the AI chatbot to help directly help people experiencing or at-risk of homelessness in Toronto.

The target population we are serving, are individuals who are at-risk of or have recently fallen into homelessness. These are people who can be best exemplified as living on friends' couches, staying in temporary shelters, or sleeping in their cars. They may have part-time employment, and they are actively looking long-term, stable housing.

We have taken an approach that is considered novel in tech, but what we consider critical to build a product for Social Good: we have been and have planned to continue co-designing our chatbot with our target population since the start. This allows us to build our community at the core of our product, and to make sure that it best serves the interests of our users. We also compensate individuals who participate in our co-design sessions, which values them for their time and helps them learn about how to use tech to navigate homelessness. On an agency level, we work with numerous homeless service providers and municipalities across Ontario for feedback. We also collect feedback in-app, which allows us to act in an agile manner regarding suggestions. Over time, we will continue to build on our strengths and learning, to scale our product.

Traditionally, people facing homelessness have to use a resources directory (e.g., 211 Toronto) or online search extensively, just to find a single service or program they need. Things like a hot meal in the cold, a place to stay the night, and immediate yet non-urgent mental health help. Our answer is a chatbot based on artificial intelligence, which allows us to streamline the resource-searching journey and direct people to what they need in real-time. We gave the chatbot the name “Chalmers”, like a person, as well as some personality traits to make it more approachable.

We begin with a direct, simple, and intuitive conversational user interface. Our users can either tap through suggestions or type their request directly into the chatbot, instead of struggling through nested menus. We also built Chalmers with a human-centric language that makes the experience not unlike texting a close friend.

Chalmers is built with differences in mind. We learned that youth, women, and people who identify as LGBTQ+ have very distinct needs, so we added ‘age and gender options’. This helps people stay safe in drop-ins that are set up specifically with their backgrounds and experiences in mind.

We designed Chalmers with technologies that we use daily, so it could be the most useful for the majority. It relies on GPS positioning and mobile data, so we can give noticeably more useful results than a directory, which usually shows a map with non-personalized results. In comparison, Chalmers returns a single result that is nearby and available now. Chalmers takes in real-time inputs, which allows it to respond to users needs, feedback, and even feelings almost instantly, instead of returning a blank results page or no-results. This way, Chalmers can deliver human-like empathy and receptiveness that legacy platforms cannot.

We are currently building more advanced capabilities for Chalmers via Natural Language Processing and Machine Learning. Chalmers can learn from every interaction, and use them to prioritize app requests. We envision Chalmers to be able to handle more complex and granular questions, and triage responsively with frontline workers.

Last but not least, the use of chatbot technology comes with much less stigma, and more privacy and control. By giving respect for people and their conditions, Chalmers helps them feel more like a part of our community.

There are a few directories in North America that are designed to help people facing homelessness, but we are the only one to take a data-driven and human-centric approach in problem-solving. 

The first thing that sets us apart is our commitment to making positive social impact. We have conducted research with people we are trying to serve. Through this process, we learned that we are not just building another directory that could only make the journey more confusing; instead, we want to present a solution that is desirable to use, technologically robust, and improves over time. Eventually, we also want something that can connect people facing homelessness directly to helpful professionals. Our chatbot, thus, is built with ease of use, technology, and learning abilities in mind. Analytics, AI, and Machine Learning can help us more effectively capture changing trends in homelessness and provide valuable insights that otherwise remain unseen by service providers. This helps us, service providers, and academia to become more informed, and better design future systems.

Chalmers also uniquely serves at-risk youth, who are more likely to have access to mobile technology, and are likely to go first to their phone for answers as to what can help.  

We will always put users at the core of research and ideation. Instead of positioning ourselves as a tech-first company, we put much more care and thoughtfulness in designing our product, which is critical in what makes social tech and innovations successful.

The foundation of Chalmers is Amazon’s Lex platform, which allowed us to create off-the-shelf chatbots on our own. With our stack, our custom components include a natural language processing engine (which allows Chalmers to accept pre-configured inputs and direct user inputs) and a language engine (which enables multiple languages in the chatbot). We also proprietarily use a few real-time data APIs (application programmable interfaces) that plug directly into 211 Toronto and Ontario’s database of homeless services. We also maintain a regularly updated database consisting service provider information to complement the use of APIs.

The conversations that happen in Chalmers are loop-based. It filters keywords to provide a response, and uses geolocation through GPS and data and time to accurately narrowing down to a single result. 

On the outside, Chalmers takes the form of a responsive web app that can scale according to screen size. Being a web app also makes it cross-platform compatible and highly accessible and useable across older or cheaper devices.

We are currently considering adding a few features: caching, which allows Chalmers to remember users’ accommodation preferences; user rating and tags, with which our users can report on the safety and quality of service locations; and last but not least, the ability to access emergency numbers, mental health numbers, and central intake information when offline.

We expected our solution to address the problem, because as a chronic social issue, homelessness calls for new perspectives and cutting-edge, innovative approaches. Homelessness also changes over time; what worked before will likely not work in the future. The premises of Chalmers respond to changing demographics, and the call for personalized journeys, which solves every person’s problems one at a time.

Directories, how-to-guides, and online search present a digitized but disconnected present, which really, is just glass half empty. A person can easily get lost in a system this complicated, especially if they had never had the experience before. Additionally, a lot of people remain chronically homeless, because the system is difficult to navigate. A chatbot that feels like someone trustworthy and personable can help out a person in these stressful situations in ways static directory systems cannot.

Meanwhile, frontline workers help people in need literally one after another. Chalmers can give them the ability to shift their capacity towards high-priority, high-urgency requests, as it can answer the simpler and more frequently asked questions with increasing proficiency.

We have a total of 3,000 all-time users that have triggered over 6,000 sessions. Right now our monthly active users vary between 700~800 people with 82% first-time users and 18% recurring users. We are seeing a steady 48% MoM (month-over-month) growth over the last 8-months.

To date, we've been able to make over 7000+ recommendations in total, about 4500+ of them are free meals, 900+ overnight shelters, 800+ clothing banks, and 700+ drop-ins. 

A key metric we track is recurring users, and right now we are at 20% recurring. We hope to double this number by end of 2019, to reach 40% recurring users. We hope to reach 10% market penetration by the end of 2019, or about 15,000 people facing invisible or transitional homelessness in Toronto.

By March 2020, we will have a extremely robust product and will be ready to scale outside of Toronto with evidence and case studies from different regions, alongside academic and secondary research. Our own research team will be conducting studies with Canadian universities to provide data to speak to mobile phone usage amongst the invisibly and transitionally homeless population in Toronto and in Canada. 

By next year, we will be ready to scale to other similar metropolitans. We see New York City, San Francisco, and Los Angeles as other cities with similar profiles in terms of invisible and transitional homelessness, where we can pilot our chatbot after we have enough evidence to prove this works in Toronto. In New York City, we hope to reach up to 200,000 people when we scale there; in San Francisco and Los Angeles, we hope to reach 20,000 and 90,000 respectively. These numbers are estimated based on 10% market penetration.

In five years, we hope to make our chatbot available in up to 20-25 cities that experience homelessness as a major issue. 

For this year, our major barrier is a lack of funding. Without funding we can't hire talent that means our product is moving at a very slow rate in terms of development and reaching our target market. The second barrier is learning how to effectively reach our first goal of 10% market penetration within Toronto that means knowing where the invisibly homeless may be and targeting those offline & online places so we are effective in reaching our target market. The following year when we've identified product-market fit and have validated that through hitting our target market would be scaling our solution. One of the challenges within scaling is working with a back-end service providers whom we can plug in the data/service to.

The first in addressing funding is to get into as many incubators and accelerators as we can to accelerate our growth and find funding opportunities. The second, in reaching 10% market penetration will be through various ways of online and offline marketing and experimentation such as paid socials, affiliate marketing, crowdfunding, and activation campaigns. The third would be to build those partnerships with back-end service providers early on so we have access to a wide database of services that we can work with when looking at scaling to other cities.

At the moment, we have three full-time employees, 20+ active volunteers who contribute between 5 and 15 hours per week. Since our inception we have had 177 unique volunteers.

We have a extremely diverse technical team. Our core team is our product team which is comprised of our Director of Product, a scrum master, 1 full-stack developer, 1 back-end developer, 1 front-end developer, 2 UI/UX designers and 3 UX researchers. Aside from that we have a full-stack marketing team comprised of our Director of Marketing, creative director, graphic designer, copywriter, social media coordinator, growth hacker, and in-house production team that does photo & video. We also have a research team comprised of our Director of Strategy which oversees the research, research manager, coordinator and 4 research assistants. 

There are also people we call Community Leads, which are individuals with lived experiences of homelessness that partake in our on-site demos, lightning talks, pitches and presentations and they inform the product development and our marketing, brand and positioning. It’s really important that our team is comprised of technical individuals that are great at what they do but also individuals that have experienced the problem first-hand so we are never designing in a black box. As the founder, my undergrad was in graphic design and I’ve worked as a designer in the industry before being a UI/UX designer for various tech start-ups in Toronto so I’ve worked in web and mobile app development prior to starting Ample.

Since our inception, we have been able to demo Chalmers with 42 other not-for-profit organizations, city offices, and corporate partners. 

Specifically, we have partnered with 211 Ontario, which holds about 300,000 service provider information in all of Ontario and that is what we are using on our back-end to pull up information. This is important for us as it will be very easy to expand outside of Toronto whenever we are ready.

The second is we’ve signed a LOI (letter of intent) with a national charity (Raising the Roof) that is interested in taking the chatbot and building a specific set of features around evictions preventions that is worth $300,000 CAD. 

We also are partnering with a smaller city in Ontario (City of Barrie) for a paid six-month pilot of Chalmers starting in August 2019. We will also be piloting these features for over a year to collect data and feedback on its effectiveness.

We've also been able to run 3 small-scale pilots of 2-weeks and 2-months to pilot Chalmers at specific city shelters to test the effectiveness of Chalmers with front-line staff to see if it delivers value. 

We are proposing a business model where we partner with other service providers where we offer Chalmers as a software-as-a-service, which helps them triage and create efficiency by automating most intake and request processes.

Our target customers are social services, municipalities, non-profits, charities. There are over 500 service providers across Toronto, 7000 in Ontario and 30,000 in Canada that could utilize what we offer to improve how they offer help towards people facing homelessness. On top of these organizations that are directly in Toronto, we plan to expand and scale across multiple regions. Our technology product has the benefit of scaling virtually with no requirement of physical residence.

Our value proposition for customer is first and foremast, a conversational AI that allows them to dedicate their staff to more urgent or in-depth issues and optimize human resources. For example, it could save time for frontline staff. Our chatbot will also allow their organization to service more clients within the same time. As a chatbot, our platform will be available 24/7 and it fills in the gap where human staff are not available. Lastly, our chatbot can help many more people than individual staff can by quickly answering commonly asked questions.

Our path to financial sustainability is based on our SaaS B2B business, continuing application to grants, and raising capital through participation in tech and nonprofit accelerator programs. In the long run, we believe our sustainability will be supported by our SaaS B2B business as we scale beyond Canada and into the United States.

The prize means we can expand our full-time product team and take our product further. We can scale our impact far beyond our current offerings and beyond Toronto. We are working hard to validate product-market-fit, and then take our solution and deploy it globally, starting with three major cities in the United States: San Francisco, Los Angeles, and New York City. The outcome we envision is impact being scaled 10x in terms of times people helped, at 5x to reach that goal. Funding will, for a long time, be the key to accelerating our growth and maximizing reach.

But we are seeking more than just funding - we are also seeking the best talents and mentors across the world, to join and guide us in this journey of innovation and uplifting of other members of our communities. We also would love to work with top research universities, building new ideas on sound research, and donating our analytics for studies. 

Ultimately, we need to bring our communities together again through exchanges of ideas with both the public and agencies/municipalities.To do that, we need to have open conversations with social service agencies across the world. We need to listen to what people have experienced and learned over the years, and build upon our knowledge base of homelessness, to better educate ourselves who are capable of helping. We want to spark a conversation with the general public to show them why and how solving homelessness can use all of our help.

There are several types of organizations we would love to work with.

First and foremost, we want to branch out to the tech giants: Google, Facebook, Microsoft, PayPal, and Airbnb. As large tech-driven enterprises, they can either supply us with resources that can directly accelerate our growth, or partner with us for innovations in technology that can be used in our chatbot or other projects.

As mentioned earlier, we would also love to partner with top research universities: MIT, New York University, Columbia University, University of Toronto, Stanford University, UCLA, and many more. They can help us study how homelessness has changed over time and show us how it varies geographically, which can really help us rethink how we build our product for each new location we expand into. In return, we can share some of our user analytics if it could be any useful for the researchers and scholars. This collaboration will benefit us both.

Lastly, we hope we can gain more exposure in the mainstream press, which will be the precursor for our open call for honest discussion for homelessness with everyday people who have the capacity to help. We would love to start a organized campaign with any major press who are interested.

We would like to apply for the AI Innovations Prize as we are already deploying AI in our chatbot, but we want to take it a step further. We want to develop more complex AI systems that will be able to handle requests and respond to questions in a more human-like manner, and with characteristics like empathy and kindness. We want to build a chatbot platform that not only makes resource searching easier, but also a pleasant and more inclusive experience for all of us.