New
AI outbound agent
- Places calls and runs the dialogue by a scenario
- Lead qualification, surveys, reminders and reactivation
- Call recordings flow into the shared analysis loop
Full-cycle conversational AI
The full-cycle conversational AI platform
A voice AI agent places and answers calls while the platform listens, understands and learns from every conversation. Five pillars in one loop: an AI outbound agent, an AI secretary on the line, call analysis, advanced analytics and an AI copilot.
Free to start — no card required.
Call transcript
- Agent 0:02
Meridian Clinic, good afternoon. How can I help?
- Client 0:06
Hi, I'd like to book an appointment with a GP this week.
- Agent 0:11
Of course. There's a slot Thursday at 3:30 pm — does that work?
- Client 0:17
Yes, Thursday works. Please book it for me.
- Agent 0:21
Booked. I'll send a confirmation by SMS. Have a great day!
100% of conversations covered
Every call is analyzed automatically — the whole stream, not the 2–3% a team can usually review by hand.
Dozens of recognition languages
Multilingual speech engines and an interface in 30 languages.
Data isolation and on-prem
Full account isolation, a self-hosted worker and BYO keys — audio never leaves your perimeter.
The AI flywheel
Five pillars form a closed loop that reinforces itself
Not a feature list but a self-reinforcing loop: every conversation makes the system smarter for the next one.
- Speaks The AI outbound agent and AI secretary run the calls
- Listens Every conversation is labeled automatically
- Understands Metrics, topics, QA and data extraction
- Knows A knowledge base is built from real conversations
- Improves The copilot and automations tune the system, and knowledge feeds back into training the agents
learns from its own conversations
Five pillars
Everything you need to speak, understand and improve conversations
Each pillar is a distinct platform capability. Together they form the flywheel.
New
AI secretary on the line
- Answers the inbound call, qualifies and transfers to an agent
- Stays on the line and prompts the agent in real time
- Takes notes and checks the script and compliance
Call analysis
- Transcription, diarization and 49 system metrics
- Topics, sentiment, QA and PII detection
- Custom metrics and AI-powered data extraction
Advanced analytics
- Dashboards from 10 chart types and slices
- Semantic search by meaning and per-speaker metrics
- An AI series builder: describe in words what to count
AI copilot
- Analytical questions and exact counts over your recordings
- Configures filters, tags and automation rules in plain language
- Reversible actions with Undo and strict data isolation
The starting tier of AI transformation
Start with analysis — arrive at AI
AI in calls starts with data, not a bot. First understand your conversations and build knowledge from them, then train the AI on it.
- 01
Assess readiness
Upload your call archive and get a reality check: topics, languages, audio quality and the share of scenarios where AI actually fits.
AI readiness - 02
Build a knowledge base
The platform extracts FAQs, objections, product facts and scripts from real dialogues and files them into living reference docs.
Knowledge base from calls - 03
Train AI on your conversations
From that knowledge and labeled calls it tunes the bot and secretary: scenarios, objection handling, tone and boundaries.
Train AI on your calls
Where it's used
The platform tuned to your industry
Pick a field — we'll show the pain, what the platform delivers and concrete scenarios.
Pain
Leads go cold, and leadership only ever hears 2–3% of calls.
What the platform delivers
The AI outbound agent qualifies and reactivates leads, the AI secretary prompts during live calls, and analytics shows conversion by agent.
Scenarios
- Outbound qualification and reactivation of dormant customers
- Coaching agents on data from real calls
Pain
The queue grows, quality drifts, and reasons for contact stay invisible.
What the platform delivers
The secretary answers and routes calls, analysis breaks contacts down by topic and CSAT, and the copilot fires alerts on negativity.
Scenarios
- 24/7 call intake and NPS/CSAT surveys after contact
- Automatic alerts on unhappy customers
Pain
No-shows and missed calls hit the schedule and revenue.
What the platform delivers
The bot confirms and reschedules appointments, the secretary answers calls around the clock, and analytics surfaces cancellation reasons.
Scenarios
- Confirming and rescheduling appointments by voice
- 24/7 call intake without growing the front desk
Pain
Calling debtors is expensive, and compliance is easy to breach.
What the platform delivers
The bot sends gentle payment reminders, analysis captures payment promises and tone, and the copilot sets escalation rules.
Scenarios
- Gentle voice payment reminders at scale
- Script and compliance control on every call
Pain
Plenty of leads, little time — some requests slip through.
What the platform delivers
The bot qualifies buyers, the secretary takes calls and books viewings, and analytics reveals lead sources.
Scenarios
- Qualifying buyers with outbound calls
- Taking requests and booking viewings without losses
Pain
Abandoned carts and order-status questions overload the line.
What the platform delivers
The secretary takes orders and questions, analysis exposes return reasons, and the copilot launches triggers.
Scenarios
- Taking orders and answering status questions 24/7
- Analyzing return reasons across the whole contact stream
AI secretary on the line
- 0:00AI answers
- 0:32Collects data
- 1:10Transfers to agent
AI stays on the line
Prompts for the agent
- Client already called yesterday — topic: return
- Don't forget to confirm the delivery address
Result: Recording Transcript CRM note
AI Copilot
Show calls where the client asked for a callback, and create a tag rule
Done. Found the records by meaning and prepared a rule:
Filter applied: 37 records Rule created (disabled)
Analytics
Contact topics, this week
Contacts Negative share ↓
over 4 weeks
What you get
Control instead of a sample
- of conversations covered
- 100% of conversations covered
- not hours of manual listening
- minutes not hours of manual listening
- inbound and outbound
- 24/7 inbound and outbound
- interface languages
- 30+ interface languages
How it works
From recordings to a trained AI — in four steps
You can start with no integrations: just upload recordings. Voice agents plug in once the base is ready.
- 1
Connect sources or numbers
Upload an archive, connect telephony, CRM or email — or give the bot a number to take and place calls.
- 2
AI speaks and listens
The AI outbound agent and AI secretary run the calls, and every conversation is transcribed and labeled automatically.
- 3
The platform understands and organizes
Metrics, topics, sentiment and extracted data — across the whole stream of conversations, not a sample.
- 4
You improve service and train the AI
The copilot and automations configure the system, and the knowledge you gather feeds the training of your voice agents.
Pricing
Simple plans, pay as you go
Start for free — no card required. Every plan comes with an included balance that operations draw from at cost, with no markup.
Recommended
Pro
For regular use: a larger included balance every month.
$10.00 / mo
$10.00 balance
Choose ProOperations draw from the balance at cost: transcription depends on the engine, while ingestion, diarization and storage use fixed rates. Live prices are always available in the app.
All plans and per-operation pricesPer-operation cost
Charged from your balance as you use it — with no markup.
| Operation | Price |
|---|---|
| Speech recognition | varies by engine |
| Audio ingestion | $0.003 per minute |
| Speaker diarization | $0.02 per minute |
| Storage | $0.003 per GiB / day |
| Extraction-rule generation | $0.15 per request |
| Extraction evaluation | $0.01 per sample |
At cost. The exact transcription rate depends on the engine and language you choose.
FAQ
Frequently asked questions
A quick take on what the platform is and how to get started.
What is CallScribe?
CallScribe is a full-cycle conversational AI platform. Voice AI places and answers calls, while the platform automatically transcribes, labels and analyzes every conversation: diarization, metrics, topics, analytics, a knowledge base and a copilot. It all closes into a flywheel — the system learns from your own conversations.
How is this different from “just transcription”?
Transcription is only the first step. On top of the text the platform computes 49 system metrics per recording (conversation dynamics, prosody, sentiment, compliance) plus your own AI metrics, builds dashboards with 10 chart types, searches by meaning and runs automations. Voice AI agents even generate the calls themselves. It's a system that understands and acts on your conversations — not just a transcript.
How does the voice AI agent work, and what about handing off to a human?
The AI agent answers inbound calls by a script, holds the conversation, collects the fields you need with validation, and transfers to a live agent by your rules. After the transfer it stays on the line in a conference: it keeps recording and prompts the operator in real time — summary, next action, warnings, collected fields. You can interrupt the bot (barge-in). Realtime telephony is new.
Can I start with analysis only, without any bots?
Yes — and it's the recommended way to begin. Upload your call archive and get a reality check: top topics, durations, languages, audio quality, risky spots. That's an AI-readiness assessment on your own data: you see which scenarios to automate first and where the ROI is. Voice agents come later, once the foundation is ready.
Which languages are supported?
The product interface is available in 30 languages. Speech recognition runs on system multilingual engines (Yandex, Google chirp_2, Amazon, Azure, Whisper large-v3) covering dozens of languages; you can also bring your own key for Deepgram, AssemblyAI, Speechmatics and others. The AI agent's voice and language are set via BCP-47.
How secure is it? Is there an on-prem option?
Accounts are fully isolated (the 2026-06 and 2026-07 audits found no cross-tenant leaks): only the owner and those they grant access to can see recordings. On-prem options are available: your own worker for transcription and diarization (audio never leaves your perimeter), an on-prem Asterisk agent behind NAT (outbound HMAC connection), and BYO keys for LLM/STT.
How much does it cost, and is there a free start?
Yes: the Free plan is $0 with a $3 included balance — no card required. Then Pro at $10/mo ($10 balance) and Business at $50/mo ($60 balance, for teams). Operations draw from the balance at cost: ingestion $0.003/min, diarization $0.02/min, storage $0.003/GiB/day, transcription depends on the engine. No markup.
Are there telephony and CRM integrations?
Yes. Telephony: Asterisk/FreePBX/VitalPBX/Issabel (on-prem agent), MANGO OFFICE, Telfin, Sipuni, Zadarma/Novofon, UIS, MTS Exolve, a universal webhook and inbound email. CRM: Bitrix24, amoCRM/Kommo, RetailCRM, Megaplan. Plus databases, S3-compatible storage, a REST API and an MCP server so you can work with the platform from your own code and from AI agents.
One platform
One platform, two editions
Conversational AI for voice and text in a single loop: one login, shared analytics, one knowledge base. Pick a channel — or run both.
CallScribe
Voice
Calls: AI outreach, an AI secretary on the line and full conversation analysis.
Textura
Text
Written conversations: chats, emails and tickets — data extraction, analytics and a copilot.
Open TexturaReady to start?
Turn every conversation into data, knowledge and action
Start by analyzing your conversations — no risk, no bots required. The platform turns your archive into data and a knowledge base, and voice agents plug in when you're ready.
Free plan, no card required.