Because frontline work is already hard enough.
“You’ve helped me feel validated as a human being with respect and support.”
— From a participant at a Homeless Connect Toronto event
Send us a note ✨ See Lemy in actionOntario alone has 60,000 community programs and services in databases like 211 — the community services network operating across every US state and Canadian province — systems that don’t understand your context and don’t remember your last search. Navigating this friction can disconnect workers from the human being across the table.
We know that broken search isn’t the hardest part of nonprofit frontline work. But it’s a part we can do something about.

An AI-powered resource discovery tool — live and in use at a Toronto community organization. One search bar. Natural language. Verified results from three trusted sources, searched in parallel. Connected to Ontario 211’s live API — 60,000+ verified services searchable in real time.
Four trusted tiers. One search.
Each organization’s internal resource list — contacts, intake tips, relationship notes that exist nowhere in 211 — becomes the highest-priority search tier. Workers see their own vetted resources before anything else.
Lemy searches only verified data: HCT’s curated directory, Ontario 211’s 60,000+ services via live API, and curated community listings. AI reasons about meaning without inventing facts. A worker can’t hand a client a referral that might be fictional.
Workers annotate services with one tap — “good intake,” “long wait,” “ask for Maria.” That knowledge accumulates across all workers, survives staff turnover, and makes the next worker better. Individual expertise becomes organizational memory.
Every search, save, and referral generates data the sector has never had — which services are overwhelmed, which populations can’t connect, where the gaps are. No extra data entry. The intelligence is a byproduct of the work.
Toronto is the pilot. The model scales to any city with 211 — which is every major city in the United States and Canada.







The community testbed is funded and underway. Four frontline community workers at a Toronto organization are co-designing Lemy to fit their actual workflow. The program manager is sharing the organization’s internal resource list — contacts and insider knowledge built over years that exists nowhere in 211.
Lemy searches the organization’s private resources alongside HCT’s Everyday Connect listings and Ontario 211’s 60,000+ verified services. Workers see their own knowledge first, verified public data second.
Lemy is portable. What we prove in Toronto can go anywhere there’s a 211.
The testbed produces published findings — search quality data, worker feedback, annotation patterns — regardless of outcome. The evidence base for the proof of concept.
Frontline community workers hold communities together, sitting with people during some of the hardest moments in their lives. They listen, search for help, make referrals they hope will land. Nobody captures what they know. When they leave, it leaves with them.
The systems they depend on weren’t built for them. 211 is slow and absent practical knowledge. General-purpose AI hallucinates. Workers deserve tools that search verified data, preserve what they know, and reveal what communities need.
This is more than a technology project. It is about amplifying what happens when a person in crisis sits down across from someone who can help — by finding what they need, instantly. And leaving time for the compassionate conversations that matter most.
Every city in the US and Canada has 211. What we prove in Toronto can go anywhere. The tool is free to organizations, always.
Time enough to listen — which is what the person across the table will remember.
Reduced administrative friction. Less time navigating broken systems means more capacity for the human work they came to do. Not a cure for burnout — but one less thing adding to it.
A worker who has time to listen, who knows where to send you, and whose referral carries weight because a real person made it. That’s what dignity in a moment of crisis looks like.
We can’t create housing or meals. We can make sure the systems aren’t standing between a worker and the person who needs them.
We don’t claim to solve homelessness. We claim to make the people trying to solve it more effective at their work.
Every city throughout North America has frontline community workers navigating fragmented systems. Most have 211. If this works in Toronto, the model is portable — to any city, any 211, any frontline organization.
Better technology won’t solve homelessness. But it can amplify what happens when a person in crisis sits down across from someone who can help — by finding what they need, instantly.
The community testbed is funded. The proof of concept — production AI search, Cohere integration, second deployment — requires $25,000. Any contribution moves this forward.
Support This WorkIf any of this is relevant to your work, we’d welcome hearing from you.
Or email directly: john@nuample.org
US 501(c)(3) · Ontario Not-for-Profit