Voice AI agents and call analytics: a platform comparison
Ten solutions that work with Uzbek speech: which ones talk to customers themselves, which only review recordings and which provide a speech API. Based on public data checked on 12 October 2026.
Launch it yourself — from 2 hours. A demo on your own channels.
What a voice AI agent is and how it differs from call analytics
A voice AI agent is software that runs a phone conversation in place of an operator: it answers an inbound call or dials a number, listens to the reply, understands what the caller means and answers in a synthesized voice. Unlike robocalls and IVR, the agent does not play a recorded menu; it holds a dialogue and hands complex cases to a person. Call analytics (speech analytics) is a neighbouring category: the service never talks to the customer, it transcribes and scores calls that operators have already had. In Uzbekistan both kinds of product face the same test: Uzbek speech over a phone line, switching from Uzbek to Russian mid-sentence, dialects and two scripts. The market splits into three groups: platforms with a voice agent (Acclaim, Aisha AI, CyberNet AI, Zukko.AI), call analytics services (Leadlar AI, ovozAI, U-BSS, ZorCall) and speech APIs for developers (Muxlisa AI, Yandex SpeechKit). Below they are compared on public data as of 12 October 2026.
Data checked on · spotted an error? Write to us
Platforms with a voice agent
Talk to customers themselves: they answer inbound calls, and some of them make outbound calls too. Many also review calls.
In this listAcclaim · Aisha AI · CyberNet AI · Zukko.AI
Call analytics services
Never talk to the customer: they transcribe and score recorded operator calls for quality control and coaching.
In this listLeadlar AI · ovozAI · U-BSS · ZorCall
Speech APIs
Speech recognition and synthesis for your own build. Your team assembles the dialogue, telephony and CRM.
In this listMuxlisa AI · Yandex SpeechKit
How to choose: 7 criteria
Put these questions to any vendor on the list — before the demo and during it.
1.Uzbek speech on a live call
Ask how the system handles switching from Uzbek to Russian mid-sentence, dialects, names and amounts on a phone line. Test only on your own recordings: vendor accuracy figures are self-reported and measured on different datasets.
2.Latin and Cyrillic script
The same word arrives in two scripts and with different apostrophes. Check whether transcripts are normalized to one script; otherwise search across calls and reports drift apart.
3.Where the data is stored
Uzbekistan’s personal data law (ZRU-547) requires personal data of Uzbek citizens to be stored inside the country. Ask where the server is, whether on-prem or private cloud is available and which subcontractors — cloud language models, speech recognition — receive the audio.
4.CRM and telephony integration
SIP trunk or PBX (Sipuni, OnlinePBX, Asterisk, 3CX), which CRM — built-in, amoCRM, Bitrix24 — and whether the call outcome lands in the customer record automatically.
5.Analysis only, or conversation too
A call analytics service will not call your customer; a voice agent will. If you need outbound calling (reminders, collections, reactivation), check that it is part of the product, not “on the roadmap”.
6.Price and billing model
Per minute, per call, subscription or a project with a one-off rollout. Count the full cost: minutes, call reviews, rollout and support.
7.Time to launch
From hours for self-serve services to weeks and months for an enterprise project with a security review.
Comparison table: ten solutions
Ordered by solution type, alphabetically within each group. This is not a ranking.
| Solution | Type | Uzbek speech | Inbound / outbound | Call analytics | Data and deployment | Price | Launch |
|---|---|---|---|---|---|---|---|
| Platforms with a voice agent | |||||||
| Acclaim | Enterprise voice agent platform (formerly Aiphoria) | Yes — per media reports (kun.uz, 2025) | Inbound and outbound (sales, collections, support) | Not stated | On-prem or private cloud; SaaS also mentioned | On request | Demo in days, first use cases in weeks (company’s claim) |
| Aisha AI | Voice agent and speech APIs (STT, TTS) | Yes; the API lists Uzbek, Russian, English | Inbound; outbound sales use cases stated | Yes — Call Metrics call scoring | Data hosted in Uzbekistan; on-premise on the Enterprise plan | Public API price list; voice agent on request | “In days” (company’s claim) |
| CyberNet AI | Voice agents for call centers | Uzbek and Russian (per uzdaily.uz) | Inbound and outbound, including mass calling (company’s claim) | Yes — analytics of recorded inbound calls | Cloud or the customer’s servers | Not published | Not stated |
| Zukko.AI | AI agent platform with a built-in CRM | Yes: mixed Uzbek-Russian speech, dialects, transcripts normalized to one script | Inbound and outbound | Yes — 100% of calls reviewed | Data of customers in Uzbekistan is stored in Uzbekistan; on-prem or private cloud for CIS banks | Inbound and call reviews from the Business plan; outbound calling on Pro, from 1,800 UZS/min | SMB from 2 hours, enterprise up to 30 days |
| Call analytics services | |||||||
| Leadlar AI | Sales call and chat analytics (Mirai Step) | Uzbek and Russian, mixed speech | No voice agent — reviews recordings | Yes | Not publicly stated | Public price list | 1–3 days (company’s claim) |
| ovozAI | Call analytics for call centers | Yes; 40+ languages, auto-detected | Reviews recordings | Yes | Not stated; per the website, recognition and analysis run on Vertex AI Gemini (Google) | Public price list | About an hour (company’s claim) |
| U-BSS | Speech analytics, virtual assistant, voice biometrics | Yes: Latin and Cyrillic | Assistant (text and voice); outbound not stated | Yes — speech analytics and quality control | On-prem or managed service | On request | Not stated |
| ZorCall | Call quality control for bank call centers | Russian and Uzbek | No voice agent | Yes | Not stated | On request | Integration in “1 day” (company’s claim) |
| Speech APIs | |||||||
| Muxlisa AI | National speech platform (UZINFOCOM): STT, TTS, API | Yes — the main language | Inbound automation; outbound planned | Yes — the Speech Analytics product | Cloud or on-premise (per the anhor.uz review) | Public price list | Not stated |
| Yandex SpeechKit | Cloud speech recognition and synthesis API | Recognition: Uzbek (Latin); synthesis: three Uzbek voices | Not an agent — API only | No — API only | Yandex Cloud | Public price list (in RUB) | Self-serve, by your developers |
Data from company websites and media reports as of 12 October 2026; sources are listed in the cards below. Vendor claims about accuracy and efficiency are left out of the table: they are measured on different datasets. This page was prepared by the Zukko.AI team.
Data checked on · spotted an error? Write to us
Which solution suits whom
Facts only from the vendor’s website or the media, with a source link in every card. No “better” or “worse” verdicts.
Platforms with a voice agent
Acclaim
Enterprise voice agent platform (formerly Aiphoria)
A voice AI agent platform for banks, fintech and other regulated industries: collections, sales and support, with use cases built in the Agent Studio builder. Deployed on-prem or in a private cloud — data, model weights and logs stay with the customer. Uzbek and Uzbek-Russian switching are confirmed by media reports; the company website says “any language”. Headquarters: Miami.
Suits: A large financial company with its own IT team and a project budget that needs deployment inside its own perimeter.
Sourcesacclaim.ai · kun.uz · anhor.uz
Aisha AI
Voice agent and speech APIs (STT, TTS)
A platform for voice agents and speech APIs: speech recognition and synthesis, Call Metrics call scoring, Meeting AI and a separate product for banks. Data is hosted in Uzbekistan, on-premise is available on the Enterprise plan. The home page states it works with your existing number and PBX and connects to Bitrix24, amoCRM or a custom CRM.
Suits: Teams with developers who need Uzbek STT/TTS through an API with a public per-minute price, and companies that want a voice agent on top of their own CRM.
Sourcesaisha.group · aisha.group/en/voice-agent · aisha.group/en/pricing
CyberNet AI
Voice agents for call centers
Voice AI agents for bank, retail and telecom call centers: scripted agents, a generative agent without a rigid script and analytics of recorded inbound calls. Runs as a cloud service or on the customer’s servers. We could not find a website describing the product — the facts come from a uzdaily.uz report of 6 May 2026.
Suits: A large call center prepared to request terms directly: there is no public pricing or product description.
Sourcesuzdaily.uz
Zukko.AI
AI agent platform with a built-in CRM
An AI agent platform with its own CRM: the voice AI agent answers inbound calls and calls customers itself, call analytics transcribes and scores 100% of conversations and writes the outcome into the customer record, and the same agents reply in Instagram, Telegram and WhatsApp. Connects to a SIP trunk, a virtual PBX and your existing numbers, to Sipuni or any PBX and to amoCRM; without IP telephony, the Zukko Recorder Android app collects the recordings.
Suits: A business that wants calls, chats and CRM in one system and a self-serve launch. If you only need a speech recognition or synthesis API for your own build, Aisha AI, Muxlisa AI or Yandex SpeechKit are a closer fit.
SourcesVoice AI agent · Call analytics · Speech to text · Pricing
Call analytics services
Leadlar AI
Sales call and chat analytics (Mirai Step)
AI analysis of sales team calls and chats: lead “temperature”, manager score, next step. Works on top of amoCRM and Bitrix24 and Sipuni, OnlinePBX and Mediatel telephony; reports arrive in Telegram.
Suits: A sales team already working in amoCRM or Bitrix24 that wants to review its managers’ calls without a voice agent.
Sourcesleadlar.uz
ovozAI
Call analytics for call centers
A call analytics platform for call centers: transcription, analysis and scoring of conversations. Connects to OnlinePBX and Sipuni, recordings can be uploaded manually, alerts arrive in Telegram. Per the website, recognition and analysis run on Vertex AI Gemini (Google).
Suits: A mid-size or large call center that needs call scoring for supervisors and the QA team with per-minute billing.
Sourcesovozai.uz
U-BSS
Speech analytics, virtual assistant, voice biometrics
Speech analytics and quality control, an omnichannel virtual assistant, Uzbek speech recognition in Latin and Cyrillic, voice biometrics and anti-fraud. Integrates with core banking, remote banking and processing; deployed on-prem or as a managed service.
Suits: Banks, fintech, telecom and the public sector that want speech analytics and voice biometrics from a single vendor.
Sourcesubss.uz
ZorCall
Call quality control for bank call centers
AI quality control of calls for bank call centers with 50+ operators: transcription, emotions, script compliance, dashboards. Connects via API or SIP trunk; the website names Asterisk and 3CX. An early-stage project: part of the features on the website is listed on the roadmap.
Suits: A bank call center that needs quality control and is ready to develop the solution together with an early-stage team.
Sourceszorcall.uz
Speech APIs
Muxlisa AI
National speech platform (UZINFOCOM): STT, TTS, API
UZINFOCOM’s national speech platform: Uzbek speech recognition and synthesis, diarization, an API and a ready Speech Analytics product for contact centers. Outbound calling is planned, according to UZINFOCOM as of April 2026.
Suits: The public sector and developers who need an Uzbek speech recognizer through an API, and contact centers that need speech analytics.
Sourcesmuxlisa.uz · uzinfocom.uz · anhor.uz
Yandex SpeechKit
Cloud speech recognition and synthesis API
A cloud API for speech recognition and synthesis: recognizes Uzbek in Latin script and offers three female Uzbek voices for synthesis. It is infrastructure, not a ready agent: your team assembles the dialogue, telephony and CRM.
Suits: Developers and integrators building their own solution and paying in rubles.
SourcesSTT models · TTS voices · pricing
Test a voice agent on your own scenario
Leave your number and the Zukko.AI agent will call you itself. It is the simplest way to hear Uzbek speech on a live call and compare it with the other solutions on this list.
Questions about voice AI agents and call analytics
Which voice AI agent should I choose in Uzbekistan?
Start with the job. If you need a conversation with the customer — inbound and outbound calls — look at platforms with a voice agent: Acclaim, Aisha AI, CyberNet AI, Zukko.AI. If your operators stay and you need their calls reviewed, look at call analytics: Leadlar AI, ovozAI, U-BSS, ZorCall. If you are building your own solution, look at the Muxlisa AI and Yandex SpeechKit speech APIs. The deciding test is your own calls: vendor accuracy figures are self-reported and measured on different data.
Are there voice AI agents that speak Uzbek?
Yes. Aisha AI and Zukko.AI state Uzbek support; per media reports, so do CyberNet AI (uzdaily.uz) and Acclaim (kun.uz, 2025); U-BSS has a virtual assistant with a voice channel. Uzbek speech synthesis is also available through APIs: Aisha AI, Muxlisa AI, Yandex SpeechKit. Test how each system handles Uzbek-Russian switching mid-sentence and dialects on your own recordings.
How is a voice AI agent different from call analytics?
A voice agent talks to the customer itself: it answers a call or dials a number and holds a dialogue. Call analytics never talks to the customer — it transcribes and scores conversations operators have already had. Aisha AI, CyberNet AI and Zukko.AI do both; Leadlar AI, ovozAI and ZorCall do analysis only.
Can call data be stored in Uzbekistan?
Yes. Aisha AI states its data is hosted in Uzbekistan; at Zukko.AI, data of customers in Uzbekistan is stored in Uzbekistan. Deployment on the customer’s infrastructure (on-prem) or in a private cloud is offered by Acclaim, Aisha AI (Enterprise plan), CyberNet AI, U-BSS, Zukko.AI and, per the anhor.uz review, Muxlisa AI. Uzbekistan’s ZRU-547 law requires personal data of Uzbek citizens to be stored inside the country, so ask each vendor where the server is and which subcontractors receive the audio.
How much do a voice AI agent and call analytics cost?
Billing models differ: per minute, subscription or a project with a one-off rollout. Public price lists are published by Aisha AI (for its recognition and synthesis API), ovozAI, Leadlar AI, Muxlisa AI and Yandex SpeechKit; Acclaim, CyberNet AI, U-BSS and ZorCall quote on request. At Zukko.AI, inbound calls and call reviews are available from the Business plan, while outbound calling connects to the Pro plan and is billed per minute, from 1,800 UZS a minute. Count the full cost: minutes, reviews, rollout and support.
How long does it take to launch?
Self-serve call analytics services go live in hours or days: ovozAI states about an hour, Leadlar AI 1–3 days. For a voice agent: Zukko.AI from 2 hours for SMB and up to 30 days for enterprise with integrations, Aisha AI “in days”, Acclaim a demo in days and first use cases in weeks. In larger projects the timeline also includes a security review.
Compare on your own calls
A demo on your scenario: the Zukko.AI agent calls the number you leave, and we send you a recording of a real call on your data. Self-serve launch from 2 hours, enterprise up to 30 days.
Uzbek, Russian and English. Local payments and channels out of the box.