Redbo AI

Redbo Meshiv · source-grounded service and intelligence agent

An approved answer for the resident.A clear picture for management.

Redbo Meshiv answers residents, customers and employees on your website, on WhatsApp and on Telegram - only from sources you approved, and with a link to the source. At the same time it surfaces the questions that went unanswered, the gaps in your content and the trends in your service - so you know what is actually missing, not only what was asked.

Start without replacing your website or your systems · Your content stays yours · Hebrew, Arabic and four more languages

  • WhatsApp
  • Telegram
  • Website
  • 6 languages
  • 24/7
  • Source-grounded
A product, not a development project

Service starts immediately - not after months of implementation

The system is already built and running. No weeks of scoping, no bespoke development, and you start with no integration project. You connect the content you approve, and the agent answers. Back-office systems get connected only if and when you want them to, after the service is already working. A change you decide on is live immediately - within seconds, even mid-day, with no maintenance window and nothing taken offline.

Why it is immediate

  • The capabilities already exist in the product; they are not built for you from scratch
  • Putting it on your site is one line of code
  • You start with no access to internal systems and no integration project
  • An update goes live immediately - within seconds, even mid-day, with no maintenance window
  • If you want to stop, you remove one line

Your independence

Every decision about design, wording, branding and content is yours - and you make it yourself, without waiting for us.

  • The design and branding of the agent - colours, logo, name and avatar
  • The wording of the answers, including sensitive topics
  • Updating content and adding answers without touching your website
  • All of it live immediately, within seconds, with no support ticket and no waiting for a release

Would you rather we did it for you? One request and we do it immediately. That is your choice, not a dependency.

Who uses it

  • Municipalities and councils
  • Water and infrastructure utilities
  • Law firms
  • Companies and organisations
  • Institutions and non-profits
Where it works

How Redbo Meshiv Works

One AI service layer. Multiple channels. Full operational control.

Channels

Agent on your site (widget)
WhatsApp
Telegram

Redbo Meshiv - the AI agent

Answers grounded in your sources Every answer cites its source Points to you when it has no answer 24/7 in 6 languages An approved answer overrides old documents Pre-set replies for sensitive topics One consistent answer on every channel Answers in the asker's language

Knowledge Sources

Your website
Documents & procedures
Content you approved
Crawl & sync - scheduled or instant
Push content updates

Performance, Quality & Security

24/7 availability with continuous monitoring
A+ security grade (Mozilla Observatory)
57 policies and procedures built to ISO 27001 (no certificate)
Backups every 6 hours plus a full restore drill run every week
Customer data separated at the system level
Full infrastructure in Israel, the EU, or both - under your control

From setup to first answer

Crawl the site, build the knowledge, answer - with no months of onboarding.

Video transcript (text)
  1. The website address is entered once.
  2. The system reads the pages and documents and builds the knowledge base from them.
  3. A first question is asked on WhatsApp and gets an answer with a link to its source.

ChatGPT exists. Why does an organisation need a dedicated agent?

A general model guesses by document majority. We answer from the right document - and name it.

It is a fair question, and it has four measurable answers. The first three are why an organisation that points people at a general chatbot ends up publishing wrong information in its own name.

  1. Answer quality has been measured - and it is not enough

    The Tow Center at Columbia University tested eight AI search engines and found over 60% of answers were wrong. An international EBU/BBC study of more than 3,000 responses found 45% contained at least one significant issue, including outdated information. The serious part: the models answer almost every question - including ones they cannot answer correctly.

  2. A general chatbot cannot see your documents

    Scanned protocols, by-laws, plans, internal procedures and archives are mostly out of reach for a public chatbot, and some need Hebrew OCR. We crawl the site, run OCR, and sync the documents you approve - including content that is not open to the public.

  3. In any established organisation, information is duplicated by design

    A superseded procedure still appears in dozens of older documents that cannot or should not be deleted. The replacement appears in two. A general model decides by document majority and states the old one with full confidence. In an authority that is the former treasurer; in a company it is an outdated refund policy. With us, an answer you approved overrides every historical document: you decide what is true today - not the file count.

  4. Accountability, control and consistency

    Answers approved by a named owner, a full audit log, role-based permissions, and predefined responses for sensitive topics. You control your own channel on WhatsApp, Telegram and the website - and get one consistent answer instead of a different answer from every department.

Sources: Tow Center for Digital Journalism (2025) · EBU/BBC News Integrity in AI Assistants (2025) · Open Data Institute, CitizenQuery-UK (2026).

Want to see this on your own content?

A short demo on a real case of yours. No installation, no commitment.

A demo on your own content

Beyond answering questions

The agent answers the people who ask. The bigger value is what the organisation learns about itself.

A council that buys "a bot that answers" gets one more channel. A council that runs a real agent gets benefits take-up, institutional memory, and a mirror that shows it - for the first time - what is missing and what contradicts itself.

Benefits take-up, not just information

Many eligible residents never claim the discount or benefit they are entitled to - they do not know it exists, or the form deters them. The agent shows the conditions, the exact form and what to attach - the determination stays with the authority.

From the field

About 41% of property-tax complaints filed with the Israeli Ombudsman between 2023 and 2025 came from eligible residents who never received the discounts and exemptions due to them (State Comptroller of Israel, 2026).

The organisation’s knowledge - reachable by its own staff

The decision exists in a protocol. The procedure exists in a document. Nobody can find them. The agent turns the archive into something you can ask a question - scanned documents included.

From the field

A council CEO now retrieves decisions from old protocols in seconds; a council head searched for - and found - the council anthem. A UK government trial with 20,000 civil servants measured an average saving of 26 minutes per person per day.

A mirror: what is missing, what contradicts itself

Every unanswered question is a measurable hole in your site. Every contradiction between documents is a fault nobody knew about, because nobody ever read all the content at once. The agent surfaces both as a ranked task list.

From the field

At one customer it turned out the site had no public reception hours at all, and that recycling information contradicted itself between towns. No content review had caught either.

A demand map for management

What people actually ask, on which topics, and in which direction it is trending. That is decision and budget input - not message counts. Delivered as a periodic report, explicitly marked as a sample of those who reached out.

From the field

Research on municipal service requests shows the mix of requests is a signature of the neighbourhood - and that reporting propensity differs between populations. The report flags that bias explicitly.

Sources: State Comptroller of Israel (2026) · UK Government Digital Service (2025) · municipal 311 research. Full detail available on request.

Illustrative example

What it looks like in practice - three sides

The three cases below really happened at a regional council we work with.

The resident

When can I come to the collections department?

Collections is open to the public on Sunday, Tuesday and Thursday, 08:30 to 13:00. There is no public reception on Monday.

From: the "Collections department" page

The staff member

Where is the 2021 decision on the public park budget?

The decision was taken at a council meeting. The full clause can be read in the original minutes.

From: council minutes, 2021

Management

A gap surfaced from residents’ questions

Residents repeatedly asked about public reception hours, and the site had no answer. The gap was raised with management and the content was completed.

Type: content gap identified

The interface shown here is a styled reconstruction of those cases, not a screenshot of a live system. It carries no authority name, customer detail or real figure.

Serving people, not piloting

A system already working - and still improving

The agent runs on your website, your portal, WhatsApp and Telegram - and inside a closed organisational environment, for staff only. Council bots have existed for years - but not LLM-based ones, and that is the difference between a menu of buttons and a real answer.

Over a year

serving residents and staff - at a local authority and inside organisations

Over 300

capabilities live and verified

Among the first in Israel

an LLM-based agent serving residents

6

live languages

Every capability can be itemised on request. The catalogue was verified on 27.08.2026 and is pinned to a code revision.

What you get beyond the answers

  • Measurement and oversight of answer and service quality
  • Dashboards and reports: questions, topics, usage, trends and costs
  • Identification of unanswered questions and content gaps
  • Feedback collection and insight into what users actually need
  • Continuous, controlled improvement of answer quality
  • Independent management of content, languages, design and permissions

Someone asks. Someone answers. With a source.

Sixty-five seconds on what the agent does, and on what it does not do. Silent.

Video transcript (text)
  1. A resident asks how to sign up for an activity.
  2. The same question multiplies across the screen, and the channels it arrives on light up beneath it: phone centre, email, form, WhatsApp, social. The caption reads: one question, five places, nobody counting.
  3. The screen empties completely for three seconds.
  4. The conversation panel opens. The question is asked, the agent thinks, the answer is written - and only then does a source chip arrive, naming the page it came from.
  5. The same exchange cycles through six languages: Hebrew, Arabic, Russian, English, Spanish and French. The source chip stays, translated.
  6. A line on a light ground: and for the first time, you see what they actually asked.
  7. A dark ground, and three lines arrive one after another: it does not invent. It does not guess. It does not answer without a source.
  8. An end card with the logo and the address redbo.ai.

Want to see this on your own content?

A short demo on a real case of yours. No installation, no commitment.

A demo on your own content

Not just technology - adoption

A system nobody uses is a cost, not an asset

The real gap in enterprise AI is not the model, it is the people. We work on the change itself: who uses what, when, and what happens to the process they have today. Over twenty years of putting tools and processes INTO organisations, not only building them.

Map before you install

Who asks what today, where the process stalls, and what actually changes for the employee. Skip this and you get an excellent agent nobody opens.

Train the staff, not only the managers

Training for the people who use it daily, in their own language, on their own scenarios. At kick-off, and refreshed quarterly.

Change at the pace the organisation absorbs

Start with one channel and one content area, measure, then widen. Rolling everything out at once is the fastest route to resistance.

Measurement that shows it works

Volume, topics, unanswered questions and satisfaction - so you can show leadership what changed instead of relying on a feeling.

The weekly report that lands in your inbox

A silent video with on-screen Hebrew titles. What was asked, what was answered, what is missing - without opening any system.

Video transcript (text)
  1. The report reaches your mailbox every Sunday morning - no login anywhere.
  2. How many residents asked, and how many got an answer: questions, answer rate and response time per window.
  3. What residents asked most and did not find - the leading unanswered questions.
  4. And exactly which content you are missing, and who already looked for it - with a box to write the answer in.
  5. The figures shown are a sample council, not a real customer.

The guidance is included in the licence. It is not a separate service and not an extra charge. And the agent answers what repeats - it replaces no one.

Focused team. Production-grade product.

Redbo combines a focused team with a deep operational product built for real service - to residents, customers and staff.

More than a chat window

  • Multi-channel service
  • Admin control
  • Reports and insights
  • Content operations

Built for daily operations

  • Chat history
  • File sync
  • Web crawl
  • Audit log

Safe and controllable

  • Answers from approved sources
  • Role-based permissions
  • Sensitive-case responses

Ready to grow

  • Website, WhatsApp, Telegram
  • Hebrew-first, multilingual support
  • Customer-controlled content

The management portal - a short tour

What your team sees: questions asked, content gaps, and a source behind every answer.

Video transcript (text)
  1. The dashboard: how many conversations came in today, this week and this month.
  2. Conversation history: every question asked, the answer given, and the response time.
  3. Approved answers: exact wording for the questions that keep coming back.
  4. By language: the same answer, in every language you have switched on.

Before you ask

The objections we hear - and the answers

Six questions that come up in almost every procurement conversation. The short answer here, the full one behind the link.

How long will implementation take us?

Putting it on your site is one line of code. The real time goes into approving your content - not into development, and not into connecting systems.

How long implementation takes

What if the agent invents an answer?

It answers only from sources you approved, and shows the source on every answer. For common questions you approve the wording in advance, and where there is no verified information it says so.

How we make sure answers are correct

What happens to residents’ data?

The models are not trained on your content. Data is separated between customers at the system level, and sensitive numbers are masked before storage. Infrastructure in Israel, in the EU, or both - your choice.

Security and privacy

Will it meet accessibility requirements?

The site, the portal and the chat are built to WCAG 2.1 Level AA. Our accessibility statement is public and sets out exactly what has been checked.

Accessibility statement

Who is responsible if a resident gets a wrong answer?

The content and the wording stay under your control, and on sensitive topics you decide in advance what is said. How responsibility is divided is written into the terms, not left to interpretation.

Terms of use

And what is the commercial model?

A fixed monthly price based on message volume and active channels. Our standard offer includes a 30-day pilot and cancellation on 30 days’ notice, with no penalty. The binding text is the signed agreement.

What is included in the price

Our business services

Security - independently tested

the platform · client.alephaleph.ai ImmuniWeb SSL Security Test - client.alephaleph.ai - grade A ImmuniWeb Website Security Test - client.alephaleph.ai - grade A

Measured 2026-08-17 · Each one opens the live report, so you can verify it independently

Why we are here

Vision

To raise the quality of public service - in Israel and worldwide.

Mission

To give every organisation - authority, company or firm - the highest quality of service, at a cost that lets them actually run it. Technology that does not stay in the slide deck.

Our principles

Service for everyone, in their own language

Nobody should need Hebrew to get an answer from the organisation in front of them. Six languages, in wording you approved.

Answers you can rely on

The agent answers from approved sources, and every answer can be traced back to where it came from.

No implementation project

One tag on your site. No server, no integration, no six-month specification phase.

We measure, we do not promise

Availability, security and quality are measured by external tools you can re-run to check us.

Talk to Us

Schedule a live product demo, ask a general question, or set up a business consultation. Pick what fits and we'll come back to you shortly.

I'm interested in:
Talk to us on WhatsApp