{"slug": "getting-voice-infrastructure-right-when-deploying-voice-ai-for-cx", "title": "Getting voice infrastructure right when deploying voice AI for CX", "summary": "Contact centers adopting Voice AI must still route most voice traffic over the PSTN, making telephony integration central to customer experience deployments, according to an analysis of voice infrastructure for CX. The article states voice remains the primary channel for contact center interactions even as digital channel usage grows, and argues that contact centers should own their voice infrastructure when adopting AI. Voice AI, described as the latest AI wave after ChatGPT and Agentic AI, is at an early adoption stage but is positioned as foundational to modern CX.", "body_md": "Across all verticals and regions, contact centers are under pressure to modernize and improve customer experience (CX). With today’s consumers being digital-centric and mobile-first, CX expectations are higher than ever, and IT and contact center leaders are struggling to keep pace, especially when it comes to adopting new technology.\n\nVoice AI is the latest technology wave for improving CX, and while most contact center leaders want to adopt it, they need a fuller understanding of all that is needed for voice and AI to work together effectively. This is especially true for enterprise-scale contact centers that must support a global customer base.\n\nWhile AI can easily scale in these environments, most voice traffic must still traverse the PSTN. Customer usage of digital channels [keeps growing](https://www.salesforce.com/service/what-is-digital-customer-service/) – including voice-capable apps like WhatsApp and Viber, but voice remains the [primary channel for contact center interactions](https://www.contactcenterpipeline.com/Article/predictions-for-the-us-contact-center-industry-in-2025).\n\nBoth modes of voice are important, but telephony-based voice is central for CX, and as contact centers adopt AI, the integration of voice with AI will be essential. This article will review the benefits of this integration, the key capabilities needed for Voice AI and why I think it’s so important for contact centers to own their voice when adopting AI.\n\nFollowing in the footsteps of ChatGPT and Agentic AI, Voice AI is the latest AI wave. While Voice AI is currently at the early adoption stage, I see it becoming foundational to modern CX, just like these predecessors quickly became.\n\nBeing so new, some explanation is in order. Voice AI refers to AI applications or platforms that are voice-based, meaning that voice can be used for both inputs *and* outputs when human customers make inquiries with a contact center.\n\nCustomers can use several modes to make these inquiries, but the “input” usually starts with voice, which the Voice AI platform is able to understand – and now can do so at a much higher level than with legacy voice technologies. For the “output”, Voice AI can respond to the customer using speech (or text) in a conversational manner that can often take the inquiry through to completion.\n\nIn more familiar terms, Voice AI applications are known as digital assistants, automated voice agents, or intelligent chatbots. Customers have been using legacy versions of this for years – think of the traditional IVR – but with today’s AI, the voice experience feels more natural. Furthermore, these applications can understand voice at a much deeper level, by integrating more extensively with back-office systems, leading to better CX outcomes, both with human agents and for self-service.\n\nOnce contact center leaders are comfortable with the Voice AI concept, they need to think about it more practically – how it impacts CX. This is what drives business value, and to frame that, here are some leading examples:\n\nWith the above in mind, I think the converse outcomes speak even more loudly when either voice or AI – and especially together as Voice AI – do *not* perform as intended. Not only would this negate the above benefits, but the negative impact will be even more consequential. As noted, most contact centers urgently want to modernize with cloud and AI – precisely to avoid these negative consequences.\n\nWhile this article focuses on AI-ready contact centers, the broader intention is to build a case for Voice AI that will help legacy-based contact centers advance their plans for getting on this much-needed path. No other channel has a greater impact on CX, and the Voice AI opportunity adds a new dimension of business value with voice, along with leveraging AI for automation and customer engagement wherever they are, and whatever language they speak.\n\nTechnology aside, my underlying message for *all* contact centers is to *own* your voice. The best way to own your voice is by having the right infrastructure in place. To varying degrees, contact centers have long relied on purpose-built voice infrastructure, since so much of their voice traffic is still PSTN-based. However, as contact centers adopt AI, it’s easy to overlook the complexities of integrating telephony-based voice into AI applications.\n\nVoice infrastructure is technical by nature, and there are some key elements that IT and contact center leaders must keep in mind when deploying AI in tandem with voice. In [AVOXI’s 2025 State of International Voice Report](https://www.cio.com/www.avoxi.com/wp-content/uploads/2025/09/State-of-International-Voice-for-the-Contact-Center_2025-Report.pdf), 77% of enterprises cited difficulty consistently achieving high-quality voice, highlighting an existing challenge that becomes even more important as contact centers integrate voice with AI. This is especially relevant for contact centers serving a global customer base, and at minimum, voice infrastructure must include the following:\n\nThere are other elements to voice infrastructure as well, and the key idea here is to explain why these are fundamental capabilities; not just to support the most important channel your customers rely on, but for voice to be effectively integrated with the AI platforms and applications contact centers are now rushing to deploy. The visual below illustrates how the main infrastructure pieces must interwork to support Voice AI.\n\nJon Arnold\n\nMore than anything else, voice infrastructure is about minimizing latency, given the real-time nature of voice communication. When customers speak to human agents, some incremental latency can be tolerated by both parties, but when the customer is relying on an AI agent to understand them, that same degree of latency – or worse – will lead to a multitude of problems, both during and after the interaction.\n\nThe AI part of Voice AI isn’t advanced enough to respond accurately when there are gaps in the conversation, echo, choppy audio, garbled speech, etc. That’s the scenario contact center leaders need to avoid – not just for poor quality bot-to-human communication, but for inaccurate speech-to-text outputs like transcriptions and call summaries. This will be the result when overlooking voice infrastructure, and it will completely defeat the purpose of using AI.\n\nAgain, the best way to do that is to put just as much consideration into the voice infrastructure piece as with AI platforms. By nature, neither AI nor CCaaS platforms can natively address the voice capabilities outlined above, hence the importance of owning your voice.\n\nWithout a proper voice infrastructure, Voice AI will simply not be effective. No matter how “intelligent” the AI platform, these capabilities are not native or comprehensive enough to manage telephony-based voice, especially at scale and on a global basis.\n\nThe key point to recognize is that both AI and CCaaS platforms will be at their best when they receive voice streams that have already passed all the hurdles required to make them AI-ready. Considering all the things that can go wrong with this traffic if not managed properly, and the computational resources needed for this, bolting voice on to one of these platforms is risky at best.\n\nIT and contact center technology leaders *could* choose to leave the job of monitoring voice performance up to an AI or CCaaS vendor. However, many contact centers rely on multiple vendors, which is challenging when adopting new technologies. There are various issues to consider:\n\nMaking good decisions here really comes down to what’s most important. When adopting AI, IT and contact center technology leaders must balance two priorities, neither of which should be compromised: providing the highest possible quality voice experience with telephony, *and* providing the most impactful forms of automation with AI – especially for self-service.\n\nEven though telephony-based voice isn’t the only mode for voice, it’s the most important one for CX. PSTN calling will be with us for years to come, and when architecting your voice stack, telephony must be a core consideration. App-based voice options have their virtues, but for today’s AI deployment decisions for CX, telephony will have a far greater impact.\n\nThis brings us to the familiar best-of-breed versus all-in-one trade-off that contact center leaders must consider with Voice AI. In short, telephony-based voice is best handled by voice infrastructure vendors, just as AI applications or CCaaS applications are best handled by those platforms. Not only will you get the best AI outcomes with best-of-breed for voice infrastructure, but the contact center will have much more flexibility in a constantly changing environment:\n\nTo make the right decisions about voice infrastructure, contact center leaders must consider what each of the following options can offer:\n\nThe core theme here has been the importance of owning your voice, and how doing so requires an understanding of the role voice infrastructure plays with Voice AI. As contact center leaders look to add Voice AI to their expanding portfolio of AI capabilities, my intention is to make clear how considerations for supporting voice infrastructure are just as important as those for conversational AI.\n\nWhile complementary, the requirements for each are quite different, and both are best served by best-of-breed partners. As AI platforms become CX differentiators, and as long as voice remains the most important communications channel for CX, voice infrastructure is fundamental to the success of Voice AI. This article makes the case for best-of-breed, not just for getting the best outcomes with Voice AI, but to provide the most flexibility for contact centers as their deployments continue evolving.", "url": "https://wpnews.pro/news/getting-voice-infrastructure-right-when-deploying-voice-ai-for-cx", "canonical_source": "https://www.cio.com/article/4221810/getting-voice-infrastructure-right-when-deploying-voice-ai-for-cx.html", "published_at": "2026-09-15 10:00:00+00:00", "updated_at": "2026-09-15 10:09:29.502915+00:00", "lang": "en", "topics": ["ai-products", "natural-language-processing", "ai-agents", "ai-tools"], "entities": ["ChatGPT", "WhatsApp", "Viber", "PSTN", "IVR"], "alternates": {"html": "https://wpnews.pro/news/getting-voice-infrastructure-right-when-deploying-voice-ai-for-cx", "markdown": "https://wpnews.pro/news/getting-voice-infrastructure-right-when-deploying-voice-ai-for-cx.md", "text": "https://wpnews.pro/news/getting-voice-infrastructure-right-when-deploying-voice-ai-for-cx.txt", "jsonld": "https://wpnews.pro/news/getting-voice-infrastructure-right-when-deploying-voice-ai-for-cx.jsonld"}}