Why residents’ questions are repeated
Every municipality, polyclinic, library or museum receives the same questions every day. When do you open? What documents are needed? When is waste collected in my street? Can I come with a group? The answers are usually already on the website, but hidden deep in a menu, in a PDF document or in an appendix to a decision. People cannot find them and call or write an email.
That is how a queue appears that nobody wants: the resident waits for an answer, while the specialist repeats the same thing again and again. A AI chatbot is useful here not because it is fashionable, but because it can answer repeated questions immediately — using what has already been written on the organisation’s website and in documents.
How a chatbot trained on an organisation’s content works
The key difference from general chat tools is that the bot is trained only on the organisation’s information. It indexes website texts and uploaded PDF documents: procedures, regulations, decisions. Every answer is provided with a link to a specific page, so the resident can check the source themselves.
When a page is updated on the website or a new document is uploaded, the bot re-indexes the content and the answers update automatically — nothing needs to be reinstalled. If the information is not in the organisation’s content, the bot is set not to make it up and to say that it does not have such information.
For municipalities: fewer repeated calls
Municipal websites are large, and residents’ questions are very specific. The bot can not only find an answer in the texts, but also query structured municipal data directly — for example, bus timetables or waste collection schedules. The resident gets an exact answer, not a link to a PDF.
During the conversation, the resident can also report a problem: an infringement, a fault or a request. The system creates a report number, encrypts the contact details and routes the report to the responsible department by category. The call centre is left with complex cases, not repeated questions. If the volume of reports is high, it is also worth looking at a separate city reporting system.
For public bodies and health centres
For organisations with a complex service structure, the bot becomes a quick search instead of getting lost in menus. In health centres and polyclinics, the most common questions are standard — opening hours, services, prices, what to bring. The answers are short, but everyone needs them, often outside working hours.
Here it is important to set a clear boundary in advance: the bot answers general questions from the organisation’s information, but does not provide health advice and does not ask for personal health data — in such cases it directs people to reception or a specialist.
For libraries and museums: answers in the evening and at weekends too
Library and museum visitors often plan their visit in the evening or at the weekend, when nobody answers the phone. The questions are similar: opening hours on public holidays, ticket prices, whether a tour can be booked for a group, whether the building is accessible to people with disabilities, where to find the events programme.
If this information is already on the website, the bot answers it immediately and provides a link. For a small team, that means fewer interruptions, and for the visitor — an answer when they need it.
What the organisation sees in the admin dashboard
The chatbot is useful not only for residents. In the dashboard, you can see what people ask, what they could not find an answer to, and where information is missing. This gives a concrete list of which website pages to expand or rewrite more clearly — the website improves based on real questions, not guesses.
Responsible deployment: transparency, GDPR and accessibility
For a public-sector organisation, a chatbot must be not only convenient, but also transparent and secure:
- People must know they are talking to AI. Article 50 of the EU AI Act requires AI systems that interact with people to be designed so that a person knows they are talking to AI, unless this is obvious from the circumstances. In practice, it is best to say this clearly at the very start of the conversation, not in small print.
- Personal data — only as much as necessary. Personal data is not needed to answer questions, but people sometimes enter it themselves, so the organisation must decide how long conversations are stored and who can see them in the dashboard. Contact details from reports are stored encrypted. All this is described in the organisation’s privacy policy.
- Answers from the organisation’s sources. The bot is set to answer from the organisation’s information and provide a link with every answer that can be checked. AI can make mistakes, so for important questions it is worth checking the source or contacting a specialist; when there is no answer, the bot says so and directs the person to a human.
- Accessible to everyone. The chat window must work with a keyboard and screen readers, just like the whole website, which must meet WCAG 2.1 AA requirements.
Where to start
The best place to start is with the most common questions the organisation already knows: review what people ask by phone and email, and check whether the answers are on the website. If they are not — first update the website, because the bot only answers with what is written there.
Tobalt chatbot for the public sector is deployed in 3–5 days and trained on your website content and documents. Want to see how it would answer your residents’ questions? Write to us — we will show you a demo with your website content.