CCW Vegas 2026 is done. And I will say it plainly: this is the best customer contact and CX event I have ever attended. Sad to see it finish, but for those who could not make it to Caesars Forum this week, here is what you missed — and why it mattered.
As a technology journalist covering this space, I have never seen a year quite like this one. The vendor innovation on show was extraordinary — agentic AI dominated every conversation, every booth, every keynote, with a level of genuine product substance that felt different from previous years. But what was equally striking, and perhaps more telling, was the tension in the room. The organizations trying to deploy this technology are going through something genuinely painful right now. They are damned if they do and damned if they don’t. CX leaders are under significant pressure to move on AI and fast. And yet deployments are struggling, reversals are common, and the gap between the promise and the live reality is wider than most will admit publicly.
What was truly admirable, though, was the spirit in the room. Nobody had given up. Everyone was motivated to experiment, to learn, to figure it out. And that is ultimately what CCW Vegas does better than any other event in this space — it grounds the conversation in reality while keeping the ambition intact. CMP Research’s Nicole Kyle, whose work is pivotal to how this event is shaped and grounded, put it best in our interview: the research captures both the gritty edge of truth — how hard customers are finding this — and the insights that give CX leaders a genuine path forward. That combination is rare, and this event had it in abundance.
Over the course of the week I was fortunate enough to interview some of the most important vendors in this space — Salesforce, Microsoft, Zoom, RingCentral, Amazon Connect Customer, UJET, CallMiner, and Content Guru — and to speak with their customers, including EasyJet and Casio, about what this all looks like from the buy side. In between, I dropped into sessions that ranged from brutally honest to genuinely inspiring. CCW itself is growing — significantly this year, and expected to grow further next year. The market is buoyant, and this event reflects that.
Here are the ten trends that defined the week — and the state of the industry right now.
1. The Autonomous Agent Era Is Here, and the Accountability Questions Have Followed
AI agents are no longer a roadmap item. They are in production, handling real customer interactions, carrying authority over workflows that previously required a human decision at every stage. According to CX Today data, narrative volume around autonomous AI agents grew from 23 tracked instances in February 2026 to 893 in June. That is a 38-fold increase in five months. The shift from “copilot” to “autonomous enterprise workflow” has happened in about eighteen months, and the CX industry is now answering the accountability questions that shift entails in live environments rather than controlled pilots.
From the Floor
Two of the most concrete demonstrations of this shift arrived at CCW this week. RingCentral announced the expansion of AIR Pro, embedding native AI agents directly into RingCX workflows — handling inbound and outbound interactions across voice and digital channels end to end, including appointment confirmation, identity verification, and payment processing, all within a single call. On the same day, Zoom unveiled Agent Architect for Zoom Virtual Agent: a generative tool that turns a plain-language prompt into a production-ready AI agent, complete with agentic reasoning for dynamic interactions and deterministic logic for compliance-sensitive steps. Both launches made the same argument — that the agentic era is not approaching, it is here — and both shipped with the governance and testing infrastructure that Trend 2 below explains why you cannot afford to skip.
2. Three Quarters of AI Agent Deployments Are Being Reversed
A Sinch report published this year found that 74% of enterprise AI agent deployments have been reversed after go-live, with governance failures cited as the primary cause. The failures did not emerge during deployment. They emerged afterwards, when agents encountered edge cases or caused brand damage no checklist had anticipated. Gartner estimates the average cost of a failed enterprise software deployment runs into millions when remediation and reputational impact are included. The organizations not in that 74% invested in the governance layer before they needed it. That sequence is the difference.
“The question we thought we needed to answer was, can we deploy this. The question we actually needed to answer was, can we govern this once it is running. We are only just realizing those are different questions.”
From the Floor
The Amazon Connect Customer session this week was titled “Be the 5%” — a reference to the MIT finding that 95% of AI pilots never reach production. It was one of the most practically useful sessions at the event, and deliberately so: both Hannah Bloking, Senior Manager of Solutions Architecture at Amazon Connect, and Andrei Papancea, Senior Manager of Software Development at AWS, have led large-scale AI implementations themselves. They were not presenting theory. They were pattern-matching from failures they had lived through.
Four failure patterns dominated the room’s own polling. The first: leading with “let’s use AI” rather than a specific business problem. Bloking was direct — the most common first conversation she has with customers is still “what are you actually trying to accomplish?” The second, and the one that came top of the room’s vote: knowledge is not AI-ready. Bloking’s framing here was sharper than the familiar “garbage in, garbage out” warning. The real problem, she argued, is that the knowledge you actually need is not documented anywhere. It is in people’s heads. Your best agents do not read the policy — they apply years of human judgment on top of it. Getting that into a form an AI agent can use is harder and slower than most project plans account for. The third pattern: governance becoming a bottleneck, with legal and security teams that are asked to approve things they have never seen before. Bloking’s advice was pointed — you have authority to push back, and learning to use it is part of the job. The fourth: building a new tool that nobody uses, creating a silo rather than embedding AI into the existing flow of work.
Papancea added the failure pattern that goes unspoken most often. The distance from zero to prototype has collapsed — anyone in the room can spin up an AI agent in minutes. The mistake is assuming prototype-to-production takes as little time. It does not, and the organizations that treat those two distances as equivalent are the ones that stall.
The playbook the 5% actually use came down to three things. Capture the human judgment that lives in your best agents’ heads and treat knowledge as a product with a continuous update process, not a one-time project. Embed AI into the flow of work rather than building it as a separate tool with its own name and interface — Bloking’s line was clean: “take the ego out of your project.” And move fast as a deliberate strategy, not a risk. “I don’t aim for perfection. I aim for being first, and I aim for incremental.” Ship something end-to-end, learn from it, and iterate. The organizations that wait for a complete solution before going live spend six to twelve months not learning anything.
The self-assessment Bloking left the room with is worth repeating here: Can you name the specific business problem you are trying to solve — not “use AI”? Is your knowledge AI-ready, or just documented? Can the people closest to your customers build and iterate, or does everything route through engineers? What is your change management plan for both humans and AI agents? If those questions feel uncomfortable, Bloking’s view is that the discomfort is the right place to start.
3. Zero Contact Center Agents Find AI Essential
A UJET study published this year found that not a single contact center agent describes AI as essential to their daily work, despite most using AI tools every day. At the largest customer experience event in the United States, where AI is the dominant conversation, the professionals closest to the customer are quietly finding none of it indispensable. Salesforce research found that 88% of customers say the experience a company provides matters as much as its products. The people delivering that experience every day have delivered a verdict worth listening to.
From the Floor
I spoke with Matt Clare, VP of Product Marketing at UJET, on the show floor this week, and the conversation went straight to this finding. UJET’s response to it has shaped their strategic pivot — from CCaaS company to CX AI company — built around reducing what Clare described as cognitive load on agents rather than simply layering more tooling on top of them. Their Spiral analytics product and the Headless SDK approach reflect a deliberate bet: that adoption follows simplicity, not capability. If agents do not find AI essential, the question worth asking is not whether the AI is powerful enough. It is whether it was designed around how agents actually work.
That workforce readiness gap came up again in a sharp conversation on the floor with Jeremie Abou, Sales Lead at Diabolocom. The interview is worth your time. His framing — that 89% of CX leaders expect humans to stay central to support, while only 17% believe their agents are actually ready for it — is one of the more honest data points to come out of this event. Abou draws a clean line between AI theatre and operationally useful technology, and makes the case that the contact center of the future may end up more human, not less. Presence, in his view, beats pure automation as a CX strategy. Given the finding that zero agents find AI essential today, it is an argument worth sitting with.
4. Verizon Is Replacing Agents. IKEA Is Retraining Them. The Industry Has Not Decided Who Is Right.
Verizon CEO Dan Schulman stated publicly that AI agents are already replacing customer service workers and that satisfaction scores have improved as a result. IKEA took the opposite position, redesigning agent roles entirely and retraining people for revenue-generating work rather than replacing them. McKinsey research found that companies involving frontline workers in technology transformation are 2.6 times more likely to achieve successful adoption. Meanwhile, Salesforce CEO Marc Benioff has argued that executives blaming AI for layoffs are being lazy, and that financial overextension is the real driver. These are the most senior people in the room, and they do not agree.
5. Efficiency Was the Wrong Goal All Along
The Festival of Work 2026 found that an intense focus on AI-driven efficiency is generating unintended negative consequences for businesses and employees simultaneously. Customer service automation is producing customers more frustrated at the end of an interaction than at the start, because the system was designed for deflection rather than resolution. A Front Research survey of 700 B2B CX leaders found AI tools are increasing operational burden rather than reducing it, using the phrase “coordination tax” to describe what happens when tool proliferation adds management complexity rather than removing it. Harvard Business Review has found that customers who have had a genuinely good service experience spend 140% more than those who have not. Optimizing for speed over resolution is, it turns out, also a poor financial decision.
“We were measuring deflection rate as our primary AI success metric. We were getting better at deflecting customers. We were not getting better at serving them.”
From the Floor
Pasquale DeMaio, VP of Amazon Connect Customer, is making the same argument from the product side. His position — stated plainly at the main stage panel on Wednesday alongside speakers from United Airlines and Citizens Bank — is that deflection is dead and that the industry has been optimizing for the wrong metric for years. Amazon Connect Customer’s Live Sync technology, launched this week at CCW, is the product expression of that argument: a patented capability that synchronizes a voice conversation with a real-time on-screen interface, so the customer can tap to confirm, select, or act during the same continuous call without switching channels or starting over. United Airlines deployed it to handle Irregular Operations, went from first build to production in under three months, and the results they shared in their dedicated session on Thursday made a pointed case that resolution and automation do not have to trade off against each other.
6. Only 10% of Organizations Have Achieved Omnichannel Maturity, and Adding Channels Is Making It Worse
CCW Europe research found that just 10% of organizations have achieved genuine omnichannel maturity despite years of investment. The combined annual cost of poor customer service sits at $75 billion according to Forrester. Technology fragmentation is a significant contributor. The finding that surfaced most honestly in conversations this week is that adding channels often makes things worse because the underlying data and process problems persist across every new touchpoint. PwC research found 59% of consumers will leave a brand they love after several bad experiences. Scaling a broken experience is not transformation.
7. Orchestration Is Replacing Automation as the Word That Actually Means Something
The conversations at CCW Vegas that felt most grounded were the ones using “orchestration” rather than “automation.” The distinction matters. Automation replaces a step. Orchestration coordinates the whole: AI, human agents, data, and workflows working toward a specific customer outcome with the right element handling the right part of each interaction. Forrester Consulting research found that most brands are failing to see business outcomes from AI investment despite significant spending. The gap is not a capability problem. It is a design problem. The organizations seeing results are designing connected systems, not deploying disconnected tools.
From the Floor
The partnership announced this week between Alorica and Crescendo is the clearest vendor-level expression of this shift. Crescendo describes itself as the only AI-native CX platform built to run customer experience as one unified system — AI and human orchestration across the entire customer journey, not at isolated points of interaction. Alorica brings 100,000-plus agents across 16 countries and 25 years of live-channel expertise, particularly in voice and chat, where human judgment and context remain essential. The partnership explicitly addresses the gap between capability and operationalization: the inability of most enterprises to coordinate AI and human effort across live channels at scale. Matt Price, CEO of Crescendo, put it directly: “This is what happens when AI stops being a project and becomes the way CX actually runs.” That is a reasonable definition of orchestration, and it is a harder thing to build than most roadmaps suggest.
8. Consumer Trust in AI Is Fragile, and the Margin for Error Is Smaller Than Assumed
Amazon’s Rufus AI shopping assistant achieved a 32% product recommendation accuracy rate in testing published this year, with consumer trust eroding rapidly as a result. Zendesk’s CX Trends report found that 61% of customers would switch to a competitor after one bad service experience. MaxContact consumer research found that poor AI performance in early interactions creates distrust that is genuinely difficult to reverse. AI builds trust in well-governed deployments. In poorly governed ones, it does the opposite, and the reputational cost may be higher than most business cases have modeled.
9. The Platform Architecture Debate Is the Most Practical Fight on the Show Floor
Salesforce’s Agentforce Contact Center has sharpened a debate already present at CCW Vegas: consolidation versus modularity. UJET has argued publicly against what it calls the “integration trap.” Avaya CEO Patrick Dennis has challenged the cloud-first narrative, citing the latency and data sovereignty demands of voice AI. IDC research found organizations using five or more disconnected customer service tools spend 30% more time resolving issues than those using integrated platforms. The omnichannel maturity finding, however, suggests consolidation alone does not solve the underlying problem. It is a five-year decision being made in a three-month window, and the stakes are significant.
From the Floor
Two announcements this week landed on opposite sides of that debate in instructive ways. Zoom’s Agent Performance Suite — combining pre-deployment simulation, live resolution dashboards, Quality Management for both AI and human interactions, and knowledge base gap detection — represents a consolidation bet: that performance, quality, and analytics should live natively on the same platform as the agents they govern. UJET, meanwhile, is pushing the other direction with its Headless SDK and BYO-AI architecture with partners like Parloa, arguing that developer-first flexibility and the ability to bring your own AI layer is the only way to avoid locking yourself into a vendor’s capability ceiling. Both positions are coherent. Both carry risk. The honest answer is that the right architecture depends on what your team can actually govern, and most organizations are still working that out.
10. GEO Is the New SEO, and Most CX Brands Have Not Started Yet
The most quietly significant trend at CCW Vegas 2026 received the least amount of attention. Generative Engine Optimization, making content visible within AI-powered search surfaces, is emerging as a discipline distinct from traditional SEO. Adobe’s Brand Visibility tool launched this year is a formal market signal. BrightEdge research found that over 50% of searches now return an AI-generated overview before any organic result. Semrush data found AI overviews draw from a much smaller pool of sources than traditional search, placing greater premium on authoritative, frequently cited content. Brands not appearing in AI-generated answers are increasingly invisible to enterprise buyers. Most organizations at this event have not started addressing that.
The Takeaway: CCW Vegas 2026 Wrapped
Walking out of Caesars Forum on Thursday, the question I kept turning over was the same one I heard repeated in sessions, corridors, and vendor meetings all week: have we been solving for the right thing? The technology on show was the most impressive I have seen at any CX event. The investment across the industry is real and significant. And yet three quarters of deployments are being reversed, zero frontline agents find the tools essential, and the $75 billion annual cost of poor customer service is not improving.
None of that means the AI story is wrong. It means the deployment story is incomplete. What this event made clear — more clearly than any CCW I have attended before — is that the industry knows it, is willing to say so out loud, and is genuinely working to close the gap. That is progress. The most important question in customer experience in 2026 is not “can we deploy this.” It is “will this make the experience better for the human being on the other end of every interaction we design?” It is the oldest question in CX. This week, for the first time in a while, I felt like the industry was actually trying to answer it honestly.
See you at CCW Vegas 2027.
CX Awards 2026: The Biggest Year of Innovation Yet
If the conversations at CCW Vegas this week have confirmed anything, it is that the bar for what counts as genuine CX innovation has never been higher. The CX Awards 2026 is where the outcomes get recognized. Categories will be revealed in early July, alongside the opportunity to apply as a judge for this year’s program. You can register your interest now.
Register your interest in CX Awards 2026 →
Join over 40,000 CX professionals in the CX Today LinkedIn community to continue the conversation from CCW Vegas 2026. And subscribe to the CX Today newsletter for your weekly rundown of the most important CX industry developments.
