{"id":27108,"date":"2026-07-28T12:13:27","date_gmt":"2026-07-28T12:13:27","guid":{"rendered":"https:\/\/www.acefone.com\/blog\/?p=27108"},"modified":"2026-07-28T12:13:27","modified_gmt":"2026-07-28T12:13:27","slug":"ai-voice-agent-july-2026-updates","status":"publish","type":"post","link":"https:\/\/www.acefone.com\/blog\/ai-voice-agent-july-2026-updates\/","title":{"rendered":"What’s New at AceX AI Voice Agents: July 2026 Updates"},"content":{"rendered":"

Most latency numbers you see for voice agents aren’t what the customer actually feels.\u00a0Most knowledge setups either bloat every prompt or go stale.\u00a0<\/span>
\nMost TTS engines still mispronounce your own brand name.\u00a0<\/span>\u00a0<\/span><\/p>\n

Here’s\u00a0how we\u00a0changed\u00a0that.<\/span>\u00a0<\/span><\/p>\n

Latency Now Means What the Customer Feels<\/span><\/b>\u00a0<\/span><\/h2>\n

Industry reporting typically tracks AI Infra Latency, the sum of STT Latency, LLM Latency, and TTS Latency. That number misses one thing: network latency.<\/span>\u00a0<\/span><\/p>\n

Acefone now reports\u00a0<\/span>Perceived Latency-<\/span><\/i><\/b>\u00a0AI Infra Latency\u00a0+\u00a0Network Latency.\u00a0Individual voice agent calls\u00a0might\u00a0only\u00a0exhibit\u00a0slight\u00a0noise during calls.\u00a0At scale,\u00a0however,\u00a0network latency\u00a0stops being\u00a0just noise. It starts affecting real-life\u00a0call\u00a0experiences.<\/span>\u00a0<\/span><\/p>\n

Perceived Latency is now\u00a0living\u00a0in the\u00a0platform\u00a0Latency tab. It gives teams the number that\u00a0matches\u00a0what a customer hears on a live call.<\/span>\u00a0<\/span><\/p>\n

Knowledge That Doesn’t Bloat Every Prompt<\/span><\/b>\u00a0<\/span><\/h2>\n

Stuffing every reference document into a system prompt gets expensive\u00a0real\u00a0fast. Token size balloons, ingestion costs climb, and answers get less reliable.<\/span>\u00a0<\/span><\/p>\n

Acefone’s\u00a0AI\u00a0voice agents now run on a two-level Retrieval-Augmented Generation (RAG) system instead. Global knowledge bases apply account-wide, for policy or company-level documents. Local knowledge bases apply to\u00a0a single product\u00a0or voice agent.<\/span>\u00a0<\/span><\/p>\n

Documents get chunked, embedded, and stored as vectors. Each query then pulls only the relevant chunk through\u00a0a semantic\u00a0search. Prompts stay lean, and hallucination rates drop.<\/span>\u00a0<\/span><\/p>\n

Your Brand Name, Pronounced Correctly<\/span><\/b>\u00a0<\/span><\/h2>\n

Every TTS provider has a\u00a0normalization\u00a0layer, and it\u00a0doesn’t\u00a0always get things right.\u00a0Let’s\u00a0say you hop on a test drive booking call with\u00a0“MG Motors”\u00a0and the voice agent\u00a0responds,\u00a0“Milligram Motors”.<\/span>\u00a0<\/span><\/p>\n

TTS Dictionaries let you override that default\u00a0normalization. Set “MG” to always\u00a0render\u00a0as “M-G,” manually or through bulk upload via Lexicon, CSV, or TXT files.\u00a0It’s\u00a0a small fix with an outsized impact on how professional a voice agent\u00a0sounds.<\/span>\u00a0<\/span><\/p>\n

What’s Coming<\/span><\/b>\u00a0<\/span><\/h2>\n

Acefone’s\u00a0AI\u00a0voice agent platform is moving to a full DIY model in September, covering Build, Test, Deploy, and Observe in one place. Voice agent to voice agent transfer is already live.<\/span>\u00a0<\/span><\/p>\n

Testing now spans telephony calls, chat-only LLM testing, and real-time WebRTC streaming. Teams can\u00a0validate\u00a0behavior\u00a0before a single production call goes out.<\/span>\u00a0<\/span><\/p>\n

If pronunciation errors or bloated prompts have cost you calls before, these three fixes were built for exactly that.<\/span>\u00a0<\/span><\/p>\n","protected":false},"excerpt":{"rendered":"

Most latency numbers you see for voice agents aren’t what the customer actually feels.\u00a0Most knowledge setups either bloat every prompt or go stale.\u00a0 Most TTS engines still mispronounce your own brand name.\u00a0\u00a0 Here’s\u00a0how we\u00a0changed\u00a0that.\u00a0 Latency Now Means What the Customer Feels\u00a0 Industry reporting typically tracks AI Infra Latency, the sum of STT Latency, LLM Latency, […]<\/p>\n","protected":false},"author":42,"featured_media":27111,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[325,332],"tags":[],"class_list":{"0":"post-27108","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-voice-bot","8":"category-whats-new"},"_links":{"self":[{"href":"https:\/\/www.acefone.com\/blog\/wp-json\/wp\/v2\/posts\/27108","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.acefone.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.acefone.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.acefone.com\/blog\/wp-json\/wp\/v2\/users\/42"}],"replies":[{"embeddable":true,"href":"https:\/\/www.acefone.com\/blog\/wp-json\/wp\/v2\/comments?post=27108"}],"version-history":[{"count":3,"href":"https:\/\/www.acefone.com\/blog\/wp-json\/wp\/v2\/posts\/27108\/revisions"}],"predecessor-version":[{"id":27113,"href":"https:\/\/www.acefone.com\/blog\/wp-json\/wp\/v2\/posts\/27108\/revisions\/27113"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.acefone.com\/blog\/wp-json\/wp\/v2\/media\/27111"}],"wp:attachment":[{"href":"https:\/\/www.acefone.com\/blog\/wp-json\/wp\/v2\/media?parent=27108"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.acefone.com\/blog\/wp-json\/wp\/v2\/categories?post=27108"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.acefone.com\/blog\/wp-json\/wp\/v2\/tags?post=27108"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}