How do AI Avatars improve citizen services?
1. Answers outside office hours
Most citizens look for help in the evening and at weekends, when offices are closed. The Avatar answers then, and tells them what to bring when they do visit.
What to measure: conversations outside opening hours, and the share that end with a clear next step.
2. Every resident's language
The Avatar answers in the citizen's language. Managed Agents speak 30+ languages; the Speech-to-Video API reaches 100+ through a connected text-to-speech provider.
What to measure: conversations by language, and requests for an interpreter at the counter.
3. Fewer wasted appointments
When people arrive with the wrong documents, the appointment is wasted. An Avatar that explains requirements first means more appointments succeed the first time.
What to measure: appointments rebooked because of missing documents.
4. Longer, more complete conversations
A face holds attention through a longer explanation. The national citizen-services app on our platform saw 3x longer sessions than its previous chatbot (one customer, not a controlled test).
What to measure: session length and the share of questions resolved without a phone call.
5. Staff time goes to complex cases
When the Avatar handles routine questions, caseworkers spend more time on the cases that need them.
What to measure: call-centre volume for the top 20 routine questions, before and after.
What do public bodies need from an AI Avatar vendor?
Procurement teams ask these questions first. Here is how Beyond Presence answers them:
- Where is the data? Persistent data is stored in the EU (Google Cloud in Belgium and the Netherlands). On-prem and private-cloud deployments keep conversation data in your own environment.
- Who is the vendor? Beyond Presence GmbH, a German company based in Munich.
- Security evidence? SOC 2 Type II audited, GDPR compliant and EU AI Act compliant, with documents in our Trust Center.
- Training on our data? No. We don't train on customer data or conversations.
- Data processing terms? A Data Processing Agreement is signed with every customer.
- Scale? 1,000+ concurrent conversations per deployment, for deadline days and public campaigns.
How do you keep it within the EU AI Act?
This is general information, not legal advice.
- Disclose that it's AI. Article 50 requires people to be told they're talking to AI unless it's obvious. Say it in the first line and on screen.
- Inform, don't decide. The AI Act lists AI used by public authorities to evaluate eligibility for public assistance benefits and services as high-risk. An Avatar that explains rules and requirements is not making that decision; keep it that way.
- Always offer a person. Give every conversation a route to a caseworker, a phone number or an appointment.
- Accessibility. Show answers as text alongside speech, and keep a text-only channel for people who prefer it.
How does it work?
Each turn runs through speech recognition, an LLM grounded in the authority's published information, a voice and the real-time Avatar. The Avatar layer takes about 100 ms; with Managed Agents, the full reply arrives in about 1.0–1.2 s.
Ground the agent only in approved content: service pages, forms, opening hours and FAQs. When it doesn't know, it says so and points to the right office.
How do you roll it out?
- Months 1–2, one service. Start with a high-volume, low-risk service such as address registration or ID renewal. Write the knowledge base from the call centre's top questions.
- Months 3–4, measure. Track resolution, languages, out-of-hours use and handovers to staff.
- Months 5–6, expand. Add services, languages and channels, including kiosks in service centres.
When is an AI Avatar the wrong choice?
- Decisions about people's rights or money. Eligibility, fines and permits need a caseworker.
- Sensitive personal situations. Domestic violence, bereavement and crisis support need trained people.
- Out-of-date information. If your service pages aren't current, the Avatar will repeat old rules. Fix the content first.
Frequently asked questions
How can governments use AI Avatars for citizen services?
To answer routine questions face to face, in the citizen's language, at any hour: which documents to bring, how to book an appointment, what a service costs and what happens next. Decisions about benefits and permits stay with caseworkers.
Is an AI Avatar GDPR compliant for public sector use?
It can be, with the right setup. Beyond Presence is GDPR compliant, stores persistent data in the EU, signs a Data Processing Agreement and supports on-prem deployment so conversations stay in your environment.
Can the Avatar run on government infrastructure?
Yes. Beyond Presence licenses its Genesis 2.0 model for on-prem and private-cloud deployment, with engineering support and an SLA.
Does the EU AI Act allow AI Avatars in public services?
Yes, with transparency: people must be told they're talking to AI. AI that evaluates eligibility for public benefits is high-risk under the Act, so keep the Avatar to information and guidance.
Which languages can citizens use?
30+ with Managed Agents, and 100+ with the Speech-to-Video API and a connected text-to-speech provider.
Are public bodies already using AI Avatars?
Yes. A national citizen-services app on Beyond Presence answers people face to face, in their own language, day or night, and sees 3x longer sessions than the chatbot it replaced.
Sources: Beyond Presence homepage (customer results and Security section, checked Nov 11, 2026) · Trust Center · Data Processing Agreement · Regulation (EU) 2024/1689 (AI Act)
Related reading: What is Beyond Presence? · AI Avatars in healthcare · AI kiosks and in-store concierges
Talk to us about a sovereign deployment: book a demo.
