- monday.com Harmony (19 Oct) runs outbound and inbound sales calls with task-specific models that answer in about 200 ms, then hands warm leads to a human with a briefing.
- Mistral Voxtral 3 (20 Oct) bundles speech-to-text, text-to-speech and a LiveKit-based agent SDK. It can run on your own servers or EU-hosted APIs.
- Scaled Cognition Voice (20 Oct) is a text-to-speech model for service calls: $9 per million characters hosted, $1 self-hosted plus hardware, and it reads dates and amounts the way people say them.
monday.com Harmony: a voice agent built for the sales call
monday.com launched Harmony, a voice agent platform for sales teams. Instead of one general model, each model does one job: qualifying a lead, capturing data or handling an objection. The company says they respond in about 200 milliseconds and reports a contact rate of up to 70%. The voice technology comes from OneAI and the platform runs on Twilio.
It ships with unlimited phone numbers with branded caller ID, predictive and power dialing, inbound routing, consent and compliance controls and native CRM write-back. When a lead is ready, Harmony transfers the call to a live rep with a full briefing. One customer, Bright Car Finance, says Harmony qualifies about half of its conversations and hands 80% of those to advisers live. Its director Richard Stubbs: "Harmony's AI agent now follows up with every approved applicant right away." Harmony is only offered to mid-market and enterprise customers in the US, UK, Australia and Singapore.
Your angleThe interesting part is not the voice, it is the follow-up rule: call every new lead right away, qualify, hand over the warm ones. Smaller businesses cannot buy Harmony, but they have the same problem. Mapping their lead flow, writing the qualifying questions and setting the hand-over moment is a service you can deliver on any voice platform.
Mistral Voxtral 3: listen, think and talk in one stream
Mistral AI launched Voxtral 3, a family of models for real-time voice agents: Voxtral Small Transcribe 3 for speech-to-text, Voxtral Mini TTS 3 for speech, and a voice agent SDK built on LiveKit. Listening, reasoning, using tools and speaking run in one streaming pipeline, so the agent can book an appointment or check an invoice while it is still on the call.
Transcription covers more than 50 languages and starts at $0.0035 per minute. Text-to-speech covers more than 35 languages and 50 dialects and adapts its emotion. Mistral says the whole system can run on "your own infrastructure or through EU-hosted APIs."
Your angleFor European clients that worry about where call recordings go (clinics, lawyers, accountants), an EU-hosted or self-hosted voice agent removes the first objection. The work you sell is the part the SDK does not do: connecting the agenda or invoice system, writing what the agent may and may not promise, and testing it in the client's own language and dialect.
Source: The Decoder
Scaled Cognition Voice: a service-call voice at $9 per million characters
Scaled Cognition, an AI lab that raised a $100 million Series A led by Khosla Ventures, made its Voice text-to-speech model generally available. It supports more than 25 languages and voice cloning, and plugs into existing agents, contact-centre platforms and older phone menus (IVR). It does not need the company's own language model.
Its "Smart Pronunciation" works out meaning before it speaks: "Pay 1/2 by 1/2" becomes "pay half by January second." The company reports 70 milliseconds from text to first sound. Hosted, it costs $9 per million characters; self-hosted, $1 per million plus your own hardware. The company says competing hosted voices charge $35 to $100 per million characters, but did not name them. Akshay Kannan, vice president of product management: "Self-hosting gets economical faster than people expect." Its blind listening test favoured Voice, but no sample sizes were published.
Your angleA voice that reads "€1,250 due 3/11" correctly sounds small, but it is exactly where cheap voice bots embarrass a business. If you build voice agents for clients, a test script full of their real dates, prices, names and order numbers is a deliverable in itself, and a reason to choose a voice on quality, not only on price.
Source: SiliconANGLE
The voice is becoming a commodity. What the agent is allowed to say is the part a client pays for.
Overview: three launches, one service
Based on the announcements linked above. Speed, contact-rate and price comparisons are the companies' own figures, so test before you promise anything to a client.
| Harmony | Voxtral 3 | Voice | |
|---|---|---|---|
| Announced | 19 Oct 2026 | 20 Oct 2026 | 20 Oct 2026 |
| What it is | Complete sales-call platform | Speech models + agent SDK | Text-to-speech only |
| Where it runs | monday.com on Twilio | Own servers or EU-hosted API | Hosted API or self-hosted |
| Price shown | Not published | Transcription from $0.0035/min | $9 / 1M chars ($1 self-hosted) |
| Main limit | Mid-market and enterprise, 4 countries | You build the agent yourself | Needs an agent and phone stack around it |
| Service to sell | Lead follow-up rules and hand-over | Private EU voice agent setup | Voice testing with real client data |
Before a voice agent answers a client's phone: 5 checks
- Does the caller know it is an AI? Say it in the first sentence. Check what the law in the client's country asks for calls and recordings.
- What may it promise? Write down which prices, dates and discounts it may say, and what it must hand to a human.
- When does a human take over? Agree on the trigger (angry caller, payment, complaint) and test the transfer, not just the talk.
- Does it read the client's data right? Run a test script with their real dates, amounts, names and addresses.
- What does one call cost, all in? Voice, language model, phone line and your hours. Price per month, not per character. Our pricing calculator helps, and the security checklist covers access.
Around 200 ms answers and 70 ms to first sound make calls feel natural. That is no longer the hard part.
EU-hosted and self-hosted options open the door to clients who said no to call bots.
Qualifying questions, hand-over moments and what not to promise: that is the work you sell.
Get the next briefing in your inbox
One short email when there is AI agent news that matters for people who sell to businesses. Sources included, no hype.
FAQ
What is monday.com Harmony?
Harmony is a voice AI agent platform for sales teams that monday.com launched on 19 October 2026. Task-specific models qualify leads, capture data and handle objections, and the company says they respond in about 200 milliseconds from audio in to audio out. It includes branded caller ID, dialing, consent controls and CRM write-back. It is available to mid-market and enterprise customers in the US, UK, Australia and Singapore.
Can I host a voice agent in the EU or on my own servers?
Mistral AI launched Voxtral 3 on 20 October 2026: speech-to-text, text-to-speech and a voice agent SDK built on LiveKit. Mistral says it can run on your own infrastructure or through EU-hosted APIs. Scaled Cognition's Voice text-to-speech model, also launched on 20 October, can be self-hosted as well.
What does an AI voice cost per call?
Prices are quoted per unit, not per call. Scaled Cognition charges $9 per million characters for its hosted Voice API and $1 per million characters self-hosted, excluding infrastructure. Mistral's Voxtral 3 transcription starts at $0.0035 per minute. Add the cost of the language model, the phone line and your own setup work before you quote a client.
Recorded training · real agent screens
Watch a team of AI agents do real client work
In the free training you will:
- See agents handle leads, follow-up, content research and spreadsheets, step by step.
- Learn how to use it for your own business or whitelabel it for clients.
- Leave with a next-step plan you can start the same week. No coding required.
Free training. No income guarantee: results depend on your own work and clients.
What you learn to deliverReady to start this week?
Skip the wait: get the full course
€497 one-time · instant login after paying
- 6 modules: build agents for leads, follow-up, research and reporting
- Price calculator + proposal and outreach templates
- Progress tracking, lifetime access, no coding needed
Not sure yet? Watch the short video first · compare all 3 options
Secure checkout via Stripe (+2% service fee). No income guarantee: results depend on your own work and clients.
All sources
- monday.com / Business Wire via StockTitan: Voice AI platform Harmony launches (19 Oct 2026)
- The Decoder: Mistral AI launches Voxtral 3, a full voice agent stack (20 Oct 2026)
- SiliconANGLE: Scaled Cognition targets customer service with low-cost voice model (20 Oct 2026)
- TokenPost: Scaled Cognition launches Voice API at $9 per million characters