Opening hours and location
Opening times, public holidays, address, directions or which documents to bring.
AI voice agent for answering phone calls
Digitup Studio is a technical studio in Madrid that designs and builds AI voice agents: a virtual receptionist that answers the phone, resolves frequent questions, manages appointments, takes details and hands the call to a person when needed.
Definition
What is an AI voice agent?
An AI voice agent is software that answers phone calls in natural language: it listens, understands the request, looks up what it needs and replies with a synthetic voice. It is also called a virtual receptionist, AI phone assistant or smart answering service. Unlike a voicemail box, it resolves what it can and leaves the rest ready for your team.
It has three parts: telephony, which receives the call; an AI model, which transcribes, decides and replies by speaking (a real-time voice model, or a chain of speech recognition, language model and speech synthesis); and the integrations with your calendar, your CRM or your knowledge base. If your customers prefer to write, there is the WhatsApp AI chatbot; if it must act across several systems, it is an AI agent with a voice.
Calls
Which calls can an AI virtual receptionist handle?
The ones that repeat every day and don't need a professional's judgment.
Frequent questions
Services, terms and published prices that you validate.
Appointments
Book, change or cancel according to the calendar's real availability.
Taking details
Name, phone, reason or case number, saved in your CRM.
Handover
To the right extension, with a summary of the conversation.
Out of hours and peak times
When nobody is there or the lines are busy.
Integrations
How does it connect to your calendar, CRM and phone system?
In three layers: the line, with a programmable telephony provider such as Twilio or Telnyx; your tools, through APIs; and your team, with transfers and alerts. Anton Zverev, CTO of Digitup Studio, has spent more than 10 years building products and systems connected to services such as OpenAI, Gemini, Twilio or Telnyx.
With a PBX phone system, the agent transfers to an extension; how this is done depends on the model and the carrier. Appointments are booked in Google Workspace or Microsoft 365 calendars, or in your management software if it has an API. It answers from a knowledge base you control, with RAG if there are many documents, as in our automation and AI integration projects.
What is recorded after each call?
A summary, the reason for the call, the data collected and, if you decide so, the transcript. Your team reviews it in a dashboard or in the CRM and sees which questions the agent can't answer yet.
Handover
What happens when the agent doesn't know the answer?
It hands the call to a person, takes a message or schedules a call back. It must not improvise answers that are not in its sources.
We agree with you when it hands over: if the caller asks for it, if it is urgent, if professional judgment is needed or if it doesn't understand after two attempts. Whoever picks up receives a summary and, if nobody can, a task is left in your CRM. The AI prepares; the person decides.
Transparency
Do you have to say an AI is answering, and what happens to the data?
Yes. Article 50 of the EU Artificial Intelligence Act (Regulation (EU) 2024/1689) requires that people know they are interacting with an AI, unless it is obvious, and that they are told at the latest at the first interaction. It applies from 2 August 2026, so the agent says it in its first sentence.
The call also processes personal data. Article 13 of the GDPR requires telling people, when the data is collected, who the controller is, what it is used for and on what legal basis, among other points; Article 11 of Spain's Organic Law 3/2018 allows doing this in layers. If you record or transcribe, the opening announcement must say so.
At a clinic, health data is a special category (Article 9 of the GDPR): the agent asks only for what the appointment needs. This is general information, not legal advice; the announcement and the privacy policy are validated by your adviser or your data protection officer.
Sectors
In which businesses does a voice agent make most sense?
Where the phone is the main channel and calls look alike. We don't yet have a public voice-agent case in these sectors.
Reception overloaded?
Clinics
A smart answering service for clinics: appointments, changes and opening hours. Anything clinical goes to a person, always.
See clinicsIncidents at all hours?
Property managers
Collects the incident and the resident's contact; anything urgent goes to the person on call.
See property managersCalls from listings going unanswered?
Real estate
Answers about published properties, collects what the enquirer is looking for and proposes a viewing.
See real estateFirst calls that interrupt?
Law firms
Takes details and the area of the matter, detects urgent deadlines and books the consultation. It never gives advice.
See law firmsComparison
Answering machine, phone menu or AI voice agent?
The answering machine takes messages, the phone menu routes calls and the voice agent resolves the repetitive ones.
| Aspect | Traditional answering machine | Phone menu (IVR) | AI voice agent |
|---|---|---|---|
| How it answers | Recorded message and voicemail. | Options by keypad or by voice. | Conversation in natural language. |
| Frequent questions | It doesn't answer them. | Recorded messages per option. | Answers with approved information. |
| Appointments and details | A message to listen to later. | Passes the call to a person. | Checks the calendar and records the details in your CRM. |
| Handover to a person | Your team calls back. | Through a menu option. | By rules, with a summary of the conversation. |
| Cost of use | Low and fixed. | Depends on the phone-system provider. | Telephony and AI model, by minutes. |
| Main risk | Missed calls and unreturned messages. | Long menus that people abandon. | Errors if its sources and rules are not limited. |
General comparison; each phone system depends on its provider.
Cost
How much does an AI voice agent cost?
Digitup Studio projects start at €1,000, plus two usage costs that depend on call volume: telephony and the AI model.
Development includes script, rules, knowledge base, integrations and testing; integrating a calendar, CRM or phone system raises the price. Telephony is paid to the provider at its rate, and the model is paid by usage.
You can pay as a fixed-price project, over 6 or 12 months, or with Build & Run, a monthly fee with agreed usage limits. More in ways of working.
Process
How do we start a voice agent?
With a prototype that you test with real calls before connecting it to your number. Delivery from 2 weeks.
We analyse your calls
Which calls you receive, how many are lost and which ones don't need a person. If it isn't worth it, we tell you.
We design script and rules
What it tells callers, what details it asks for, when it hands over, how it says it is an AI and what is stored.
We test with real calls
Your team calls a test number; we measure answers, times and cost per call.
We integrate calendar, CRM and phone system
Appointments, call logging and transfers, with permissions limited to what is necessary.
We launch in stages
First out of hours or unanswered calls; then we expand and run maintenance in production.
Fit
When does a voice agent fit and when doesn't it?
It fits with volume and repetition; not when every call needs professional judgment.
When it fits
- Many similar calls: opening hours, appointments, status of a procedure.
- Calls missed out of hours or at peak times.
- Stable information that someone can approve.
- A calendar or CRM with an API, or manual confirmation is acceptable.
When we are not the right option
- Emergency lines: people staff them, and in an emergency you call 112.
- Decisions that belong to a professional, such as a diagnosis or legal advice.
- Few calls: an answering machine costs less.
- An off-the-shelf tool that is already configured is enough.
- A budget below €1,000 or delivery in under 2 weeks.
If in doubt, an AI audit sets priorities before building.
Related experience
There is no public case of a telephone voice agent yet. The closest is Mallorcab, a taxi platform in Mallorca developed by Digitup, with AI for a virtual call center, translations and support.
Frequently asked questions about voice agents
Does the voice agent say it is an artificial intelligence?
Yes, in its first sentence and before asking for any data. Article 50 of the EU Artificial Intelligence Act (Regulation (EU) 2024/1689), applicable from 2 August 2026, requires people to be told that they are interacting with an AI system, unless it is obvious, at the latest at the first interaction. The same opening announcement also says whether the call is recorded.
Can I keep my current phone number?
Yes. The usual setup is to keep your number and divert to the agent all calls, only those out of hours, or only those nobody answers, as far as your carrier allows. The agent runs on a line from a programmable telephony provider, such as Twilio or Telnyx. If you prefer not to touch your line, you can start with a new number for one specific service.
What happens if the caller wants to speak to a person?
The agent transfers the call to the extension or phone you define and leaves a summary of the conversation, so nobody has to repeat their case. The handover rules are agreed before going live: an explicit request, urgency or topics that need professional judgment. If nobody is available, it takes the message with the reason and phone number and creates a task in your CRM.
Can it book appointments in my calendar?
Yes, if your calendar can check availability and book through an API, like Google Workspace or Microsoft 365 calendars, or your management software if it has that option. The agent offers free slots, confirms day and time with the caller and can change or cancel appointments. If your calendar has no API, it records the request with the necessary details and reception confirms it afterwards.
Are the calls recorded?
Only if you decide so: you can store the audio, the transcript or just a summary. In any case the call processes personal data, and Article 13 of the GDPR requires telling people, when it is collected, who the controller is and what it is used for; if you record or transcribe, the opening announcement must say so. This is general information: your adviser validates the announcement.
How much does an AI voice agent cost?
Digitup Studio projects start at €1,000. An agent that provides information and takes messages is closer to that minimum; integrating it with a calendar, CRM or phone system raises the price. Two usage costs are paid separately according to call minutes: telephony, which the provider bills, and the AI model. We estimate them with your real call volume.
How long does it take to have it running?
Delivery starts from 2 weeks for a well-scoped case; integrations with a calendar, CRM or phone system may need more phases. Before connecting it to your number, your team tests a prototype with real calls on a test number. We adjust script and rules with what we learn and launch in stages, starting with out-of-hours calls.
Which languages can it answer in?
Current voice models understand and speak several languages, so the agent can answer in more than one. We define with you which ones, normally Spanish and the languages your customers use most, and we test each language with real calls before activating it, because the quality of recognition and voice changes by language and model. In all of them it uses the approved information.
Want to try a voice agent with your own calls?
Tell us which calls you receive and which ones get lost today. We will tell you how we would approach it and also whether we think it isn't worth doing.