The municipal hotline answers every resident, at any hour
A voice agent that answers residents in Hebrew and other languages, logs each request with an address and a topic, routes it to the right department, and can run fully self-hosted on the authority's own infrastructure.
The short answer
How does a municipal hotline answer every resident around the clock?
A municipal hotline covers every resident when an AI agent answers each inbound call in Hebrew — and in the other languages you configure, such as Arabic and Russian — and handles the repeat traffic itself: opening a service request, checking the status of an existing one, reception hours, and which department to approach. The agent captures the address, topic and caller details, routes to the right department or extension, and hands off to a human when needed. Authorities subject to information-security requirements can run it fully self-hosted on their own infrastructure, aligned to Israeli directive 5.43, so resident calls never leave the perimeter.
What happens today
- After a storm or an infrastructure failure the hotline is flooded with dozens of calls about the same incident, and residents sit on hold
- Urgent reports — a fallen tree, a safety hazard, a sewage overflow — arrive outside the hotline's hours
- A resident who is not a fluent Hebrew speaker gives up on the call, and their request is never logged
- The same request bounces between nine departments because nobody routed it correctly the first time
- There is no uniform record of what a resident was told, so an audit or an information request cannot be answered
How it sounds
Hello, you have reached the municipal hotline. I am the hotline's digital assistant. I can open a new request, check an existing one, or give information about municipal services. How can I help?
What the agent captures
- Resident's name and a callback number
- Location of the incident — street, house number and neighbourhood
- Topic of the request and the relevant department
- Description of the hazard or request, and when it started
- Urgency and whether there is an immediate safety risk
- An existing request number, when the caller is checking status
Common use cases
Incident surges and the hours with no operator
A municipal hotline lives between two extremes: routine, where most calls repeat, and an incident, where hundreds of residents dial at once. The agent answers every call in parallel with no hold queue, opens requests and routes them onward — and in the hours when the hotline is closed it keeps taking requests instead of leaving residents at a voicemail prompt.
- 24/7 answering on the hotline's existing number
- Simultaneous calls with no seat limit
- The same agent on WhatsApp for text enquiries
- Transfer to an operator or to a specific department's extension
Residents who do not all speak Hebrew
In a mixed authority, a resident who cannot explain themselves in Hebrew simply gives up — and their request is never logged. You can configure an agent that speaks with the resident in their own language, Arabic or Russian for instance, and maintain the scripts and knowledge base per language. The request is logged in the same system with the same structure, so the data stays uniform regardless of the language of the call.
- A dedicated agent per service language, with its own script
- A separate, approved knowledge base per language
- The same request structure in every language, for reporting and oversight
- Handoff to an operator who speaks the language when a person is needed
Resident privacy and sensitive data
Resident requests carry names, ID numbers, home addresses and sometimes sensitive detail about hardship or a neighbour dispute. Hebrew-aware PII detection marks and masks that data in the transcript according to a settings map the authority controls, and an unrecognized entity type is masked by default. Recording access is granted by role only, and an erasure request runs through a full anonymization flow — clearing the caller's details, masking the transcript, disabling the recording and writing an audit record with a reference number you can quote back to the resident.
- Identifying data masked in transcripts, configured by the authority
- Recording playback restricted to authorized roles only
- Full anonymization for erasure requests, with an audit log
- An audit record for every action, for answering oversight
A fully self-hosted deployment inside the authority
Public bodies subject to information-security requirements cannot always move resident calls to a public cloud. The system also supports a fully self-hosted deployment — every component, including the models, runs on the authority's own infrastructure and under its control, aligned to Israeli directive 5.43. The very same system is available as multi-tenant cloud, so you choose the posture by the sensitivity of the data rather than by a product limitation.
What the agent does not decide
The agent logs requests, gives information approved in the knowledge base and routes — it does not make administrative decisions, approve applications, determine eligibility or interpret a bylaw. Any such call is escalated to an operator or a department, and the reason is recorded. When the guardrails block an answer repeatedly, the conversation is escalated to a person automatically rather than leaving the resident going in circles.
Frequently asked
Can the system run inside the authority's own infrastructure?
Yes. There is a fully self-hosted deployment in which every component, including the models, runs on your infrastructure and under your control, aligned to Israeli directive 5.43 — so resident calls and transcripts never leave the perimeter. The same system is also available as multi-tenant cloud for bodies not subject to that requirement.
Can the hotline answer residents in Arabic or Russian?
Yes — you configure an agent per service language, each with its own script and knowledge base, and requests are logged in the same data structure across all of them. A resident who is not a fluent Hebrew speaker can complete the request, and it appears in hotline reporting alongside everything else.
What happens during an incident when hundreds of residents call at once?
The agent answers all of them in parallel, with no hold queue and no seat limit, and takes each caller's address and description — producing a picture of the incident's scale instead of dozens of lost calls. Requests defined as life-threatening or a safety hazard are escalated immediately by the rules you set.
What record remains after a call with a resident?
A recording, a full transcript, a summary, sentiment and an escalation reason if there was one — with identifying data masked per the authority's settings. Every action on the conversation is written to an audit log, and when a resident asks to be erased, the anonymization flow records who ran it and when, and produces a reference number you can give them.
Call the agent now
A real call, in Hebrew. Exactly what your callers would hear.
Have the agent call me
Leave a number — the agent calls you in under a minute.
Updated: