hotwirednews Get GrokBot

Week 26

22–28 Jun 20262026-W26 · 23 filings · 22 companies · 8 net-new

The Use layer led the week with 9 of 23 filings, from Jobber, Infinitus, Samsara, Tata 1mg and 5 more.

Hearspeech-to-text4

TwinMindPlatformlaunch2026-06-26net-new

TwinMind brings meeting capture and dictation to Mac and Apple Watch

TwinMind's bot-free capture from system audio competes with meeting bots on the call itself, and the hold-Fn dictation puts it against Willow (2026-07-07-willow-frontier-pro-mini) launched eleven days later. Windows still sat on a waitlist on the 8 July download page; it arrived by 17 September (2026-09-17-twinmind-for-desktop-windows).

Also Use

Details3 sources

TwinMind users can now record meetings and calls from a Mac app that runs in the background, dictate hands-free, and start a recording with one tap on Apple Watch.

TwinMind for Mac and Apple Watch

Gnani.aiModel providermodel release2026-06-24

Gnani launches Prisma v2.5 for code-switched telephony speech

A telephony-focused STT model reports a 23.7% WER improvement after training on 14 million hours of audio, but Gnani does not name the baseline or evaluation set.

Details3 sources · 1 X post

Gnani.ai launched Prisma v2.5, an enterprise speech model trained on telephony audio and positioned for code-switching, accents, and dialects.

Prisma v2.5

Launching Prisma v2.5. Built on: → 30M+ customer conversations/day → 14M hours of real telephony audio Result: → 23.7% WER improvement Enterprise speech isn't clean. Prisma v2.5 was built for code-switching, accents, dialects & telephony. Built for production, not benchmarks. https://t.co/btf4p9PTo1

https://x.com/GnaniAi/status/2069759967134158855
AssemblyAIModel providermodel release2026-06-23

AssemblyAI ships Universal-3.5 Pro Realtime with agent_context

First AssemblyAI realtime SKU that conditions on the agent's own question — 8.9% relative WER drop on vendor agent audio, six weeks before the pre-recorded Universal-3.5 Pro drop.

Details1 source

AssemblyAI released Universal-3.5 Pro Realtime, a streaming STT model that takes the agent's question via agent_context, keeps rolling conversation memory, and covers 18 languages with mid-sentence code-switching at the same $0.45/hr base rate.

Universal-3.5 Pro Realtime

Reson8Model providerAPI2026-06-22

Reson8 adds Patterns to recover order codes and other identifiers

An 81% relative drop in order-number errors is a concrete identifier-recovery claim for contact-centre audio, but it is Reson8's own comparison and does not speak to overall transcript WER.

Details1 source

Reson8 added Patterns, an API feature that biases transcription toward structured identifiers such as order codes, invoice numbers, and licence plates.

Patterns

Speaktext-to-speech3

VoicesPlatformlaunch2026-06-23net-new

Voices launches Voices for Customer Experience branded AI voices

Treats enterprise CC voice quality as a cast-and-govern supply chain for platforms, not another generic synthetic voice pack — aimed at vendor platforms that resell to brands.

Also Use

Details2 sources

Voices launched Voices for Customer Experience, a casting-to-licence offering that supplies customer-service automation platforms with professionally recorded, brand-specific AI voices plus consent and licensing governance for live support use.

Voices for Customer Experience

SanasPlatformlaunch2026-06-23

Sanas expands language translation to 30+ languages and adds Speech Intelligence at CCW 2026

Sanas pairs a 30-language live translation stack with on-call compliance and fraud signals and a public developer path; the UnitedHealth $2M and 70% clarity figures are customer-claimed in the release.

Also HearOrchestrate

Details2 sources

At Customer Contact Week 2026 Sanas raised Language Translation past 30 languages, introduced real-time Speech Intelligence for sentiment, fraud, compliance, and PII redaction, and opened self-service plus a developer SDK.

Language Translation and Speech Intelligence

Fish AudioModel providerpricing2026-06-23

Fish Audio opens S2.1 Pro TTS as a free Fair Use API

Puts a current-generation multilingual TTS model on an uncapped free string for builders — Fair Use and no-SLA caveats apply, so treat it as eval/prototype capacity, not a production meter.

Details1 source

Fish Audio made its S2.1 Pro text-to-speech model available free via the existing Fish API under model string s2.1-pro-free, covering 83 languages with no hard character cap under Fair Use, without SLA or latency guarantees.

S2.1 Pro free API

Showavatars, video1

LiveKitPlatformlaunch2026-06-24

LiveKit Agents 1.6.4 adds Protoface avatar plugin

New avatar vendor plugin.

Details1 source

livekit-agents 1.6.4 adds a Protoface avatar plugin.

livekit-agents 1.6.4

ThinkLLMs, speech-to-speech2

KotobaModel providerfunding2026-06-24

Kotoba raises $10M seed led by Kindred to expand East Asia voice AI

Dated 2026 funding primary for a new East Asia speech lab on the hunt roster; same week family as the May API alpha.

Also HearSpeak

Details1 source

Kotoba announced an additional $10 million seed round led by Kindred Ventures with Salesforce Ventures and Sony Innovation Fund, bringing total funding to $23 million.

Koto platform

GradiumModel providermodel release2026-06-24

Gradium launches stt-translate and s2s-translate for realtime speech translation

Collapses the usual STT→MT→TTS cascade into two Gradium models with voice choice — vendor BLEU/MetricX lead vs Gemini/GPT realtime translate, on a five-language set only.

Also HearSpeak

Details1 source

Gradium released stt-translate, which returns translated text from speech in one step, and s2s-translate, which streams speech-to-speech translation over a single WebSocket with a selectable or cloned target voice across English, French, German, Spanish, and Portuguese.

Gradium Translate (stt-translate and s2s-translate)

Orchestrateframeworks, evals4

TalkdeskPlatformlaunch2026-06-23

Talkdesk launches Agent Builder for pre-deployment AI agent validation

Makes pre-production simulation and instruction debugging a first-class Talkdesk CXA surface, not an after-the-fact QA step once the agent is already on the phone.

Also Use

Details2 sources

Talkdesk announced Agent Builder, a natural-language tool that creates, tests, diagnoses, and validates customer-service AI agents from existing SOPs and policies before any live customer contact, then hands monitoring to the CXA Operations Center.

Agent Builder

ParloaPlatformlaunch2026-06-23

Parloa launches Navigator, a design copilot for building and fixing AI agents inside AMP

Navigator is in private preview from 29 June, so general availability and pricing are still undated. The survey figures are Parloa's own, and the same press release launched Lens (2026-06-23-parloa-lens) as the monitoring half. Talkdesk validated AI agents before deployment the same day (2026-06-23-talkdesk-agent-builder).

Details3 sources

Navigator starts builds from procedure uploads and industry templates, traces failures to root cause with line-level fix recommendations, and answers questions about agent analytics in plain language.

Parloa Navigator

ParloaPlatformlaunch2026-06-23

Parloa launches Lens, observability and automated evaluations for every agent conversation

Lens is an agent-quality product aimed at compliance teams, and its proof point is one unnamed travel platform. It is in private preview, and the Q3 composite score is a plan, not a ship. Zoom (2026-06-22-zoom-agent-architect) and Talkdesk (2026-06-23-talkdesk-agent-builder) launched agent-quality tooling in the same days.

Also Use

Details3 sources

Lens combines performance monitoring with up to 50 proprietary automated evaluations, including hallucination rate, sentiment trajectory, empathy score, tone alignment, and PII leak detection.

Parloa Lens

ZoomPlatformlaunch2026-06-22net-new

Zoom Virtual Agent adds Agent Architect and Agent Performance Suite

Puts generative agent build and post-launch measurement inside Zoom's own CCaaS stack on the same day CCW opened — voice and digital agents share one context layer rather than a bolted-on bot SKU.

Also UseConnect

Details2 sources

Zoom shipped Agent Architect, which turns a plain-language prompt into a production-ready voice or digital agent, plus Agent Performance Suite for simulation, quality management, and outcome tracking across Zoom CX.

Zoom Virtual Agent Agent Architect and Agent Performance Suite

Useagents in market9

JobberPlatformlaunch2026-06-26net-new

Jobber Voice becomes available to every team member in the field

Field technicians on Jobber's Plus plan can now speak job notes, quotes, and timer changes hands-free, a separate voice surface from Jobber's August 2025 AI Receptionist; some Jobber help pages still label Voice as beta.

Also Orchestrate

Details2 sources

Jobber made Jobber Voice available to the whole team on its Plus plan, turning spoken field updates into job notes, quotes, directions, client updates, and timers.

Jobber Voice

InfinitusPlatformlaunch2026-06-25

Infinitus launches Clinical Escalations for AI-agent patient triage

Continuous severity scoring adds a safety layer to patient-facing agents, with the Schmidt Thompson CPI framework and all-customer availability stated on 25 June but no clinical outcome data published.

Also Orchestrate

Details3 sources · 1 X post

Infinitus launched Clinical Escalations, a risk-detection system that continuously assesses patient interactions and can reclassify emergent clinical situations for AI-agent triage.

Clinical Escalations

Today we're announcing Clinical Escalations, our risk detection system that ensures AI agents triage patients correctly, regardless of an interaction’s stated intent. Read more in Fierce Healthcare: https://t.co/ZchFHEVzsz https://t.co/WdTOO14mEJ

https://x.com/InfinitusAI/status/2070159579997241628
SamsaraPlatformlaunch2026-06-24net-new

Samsara adds two-way AI voice to its dash cams and launches Agent Studio

The dash cam becomes the voice endpoint for drivers who cannot hold a phone, so agent calls reach the cab without a handset or dialler; the 30-minute saving per call comes from one unnamed customer with no baseline given.

Also Orchestrate

Details2 sources · 2 X posts

Samsara AI can now talk with drivers through the in-cab dash cam, for example when they enter a geofence, and fleets can build their own driver-facing agents in a new no-code Agent Studio.

Samsara dash cam two-way voice; Agent Studio

A driver's whole day, reimagined. Personalized start-of-shift briefings, CarPlay navigation in the cab, two-way voice with both AI and managers, and recognition for safe driving. All built around the driver. #SamsaraBeyond

https://x.com/Samsara/status/2069823508041895987

Meet Agent Studio. Build agents that handle routine work like delivering start-of-shift briefings or surfacing weekly coaching recaps for you. Use a pre-built template, or build one completely from scratch — no coding required. See more: https://www.samsara.com/company/news/press-releases/samsara-launches-new-agentic-capabilities-to-automate-tedious-operational-tasks #SamsaraBeyond

https://x.com/Samsara/status/2069829739955486830
Tata 1mgEnterpriselaunch2026-06-23net-new

Tata 1mg runs abandoned-cart sales calls end to end on Arrowhead voice agents

Arrowhead's second named deployment in the corpus is an outbound sales job, where Kissht (2026-09-08-kissht-arrowhead-voice-agents, W37) was inbound support; the 15% lift is vendor-reported against Tata 1mg's own team, and the API count differs between the June post and the case study.

Details3 sources

The Indian online pharmacy has Arrowhead agents call customers who leave medicines or lab tests in the cart, resolve the objection, and place the order during the call.

Abandoned-cart recovery calls · vendor Arrowhead

KrispPlatformlaunch2026-06-23

Krisp launches Voice Security and Speech Analytics for contact centres

Extends Krisp from live audio enhancement into CC governance — Speech Analytics is GA on day one; Voice Security stays early access, so deepfake coverage is not yet a full GA claim.

Also Hear

Details3 sources

Krisp added Voice Security (real-time deepfake, social-engineering, and agent-voice verification alerts) in early access and Speech Analytics (100% post-call scoring, summaries, compliance flags, dispositions) generally available on its Call Center AI platform.

Voice Security and Speech Analytics

InvocaPlatformlaunch2026-06-23

Invoca introduces Nico, a voice and SMS agent that books appointments from marketing leads

University Hospitals' 25 first-night appointments are the only outcome figure, and with voice still in beta on 23 June the release does not separate phone from SMS bookings.

Also Connect

Details2 sources

Invoca's Nico answers inbound calls, texts, and web or ad-form leads for considered-purchase sectors such as healthcare and home services, qualifying intent, booking or rescheduling appointments through Salesforce, HubSpot, and NexHealth, and routing high-intent callers to live agents.

Nico

Five9Platformlaunch2026-06-23net-new

Five9 relaunches Voice AI Agents on an agentic architecture

Moves Five9's voice self-service off scripted IVR bots onto a native agentic harness inside its carrier-grade switch — CCW-day product, not a partner-only demo.

Also OrchestrateConnect

Details1 source

At CCW, Five9 launched a new Voice AI Agents release on a purpose-built agentic stack with multi-agent orchestration, secure tool calling, AI Agent Studio, and warm handoffs to human agents on its Intelligent CX Platform.

Voice AI Agents

AmbiencePlatformlaunch2026-06-23net-new

Ambience ships Ambient Flowsheet Documentation for inpatient nursing

Ambient voice into inpatient nursing flowsheets — not just physician note drafts — with explicit nurse-in-the-loop confirmation.

Also Hear

Details3 sources

Ambience Healthcare expanded its nursing suite with Ambient Flowsheet Documentation, which turns narrated patient assessments into structured EHR flowsheet values that nurses confirm before submit, alongside Nursing Summary.

Ambient Flowsheet Documentation

Amazon ConnectPlatformlaunch2026-06-22

Amazon Connect Customer opens Agentic CX designer and Live Sync in preview

Ties Amazon's CCaaS self-service builder to a voice↔UI sync path — preview on 22 Jun, so treat go-live timing as later than the What's New date.

Also OrchestrateConnect

Details2 sources

Amazon Connect Customer put Agentic CX designer (NLX) in preview: a no-code canvas for voice and digital self-service that mixes agentic and deterministic AI, with Live Sync driving a caller's web or mobile UI in real time during the conversation.

Agentic CX designer (NLX) and Live Sync