“The stylist is with a client.”
The phone rings. Nobody picks up. The caller hangs up and dials your competitor.
Voice AI · chatbots · custom systems
Voice AI, built and run for customer service, sales, and the work behind them.
We build and run voice AI systems — from discovery through integration to daily operation. Over 30 languages, with native-level Hungarian where you need it. If the job calls for it, we also build the embedded web assistant, the chatbot, or a custom system to go with it. All of it from one place.
Inbound and outbound callsChatbots and web assistantsVision and document AISystem integrationOperations and measurement30+ languages · EU data handling
Two to four weeks to the first live call. After that the running stays with us: we measure, review, and improve every month.
Sound familiar?
Every business runs into these.
“The office closes at four.”
Evening appointment requests pile up in the inbox. In the morning someone calls them back one by one — if there is time.
“Support gets the same ten questions every day.”
The agent burns out. Answers come slower. The customer hangs up.
“Where is my order?”
A hundred times a day, the same question — and every one of them ties up an agent who could be doing better work.
None of this is inevitable. All of it is fixable.
How we work
The real work starts after go-live
Rollout is a straight line: we learn your process, connect the system to what you already run, and take it live. After that it turns into a loop — measure, review, improve, month after month, for as long as the system runs. That loop is where the difference comes from.
Rollout
Discovery
process, data
Integration
your existing systems
Go-live
step by step
Operations — monthly, ongoing
Measure Review Improve every conversation measured
Rollout takes a few weeks. The operating loop stays with us for as long as the system runs — every month you get a performance report, an updated knowledge base, and a list of proposed fixes.
We build it and we run it
- Processing and maintaining the knowledge base
- Building, releasing, and updating the system
- Round-the-clock operation, monitoring, and fixes
- Rate limits, alerts, and cost controls
Your part — this is all of it
- Handing over source material and access
- For the website: pasting in roughly one line of code
- Brand basics: color, logo
- Feedback during the test round
What lands on your desk monthly
- A performance report on the key numbers
- An expanded knowledge base covering the new questions
- Proposed fixes for the next round
By the numbers
Measured results, not promises
1000+
Live conversations measured across our completed campaigns
91%
On-topic rate in our most recent retail campaign — audited
11 days
Continuous operation without a break, at shopping-mall peak traffic
24/7
Call coverage with concurrent calls, without adding headcount
The engine
Every system we build answers from a knowledge base — not from memory.
The most common worry about AI is that it makes things up. We do not rule that out with a rule, we rule it out by design: the system answers only from the verified material you handed over, and it shows where the answer came from. What the knowledge base does not cover, it does not guess at — it hands off.
01 · Source material
Your documents
Product data, price lists, policies — what you already have on hand.
02 · Knowledge base
Chunked and checked
It becomes a searchable, versioned knowledge base with verified content.
03 · Channel
Phone, chat, kiosk
The same engine works on the phone, in chat, and on a kiosk — only the surface changes.
04 · Answer
With its source attached
Every answer starts with a search, and you can always see where it came from.
The questions it could not answer make the knowledge base smarter every month.
Illustration — source discipline in practice
I started the course but the video will not play. What should I do?
That is usually a browser setting. Refresh the page and allow autoplay — if the error stays, switch browsers. Source: User manual · Video playback
And if that does not help?
Then this one is for customer support — I will give you their contact details. What the documentation does not answer, I do not guess at.
Listen
Here is how it actually sounds.
You do not have to imagine it. Below are real situations — an AI·Squad agent handling customers across different industries. The recordings are in Hungarian.
Sales
B2B lead qualification and demo booking
0:34 Customer service
Outage report and proactive callback
0:33 E-commerce
Order status check and rescheduling
0:41 Healthcare
Appointment booking with flexible ID matching
0:33The demos for the other industries sit on their own solution pages.
🎙️ Talk to one live →
These are recordings. On our demo site you talk to the agent yourself — and afterward you see what it understood.
Solutions
Which one is your industry?
Healthcare & clinics
doctors' offices · private clinics · dental practices · physical therapy
Thirty to fifty calls a day, and 60% of them are appointments. The receptionist cannot keep up — and in the evening nobody picks up at all.
Beauty & salons
hair salons · beauty studios · nail salons · day spas
The stylist is with a client. The phone rings. The caller will not wait — they go somewhere else.
Restaurants & hospitality
restaurants · pizzerias · cafés · catering companies
Twenty to thirty calls at once during the rush. Your server is not a switchboard — the order gets lost.
Customer service
companies of 20–200 people · service providers · utilities
Your agents field the same ten questions every day. That is not work for a person — it is work for the AI.
E-commerce
online stores · marketplace sellers · direct-to-consumer brands
“Where is my order?” “When does it arrive?” — 70% of calls are about this. The AI answers them around the clock.
Sales & lead generation
real estate · solar and insurance companies · B2B sales teams
A human rep manages 40–60 calls a day. An AI·Squad agent handles 500 — qualifying, booking, handing over.
IT helpdesk & support
SaaS companies · IT providers · MSPs
Weekend and night support is expensive and hard to staff. The AI diagnoses, logs the ticket, and escalates, 24/7.
Soft collection
financial services · utilities · telecom companies
Overdue invoices, late payers, subscribers about to cancel. The AI handles soft collection calls with empathy and within the rules.
Sales call training
sales teams · financial advisors · call centers
Your rep calls an AI customer who asks back, objects — and hangs up if it is going nowhere. After the call they get a scored review, quoting their own words back to them.
System integration
An AI system is worth only as much as it can see of your business
If the system cannot reach your data, all it does is make polite conversation. So every project starts with the same question: what does this have to connect to? Then we connect what you already run. From the phone line to your webshop orders: if a system has an interface or a data connection, we can plug into it.
Phone system, SIP
VCC.Live · Telnyx · PBX systems
CRM
HubSpot · MiniCRM and others
Calendar, scheduling
Google · Microsoft 365
E-commerce, orders
UNAS · Shoprenter · Shopify
AI system
voice · logic · decisions
ERP
orders, inventory, invoicing
Ticketing, helpdesk
Zendesk · Freshdesk · ServiceNow
Databases, knowledge stores
spreadsheets · databases · documents
Email, SMS, chat
notifications, confirmations
We do not replace what you work with — we connect to it.
Phone integration
Hungarian phone numbers, inbound and outbound calls — on every mobile network and landline.
Web widget (no phone number needed)
If you would rather your customers did not have to dial: a voice and text assistant embedded in your site, callable straight from the browser.
Custom connection
If your system is not on the list, we build a custom integration. Webhook, REST API, direct database connection — we will sort it out.
What our customers say
Feedback
AI Squad develops the AI voice agent solution for Protocall, and I recommend them to any company looking for a reliable AI partner. They are flexible, proactive, and extremely thorough: they respond quickly to every request while carefully weighing the options to arrive at the best solution. The technology they deliver is outstanding and strikingly lifelike, which makes it easy and effective to use in inbound customer service, and in simpler outbound projects as well.
Working with AI Squad showed us that voice AI is not just an enterprise topic. They understood agency logic quickly — that we speak on behalf of our clients, and that this takes more than technology, it takes understanding the context. The result is something we can confidently recommend to our own clients.
Security & compliance
Your data stays in the EU. That is the default here.
GDPR-compliant
All data is stored in the EU. Nothing is transferred to a third country.
EU AI Act, Article 50
Every one of our agents tells the caller they are speaking with an AI system. Lawfully, and on the record.
EU infrastructure end to end
We do not run on US-based servers. Data residency you can stand behind in a regulated industry.
Auditable
Call transcripts, reports, and a complete activity log in the client portal.
The team
Who is behind this
László Sabján
CEO & AI Voice Architect
Business development specialist and technology entrepreneur. He has been building enterprise systems for over a decade, focused on voice AI and agent workflows since 2024.
LinkedIn →
János Pereczes
Chief Growth Officer
Economist and technology investor. Strategic lead in the fintech transformation at MKB and MBH Bank. Founder of AI Budapest and 1337.partners.
LinkedIn →Straight talk
When we will tell you not to do this
These come out by the second meeting anyway. We would rather say them up front.
When the process itself is broken.
AI does not fix a broken process, it just runs it faster. Fix the process first. We can help with that too, but it is better to know going in.
When there is nothing to connect to.
If your data lives in people’s heads and spreadsheets, the system has nothing to look up. Sorting out the data comes first, not the voice.
When the conversation calls for human judgment.
Leave the complicated, sensitive cases to people. Hand the system the lead-in and the follow-up instead, so your colleagues’ time goes to the hard part.
FAQ
Frequently asked questions
What customers ask us most often about voice AI.
A phone assistant powered by AI. It speaks in a natural human voice, responds in real time, and can handle customer calls on its own — booking appointments, taking orders, or answering questions.
No. Today's text-to-speech produces natural human speech, with emotional shading. AI Squad's agents speak Hungarian at a native level — callers usually cannot tell them apart from a live agent.
Two to four weeks to the first live use, typically. That covers building the knowledge base, designing the conversation paths, the integrations, and testing. Simpler use cases can go live in about a week.
More than 30, and one system can cover several of them at once. Hungarian is native-level rather than a translation layer — the agent thinks and speaks in it, which is not something every provider can say.
It hands the call to a human agent, along with the full context of the conversation. We define the escalation rules up front — the AI knows exactly which topics need a person.
We work with two pricing models, depending on your expected call volume.
At lower volumes you pick a fixed monthly package — each one includes a set number of minutes, so you know exactly what you pay every month.
At higher volumes, or for more complex needs, the model is a one-time setup fee plus per-call pricing: you pay for the minutes you actually use.
The exact cost depends on your industry, your call volume, and how complex the integrations are. On a short call we will show you which model works out better in your case — and put together a quote for you.
Yes. It is GDPR-compliant, runs on EU servers, and we do not use customer data for model training. Three layers of protection: system-level rules, real-time monitoring, and encrypted storage.
CRMs (MiniCRM, Pipedrive, HubSpot, Salesforce), calendars (Google Calendar, Outlook), invoicing software (Billingo, SAP), helpdesk platforms (Freshdesk, Zendesk), and plenty of others through their APIs. We build custom integrations too.
No. We handle the whole rollout, the operation, and the ongoing tuning. You do not need technical staff — keeping the system running is our job.
A dashboard tracks call volume, resolution rate, average call length, and customer satisfaction. We also send a monthly report comparing the AI's performance with the previous period.
Let us start with a conversation.
First we look at whether there is a good solution here at all. If there is not, we will say so.