Resilience Communication
Always On Call- Carol Chen, Gianna Feng, Alen Wong, Adrian Bai
AI call line that survive in high volume call flood
Life is put into danger, a rational cost-benefit analysis would make our application indispensable for government to procure.
Conclusion
During Hurricane Harvey raging Houston in 2017, there are more than 56,000 calls in 17 hours, against about 8,000 on a normal day. That’s roughly 7 days of calls in under one. The queue reached 120–250 waiting callers.
Situation of LA
LA fires 2025, LAFD
Thousands of calls with up to 100 in the queue, about one every 30–60 seconds all night. The operations center phone system was flooded until it stopped working.
Wildfires happen all the time
People actually die when they can’t dial to the help they need
What we offer:
How our app work for everyone, with ease
Four callers we designed around, and the languages LA speaks at home.
South Korean tourist
Pain point Doesn’t know local evacuation routes or emergency terms.
Need Clear directions in her language, using landmarks she can see.
Older adult who doesn’t speak English
Pain point Alerts are English-only and the instructions are complicated.
Need Accurate spoken translation and a clear summary of what to do next.
Delayed business traveler
Pain point Keeps searching for updates and isn’t sure when things reopen.
Need Short, reliable updates and a status summary he can share.
Person without stable housing
Pain point May have no data plan, and shelter availability changes fast.
Need Works on any phone, needs no address, and points to verified shelters.
Language spoken at home
LA County residents aged 5 and over
Source: U.S. Census Bureau, ACS 2020–2024 5-year, table B16004
Language, Info, tech savviness – Solving all at once
How a call flows through the line
Retell AI conversation flow, from the first “hello” to a confirmed next step
01
Understand the caller
Care Every answer, transfer and text stays in the caller’s own language.
02
Route by case
Safety Life-threatening at any point → 911 guidance. High urgency → live coordinator.
03
Hand off to the call center
Trust Says “saved” only after the backend returns a reference code, then texts a summary.
Guardrails
Never promises help is on the way · 911 override at any point · Asks only what’s needed · Every failure has a fallback: retry, callback, text or a person
Coordinators see who needs help first
Night Desk: the supervisor console every AI call feeds into
Live map Cases over fire perimeters, clustered
Ranked queue Most severe first; kids and 65+ flagged
Caller profile People, language, location, wait
New-info flags New details from repeat callers marked
Severity score, 0–100
U = min(100, N+H·c+T+V+P+A+R)
Max points per factor. Hazard is scaled by location certainty (×0.6 unknown to ×1.0 exact address).
Others help 911 desks. We answer the surge
Where today’s tools stop, and where the Resilience Line picks up
Player | What it does | Who talks to the caller | Where it stops |
Aurelian | AI voice agent for 911 non-emergency lines; live in 12+ dispatch centers ($14M Series A, 2025) | AI, for routine reports (noise, parking, theft) | Built for everyday police calls, not disaster surges |
Carbyne | Cloud 911 call-handling platform with transcription and translation; ~300 sites ($100M raise, 2025) | Human call-takers, helped by software | Every call still waits for a free person |
Everbridge | Mass notification: alerts pushed out by text, email and voice | No one: alerts are one-way | Can’t take a call or hear what people need |
Generic voice bots | Customer-service AI agents (e.g. PolyAI) | AI | No emergency guardrails, urgency triage or coordinator view |
Resilience Line | Multilingual AI front door for disaster surges, plus the Night Desk console | AI first; a human for high urgency | Scores urgency, 911 override, hands off with full context |
Sources: TechCrunch (Aug 2025) on Aurelian; Carbyne press release (Jul 2025).
Calm days pay for disaster days
One AI line, two kinds of paying customers, and grants to get the first pilot live
Cities & counties
Who Emergency management offices, 311 and 911 centers
Pricing Annual platform license; surge minutes billed only when a disaster hits
Why they buy Life-safety capacity without hiring for the peak
Airlines, utilities, transit
Who Companies whose call volume spikes on every disruption
Pricing Per-minute usage plus platform fee (airline model: $0.40/min, $180K/yr)
Why they buy $11.7M first-year savings for one airline
Grants & pilots
Who DHS Science & Technology, FEMA preparedness grants
Pricing Non-dilutive pilot funding; open prizes near $1M
Why they fund Pays for the first city pilot in LA
Flywheel
Enterprise calls keep the agent live and improving every day, so a city gets a proven surge line the moment disaster hits.
Proof point: an airline saves $11.7M in year one
Our cost model for one airline handling 5 million calls a year
74%
of incoming calls handled end to end by AI
$22.4M
labor cost avoided each year
$10.58M
annual AI voice cost, usage plus platform
109.4%
first-year return on investment
Annual customer service cost
US$ millions, recurring
Net result
$11.69M first-year net savings after a $100K one-time setup
Two twists that widen the opportunity
A federal program built for this problem, and a design that travels
TWIST 1
DHS funding fit
The DHS Science and Technology Directorate finds innovators to solve homeland security challenges, which is exactly our problem.
Next step Submit to the DHS S&T open call.
TWIST 2
Built to go global
Call floods and language barriers happen in every city, so the line adapts anywhere.
To consider Where labor is cheaper, Retell AI fees save less, so pricing must adapt.
Always on call
When every line is busy, no one in LA should be left without an answer in their language.
The problem
Call lines collapse when disasters hit, and many callers don’t speak English.
Our answer
A multilingual AI line that scores urgency and hands the right cases to people.
The model
Enterprise customers fund the line every day; cities get surge capacity on demand.
Thank you Always On Call · Carol Chen, Gianna Feng, Alen Wong, Adrian Bai