Heaven InteractiveБългарски

02 · Voice & Chat Agents

Voice & Chat Agents

Every call and message answered, in Bulgarian and English, day and night.

Book a call · All practice areas

Conversational agents on real telephony and on your website, Bulgarian first and English when the caller switches. They answer from your own knowledge base, take bookings straight into your calendar and hand off to a person the moment the conversation needs one. Nothing is invented: if the answer is not in the source material, the agent says so.

Who it is for

  • Clinics, hotels and service businesses that miss calls outside working hours
  • Reception and support teams answering the same questions all day
  • Delivery and logistics operations with many status and rescheduling calls
  • Companies whose customers speak Bulgarian, English, or both

Problems it solves

  • Missed calls become lost bookings

    Evening, weekend and busy-hour calls go to voicemail, and most callers do not leave one.
  • The same answers, all day

    Opening hours, prices, directions and order status take up staff time that should go to people who need it.
  • Bulgarian is an afterthought

    Generic assistants stumble over Bulgarian names, streets and places, and callers notice.
  • No record of what was said

    Phone conversations leave no searchable trace, so problems repeat.

Capabilities

  • Inbound and outbound voice on your own number or SIP trunk
  • Web and messenger chat on the same knowledge base
  • Booking, lookup and callback tools connected to your systems
  • Warm transfer to a person on your team
  • Pronunciation dictionary tuned for Bulgarian names and places
  • Every conversation transcribed and reviewable

What you receive

  • Voice agent and phone number

    Configured, tested and answering on your line.
  • Web chat widget

    Styled to your brand and working from the same knowledge.
  • Knowledge base and pronunciation dictionary

    Built from your real material, not a generic script.
  • Transcript dashboard and runbook

    Review every conversation; change knowledge safely after handover.

How the engagement runs

Setup, then usage-based running. Typical setup is confirmed at scoping.

  1. Week 1: Knowledge and call flows

    We collect your real material, map the calls you receive and agree when the agent must hand over to a person.

  2. Weeks 2–3: Build the agent

    Voice, pronunciation, knowledge base and tools for booking or lookup are configured and connected to your systems.

  3. Week 4: Back-test and go live

    The agent is tested against a fixed set of real questions. It answers your line only after it passes.

  4. Ongoing: Review and improve

    Transcripts reviewed, the test set re-run after every change. Available as an optional retainer.

Pricing and availability

  • Pricing band: EUR 3–12k setup
  • How the engagement runs: usage-based running cost · optional retainer
  • Lead time: 6–8 weeks from signing

Pricing bands and lead times are our published starting point. After the first call you receive a written scope and a quote.

Questions

Will callers know they are speaking to an AI?

Yes. The agent introduces itself as a virtual assistant. We recommend it, and EU rules on AI transparency point the same way.

What happens when the agent does not know the answer?

It says so. It then offers a transfer to a person, a callback or a note, depending on the rules you set. It does not improvise facts.

Can we keep our current phone number?

Usually, yes. The agent can answer your existing number through call forwarding or a SIP trunk, or we can set up a new number.

Voice & Chat Agents

Hear how a call goes.

Our voice agents answer on a real phone line, in Bulgarian first and in English when the caller switches. They work from your own knowledge, use your calendar and systems, and hand over to a person when the conversation needs one. Below are two sample conversations, written to show the behaviour.

  • Answers from your material

    If the answer is not in the source material, the agent says so instead of guessing.
  • Bulgarian first

    Tuned pronunciation for Bulgarian names, streets and places. English when the caller switches.
  • A person when needed

    Warm transfer to your team, with the context already captured.
  • Every call on record

    Each conversation is transcribed and reviewable in your dashboard.

Sample conversation: Clinic reception: booking a first appointment

Bulgarian · inbound call · 19:40, after reception has closed

  1. Agent: Добър вечер, свързахте се с рецепцията на клиниката. Аз съм виртуалният асистент. С какво мога да помогна? (Good evening, you have reached the clinic reception. I am the virtual assistant. How can I help?)
  2. Caller: Здравейте, искам да запиша час при кардиолог. (Hello, I would like to book an appointment with a cardiologist.)
  3. Agent: Разбира се. Първи преглед ли ще бъде, или контролен? (Of course. Will this be a first visit or a follow-up?)
  4. Caller: Първи. (A first visit.)
  5. System: Проверява графика на кардиолозите (Checks the cardiologists’ schedule)
  6. Agent: Имам свободен час в сряда в девет и половина сутринта или в петък в два следобед. Кое ви е по-удобно? (I have a slot on Wednesday at half past nine in the morning, or on Friday at two in the afternoon. Which suits you better?)
  7. Caller: Сряда сутрин. На името на Мария Петрова, телефонът е този, от който звъня. (Wednesday morning. The name is Maria Petrova, and the phone number is the one I am calling from.)
  8. System: Записва часа · изпраща SMS потвърждение (Books the slot · sends an SMS confirmation)
  9. Agent: Записах ви за сряда в девет и половина, първи кардиологичен преглед. Ще получите SMS с адреса. Ако имате направление, моля, носете го. (You are booked for Wednesday at half past nine, first cardiology visit. You will receive a text with the address. If you have a referral, please bring it.)
  10. Caller: Благодаря. Има ли паркинг? (Thank you. Is there parking?)
  11. Agent: Да, клиниката има паркинг за пациенти зад сградата. Мога ли да помогна с нещо друго? (Yes, the clinic has patient parking behind the building. Can I help with anything else?)
  12. Caller: Не, това е всичко. (No, that is all.)
  13. Agent: Благодаря за обаждането. Приятна вечер! (Thank you for calling. Have a good evening!)

Appointment booked in the clinic calendar, SMS confirmation sent, parking question answered from the clinic’s own information. Reception finds it in the morning; nobody had to call back.

Scripted example to illustrate behaviour. Not a recording of a real patient.

Sample conversation: Delivery line: rescheduling a parcel

English · inbound call · parcel delivery line, Sofia

  1. Agent: Good afternoon, you have reached the delivery line. I am the virtual assistant. How can I help?
  2. Caller: Hi. I have a parcel coming tomorrow, but I won’t be home. Can I move it?
  3. Agent: Of course. Could you read me the shipment number from your text message?
  4. Caller: It’s B-G four seven one nine zero two.
  5. System: Looks up shipment BG471902
  6. Agent: Thank you. I can see a parcel for Mladost, Sofia, due tomorrow between ten and two. I can offer Thursday from nine to twelve, or Thursday from two to six. Which works for you?
  7. Caller: Thursday afternoon. And can the driver leave it with my neighbour?
  8. System: Reschedules delivery to Thursday 14:00–18:00
  9. Agent: Done, Thursday from two to six. About your neighbour: I don’t have a confirmed rule for that on this shipment, so I won’t promise it. I can put you through to a colleague now, or add a note for the driver. Which would you prefer?
  10. Caller: Just add the note, please.
  11. System: Adds driver note · sends SMS with the new window
  12. Agent: Note added. You’ll get a text with the new delivery window in a moment. Anything else I can help with?
  13. Caller: No, that’s it. Thanks.
  14. Agent: Thank you for calling. Goodbye.

Delivery rescheduled, driver note added, SMS sent. When the question went beyond the source material, the agent said so and offered a person instead of inventing an answer.

Scripted example to illustrate behaviour. Not a recording of a real customer.

Voice & Chat Agents in detail

Related work

  • Production voice-agent framework

    Voice AI · 2024. The framework behind every voice and chat agent we build: Bulgarian and English, phone and web.
  • Client dashboards for voice and chat agents

    Custom tooling · 2025. One login per company to review every conversation its agents have had.

Start here

Bring a real problem.

Thirty minutes with an engineer who would build it. Tell us what is breaking, what you have tried and what would change if it worked.

Book a 30-minute call · info@heaven-interactive.com

  • Free, no sales pitch
  • Reply within one working day
  • Pricing band before any spec work