One front door to every helping hand in the city
A community resource hub run by a city and its nonprofit partners. Residents describe what they need in their own words and find help that is open, nearby and confirmed. Nonprofits keep listings current in seconds. City leaders see where needs go unmet.
A city’s safety net is mostly run by nonprofits, and residents usually find them by word of mouth, outdated PDFs or a dozen separate websites. Front Porch puts that net in one place and keeps it accurate.
This concept extends patterns from my church website work: a trimmed dashboard for the people who keep information current, an email path for the people who won’t log in, and AI that helps but never gets the final word.
Help exists, but residents can’t find it, and listings go stale because updating them is no one’s full-time job.
A shared resource hub with three surfaces: resident search, a partner portal, and a city insights view.
AI only points to verified listings. It never decides who qualifies, and it never asks about immigration status.
Help is out there. Finding it is the hard part.
A resident facing an eviction notice has to know that “rental assistance” is the right term, which agency runs it this year, whether they qualify, and whether the office is open today. Many look on a phone, under stress, sometimes in a second language.
Resource directories exist, but they decay. A food pantry changes its hours, a program runs out of funds, a phone number changes. The listing stays the same, and a resident takes a bus to a locked door.
Meanwhile, city leaders can’t easily see the gaps: what residents are searching for, in which neighborhoods, that no program currently offers.
How might we connect residents to help that is real, open and nearby, while showing the city where help is missing?
This framing is based on common patterns in public-service directories. Before design moves forward, it would be tested with residents, nonprofit staff and 311 agents in the pilot city.
Four groups, one shared map of help
Someone who needs help now
Often on a phone with limited data. May be in crisis, may not speak English first.
- Find what’s open today
- Know what to bring
- Stay anonymous
Librarians, caseworkers, 311
Help residents in person or by phone, many times a day.
- Search fast during a call
- Text a list to a resident
- Trust the information
Program coordinator
Keeps their organization’s listing current on top of a full workload.
- Update hours or capacity in seconds
- Mark a program full or paused
- Get referrals that fit
Mayor’s office, community partnerships
Funds programs and answers to residents and council.
- See unmet needs by ward
- Spot stale or inactive listings
- Report impact without exposing anyone
Three surfaces, one source of truth
Resident search
- Describe the need in your own words, in your language
- Results ranked by open now, distance and freshness
- Eligibility in plain words, what to bring
- No account, nothing tied to identity
Partner portal
- Simple dashboard for each organization
- Update by email: “We’re closed Friday”
- Monthly one-tap “Still accurate?” check
- Waitlist and capacity status
City insights
- Top needs by week, ward and language
- Searches with no nearby results
- Listing freshness across partners
- Grouped data only, never individuals
Ask in your own words
AI turns everyday language into the right service categories, then shows only verified listings. Try four situations. The city, programs and details are fictional.
Keeping listings fresh
Accuracy is the product. Every listing shows when its provider last confirmed it, and the system makes confirming nearly effortless.
Shown normally with “Confirmed X days ago.”
Partner gets a one-tap “Still accurate?” email.
Ranked lower. Residents see “Call ahead to confirm.”
Removed from results. City staff follow up with the partner.
Residents see this listing today. If nothing changed, one tap keeps it at the top of results.
Or reply with changes in plain words, like “Closed Thanksgiving week” or “Now open Fridays too.” You’ll get a preview to confirm.
Why email again
Most small nonprofits have no web staff. The same pattern from my church platform concept applies: an approved sender writes in plain words, AI drafts the change, and the sender confirms a preview before it goes live.
Quick status changes skip the form entirely. Replying “FULL” marks a program as at capacity until the partner reopens it, so residents stop arriving to a closed door.
See where help is missing
For the mayor’s office, the most valuable data is what residents look for and can’t find. Every number is grouped and anonymous.
Community needs · Last 30 days
Sample dataTop needs searched
Share of all searches. Bars scaled to the top need.
Unmet needs
Shown only when 20 or more searches share a pattern, so no one can be singled out.
Built to be safe for anyone
Residents seeking help are often at their most vulnerable. Public trust depends on what the system refuses to do.
Residents can search without signing up. Searches are never tied to a name, phone or device ID.
No questions about immigration status, ever. Eligibility notes come from the program, written in plain words.
It explains what a program says, then points to the provider. “Each program decides who qualifies.”
The AI can’t invent a program. Every result comes from a listing a partner or city staff member confirmed.
When a search suggests danger, emergency options appear above everything else, along with a quick exit.
Every listing shows its last confirmation date. Stale information is labeled, ranked lower, then hidden.
Designing for the hard moments
| Situation | What the system does | Why |
|---|---|---|
| Search suggests danger or self-harm | Shows 911 and 988 first, then local services; adds a quick exit and never saves the search | Safety outranks relevance |
| Program is full or out of funds | Shows “Waitlist” or “Paused” and suggests the next closest option | A full program shouldn’t look like an open one |
| Search in a language with fewer translated listings | Translates automatically, labels machine translation, and flags providers who speak that language | Honest about quality, still useful |
| Vague search like “help” | Offers six common needs as large buttons instead of guessing | A guess can send someone the wrong way |
| Two listings disagree (hours on a flyer vs. portal) | Uses the most recently confirmed source and notifies the partner | One answer, with a clear source |
| Shared or library computer | Offers “Clear my search” and never stores history locally | Privacy on public devices |
| Fraudulent or spam listing | New organizations need city verification before appearing | Protects residents from scams |
| Weak signal or old phone | Lightweight text-first page; “Text me this list” sends results by SMS | Help shouldn’t need fast data |
Failure states
An empty result becomes data the city can act on.
The directory works without AI. The AI is a shortcut, never a dependency.
Residents become the freshness check of last resort.
Silence never keeps old information online.
Public means everyone
Accessible contrast, keyboard and screen reader support, large tap targets.
Written around a 6th–8th grade reading level. “Help paying rent,” not “housing stabilization services.”
Key content reviewed by human translators; everything else machine-translated and labeled.
Text-first pages that load quickly on older phones and weak signals.
Text a list, or show a code a library or 311 navigator can pull up.
The body font comes from the U.S. Web Design System, built for government readability.
Why it works this way
Search in everyday words, with categories as backup
Residents describe situations, not service names. AI translates “can’t pay rent” into rental assistance, unemployment and food help.
Trade-off: language models can misread. Showing what it understood as editable chips lets residents correct it in one tap.
Freshness affects ranking
A confirmed listing two miles away beats an unconfirmed one next door. Accuracy is ranked above proximity.
Trade-off: partners who don’t respond lose visibility, which raises the stakes on making confirmation easy.
Anonymous by design
No accounts for residents. Insights use grouped counts with a minimum threshold.
Trade-off: the city can’t track individual outcomes. Partners can report completed referrals in aggregate instead.
City verifies partners, partners own their listings
The city approves who can list. Each organization owns and updates its own information.
Trade-off: onboarding takes a review step, but it prevents scams and keeps accountability clear.
Pilot in one ward first
Co-design
Sessions with residents, 10–15 nonprofit partners, librarians and 311 agents in one ward.
Directory pilot
Launch search with categories only, plus the monthly freshness check. Measure listing accuracy.
Add AI search
Turn on everyday-language search for half of visitors. Compare time to find help and wrong turns.
Insights for the city
Share unmet-needs reports with the mayor’s office and partners each month. Decide on citywide rollout.
What I would measure
Seconds from first search to calling, getting directions or saving a listing.
Share of listings confirmed in the last 30 days. Target above 85%.
Reports of closed doors or bad numbers per 1,000 visits.
Unmet needs that led to a new partner, grant or outreach effort.
Accuracy is a form of care
In a church or a city, the cost of a wrong answer lands on the person who trusted it. Across all three of my case studies, the most important design work happened in the same place: making the right information easy to keep right.
For AI in public service, the bar is higher than convenience. The system has to know what it doesn’t know, say so plainly, and route people to humans when it matters most.