- The AI Voice Newsletter
- Posts
- Apple Introduces AI Call Screening in iOS 26
Apple Introduces AI Call Screening in iOS 26
🔊 Soundcheck
Apple Introduces AI Call Screening in iOS 26
Sharper voice AI accuracy for contact centers
Character voices dataset unlocks expressive AI agents
AI voice ups Donatos’ pizza orders fast
Read time: 4 minutes
🔥 Hot Mic
Big moves, deep dives, and standout stories.
Apple just rolled out Call Screening in iOS 26, a feature designed to reduce interruptions from unknown numbers by letting iPhone users see who’s calling, and why, before deciding to answer.
Building on Live Voicemail, Call Screening automatically picks up calls from unsaved numbers and prompts the caller to state their name and reason for calling. A live transcript of the response then appears on the recipient’s screen, allowing them to accept or ignore the call without disruption. Screened calls from unknown numbers can also be filtered into a separate list, keeping the main call log clutter-free. For users stuck on hold, a companion feature called Hold Assist notifies them when a live agent finally joins the line.
Key Points:
Screen Unknown Callers – iPhone silently answers and asks for caller name + reason
Real-time Transcript – See details on screen before choosing to answer or decline
Filtered Call List – Screened calls from unsaved numbers can be kept out of the main log
Hold Assist – Alerts users when a live agent picks up after long hold times
User Control – Provides context without interruptions, reducing robocalls and unwanted outreach
Takeaway: Call Screening in iOS 26 gives iPhone users more control over their calls, reducing spam and interruptions while making it easier to decide which unknown numbers are worth answering.
Microsoft’s new constrained speech recognition enhances contact center voice AI accuracy using grammar-based rules.
Microsoft has introduced a feature called Constrained Speech Recognition in its Dynamics 365 Contact Center. It applies structured grammar rules to speech recognition, reducing errors especially with alphanumeric inputs like codes or tracking numbers. Traditional systems often misinterpret critical details, creating friction in customer interactions.
This update narrows the recognition focus to expected phrases and formats, improving accuracy in noisy or multi-accent environments. The innovation is especially valuable in sectors like healthcare, finance, or enterprise IT, where precision is essential and transcription mistakes can lead to serious consequences.
Key Points:
Introduced Constrained Speech Recognition in Dynamics 365 Contact Center
Uses SRGS-style grammars to define expected voice inputs
Improves accuracy on alphanumeric or structured voice data
Helps reduce manual corrections and agent workload
Takeaway: By enforcing structured voice input recognition through grammars, Microsoft’s Constrained Speech Recognition narrows ambiguity, boosting accuracy and reliability in enterprise contact centers—cutting down transcription errors that frustrate agents and customers.
Voices released a vast, ethically sourced dataset with 450+ character voices, featuring tones, emotions, and metadata.
Voices has unveiled a one-of-a-kind, ethically sourced audio dataset tailored for voice AI training. It includes recordings of over 450 distinct character types, each brought to life by professional voice contributors. The collection spans a rich emotional and expressive range—19 tones, 35 emotions, and 21 paralinguistic cues—making it the first multi-character dataset of its kind. Every clip includes detailed metadata and full transcriptions, ready for AI model ingestion. Recorded using Voices’ in-house virtual Recording Studio with auto-directed sessions, the dataset delivers high-quality, authentic performances. Voices emphasizes transparent terms, fair compensation, and contributor control in its ethical sourcing approach.
Key Points:
Over 450 distinct character voice types
Includes 19 tones, 35 emotions, 21 paralinguistic cues
Rich metadata and full transcriptions included
Recorded in virtual studio with auto-directed sessions
Takeaway: This launch marks a milestone in voice AI training—delivering a diverse, expressive, and ethically sourced audio dataset that empowers developers to build personalized, character-rich conversational agents with authenticity and respect for voice talent.
Donatos deployed Revmo AI voice ordering across 174 locations, achieving major gains in conversion, accuracy and labor efficiency.
Donatos rolled out Revmo’s AI voice–ordering system across 174 restaurants by June, fully integrated into its POS. The initiative came after Donatos selected Revmo following the planned phase‑out of DoorDash AI.
By the end of August, the system had handled over 301,000 calls, lifted order conversion from 58% to 71%, and delivered near‑perfect order accuracy at 99.9%. The adoption freed thousands of labor hours, redirecting staff efforts back to in‑store hospitality.
Key Points:
71% order conversion rate vs previous 58% benchmark
301,388 calls handled by AI through August
99.9% order accuracy – about one mistake per 1,000 orders
4,840.8 labor hours reallocated in August alone
Takeaway: This rollout shows that voice AI can move the needle—boosting conversions, freeing labor, and enhancing guest experience when embedded into operations.
🎙️ Mic Drop
What else is making noise in voice AI.
OpsTechPro and VENZA introduce Guest Voice AI to streamline hotel operations with automated call handling and guest services. (hotelmanagement.net)
Independent metrics confirm impressive conversion rate and accuracy following Donatos' full-scale deployment of Revmo voice ordering tech. (prnewswire.com)
Capacity expands in healthcare with acquisition of KLaunch, strengthening its conversational AI offering for customer engagement. (hitconsultant.net)
Albertsons rolls out Google Cloud’s conversational AI for commerce, aiming to enhance personalized digital grocery shopping. (finance.yahoo.com)
Tour24 prioritizes conversational AI in self-guided property touring with new sales leadership to accelerate adoption. (finance.yahoo.com)
MEGA secures $2M pre-seed funding to build specialized AI voice agents for automated order-to-cash management workflows. (startupsmagazine.co.uk)
Level announces smart lock integration with Josh.ai, enabling seamless home access via voice, text, and touch interfaces. (cepro.com)
Investor activity in voice AI startups accelerates, with global VC inflows spotlighting rapid growth across the sector. (crunchbase.com)
Researchers develop new TTS methods that improve speech clarity and user engagement, addressing synthetic voice quality issues. (sfu.ca)
Guide to running open-source Dia-1.6B TTS locally, offering a self-hosted alternative to proprietary text-to-speech platforms. (vocal.media)
Microsoft adds ‘scripted mode’ to Copilot’s TTS, improving output clarity for enterprise and accessibility use cases. (pcworld.com)
Waveshare launches ESP32-S3-Touch-LCD-3.49 board, expanding DIY HMI applications with built-in AI voice control support. (cnx-software.com)
Cars24 deploys ElevenLabs and Google Cloud voice AI to improve buying and selling experiences in automotive marketplaces. (techcircle.in)
Competitive landscape grows as multiple startups raise funds and enter the restaurant voice AI space amid rising adoption. (restaurantbusinessonline.com)