Enterprises evaluating contact center as a service (CCaaS) platforms in 2026 are not waiting for vendors to solve AI. They are building around it, layering in custom solutions to cover use cases that native CCaaS AI cannot yet address. That is the central finding from ETR's Observatory for CCaaS. In March 2026, an ETR Insights panel of senior technology leaders convened to discuss the report's findings, revealing that vendor selection today is driven less by feature comparison than by regulatory fitness, cloud alignment, and the flexibility to bring AI in from outside the platform.
The ETR Insights panel brought together a chief software architect at a large higher education institution, a CIO at a business services firm, a global head of advanced data analytics at a large financial services institution, an enterprise architect at a second global financial institution, and a chief software architect working on Defense projects. Their discussion covered contact center AI trends, agentic AI use cases, pricing model shifts, identity and access management challenges for AI agents, and market consolidation prospects. The key findings are below.
Vendor selection is driven by regulatory fitness and cloud alignment. Security posture, encryption standards, and cloud strategy determine which vendors even make the shortlist. Genesys and NICE are favored in regulated environments that early cloud-native providers could not serve; Amazon Connect gains traction but is routinely disqualified where cloud-agnostic mandates apply.
Native AI is not compelling, so enterprises are building around it. No CCaaS vendor has done anything remarkable with AI yet. Instead, enterprises are layering in custom or third-party solutions, such as Bedrock and Infosys Cortex, to serve use cases that extend beyond the contact center. Integration flexibility, not embedded AI functionality, is the current differentiator.
Agentic AI will land in back-office operations before it touches customers. Workforce management, quality monitoring, and call deflection are near-term targets: workflows where imprecision is tolerable. Customer-facing deployment demands much higher accuracy and stronger governance, particularly around identity, auditability, and regulatory compliance.
AI is beginning to pressure pricing models toward hybrid structures. Per-seat licensing still dominates CCaaS, but panelists note hearing of models that pair agent licenses with token-based consumption. As AI agents absorb work historically tied to headcount, the shift will be cultural as much as financial.
Too many vendors, too little differentiation. Feature and price parity makes RFPs harder without making outcomes better. Panelists expect consolidation to accelerate AI-driven innovation, though some argue the real disruption will not come from incremental platform upgrades but from architectures rebuilt around AI from the ground up.
The panel of senior technology leaders described a CCaaS landscape where vendor selection is often driven less by feature comparison than by regulatory fitness, existing cloud commitments, and the gravitational pull of merger-driven technical debt. To set the stage, the business services CIO mentions that traditional Avaya physical infrastructure was the starting point for many organizations. "If you've had a contact center for years and years—which most people have—Avaya for the longest time was really the only player in town." When it came time to migrate to the cloud, they chose RingCentral, specifically on ease of configuration, support, and integration, qualities they claim Avaya lacks. "Avaya's current CCaaS platform, like their cloud-centric experience platform, I would not describe that as an easy platform to stand up, to support, to use, and to do all kinds of point integrations, like workforce management, call recording platforms, sentiment engagement, and things of that nature. RingCentral offered that." Cloud provider alignment also matters: this CIO's organization is Microsoft Azure-centric, effectively disqualifying Amazon Connect. "RingCentral, and many of the others that are listed here, are neutral players. You could be any kind of cloud—any kind of other products for that matter—and then they'll bolt bespoke around it."
The global head of advanced data analytics describes a more complex environment spanning Genesys, NICE inContact, and Aspect Software, with both on-premises and cloud segments. Their organization could not leapfrog from Avaya directly to cloud-native providers, because those platforms were not ready for regulated environments at the time. "The encryption posture was not adequate, and the security posture was not adequate. They had to almost play catch up and bolt on." Again, Amazon Connect's traction is notable but remains a non-starter for organizations committed to cloud-agnostic architectures. "Taking a native hyperscaler offering is just not aligned with our mandate and directive." The enterprise architect echoes this, with a technology stack driven primarily by mergers and acquisitions, now including Avaya, Twilio Flex, NICE, and Genesys. Consolidation is the goal, but regulatory requirements around country-specific sentiment analysis and telephony compliance make a single-platform strategy difficult. "When you are doing, for example, sentiment analysis of contact components from a customer perspective standpoint, some of the regulations go that you have to be country-specific for some of them, so it's a heavily complex type of environment."
The chief software architect represents the most security-constrained use case. Their organization migrated from Zoom to Amazon Connect not because of feature superiority, but because of infrastructure alignment. "For all our cloud computing, our move to the cloud, AWS is our partner. So, what we are trying to do is lock down the infrastructure exhaustively, so that nothing gets out of a single in-and-out network. Zoom could be better than AWS—I don't know if that's true or false—it's just a lot of integration we [would have] needed to do."
AI capabilities are increasingly central to CCaaS vendor evaluation, but panelists broadly prefer integrating custom AI solutions into CCaaS platforms over relying on what they view as immature native offerings from platform providers. The business services CIO is blunt in their assessment: no CCaaS vendor has done anything remarkable with AI yet. More compelling is to build custom solutions that span channels beyond contact center, like self-help on customer-facing websites, email platform integration, and other support mechanisms that a single CCaaS provider's native tools would never cover. "I like the idea of being able to bring a custom solution in to fit our very specific use case, and to provide an omnichannel solution that may be bigger than what's going on in CCaaS. For example, self-help on a customer-facing .com, or an integration with an e-mail platform that might also be part of the support mechanism."
In ETR's SaaS Displacement Drill, contact center is at risk of AI-driven displacement. Categories in the upper right are the most disrupted, seeing both high AI adoption and high vendor displacement. Customer support/contact center and vertical/industry-specific software show notable rising forward-looking displacement.
The financial services data analytics leader describes a more layered reality where multiple AI delivery methods coexist and compete across different contact center segments. Their Genesys platform runs an instance of NICE inContact, and Aspect Software, which is already outdated, creating a patchwork that is expensive and complex to maintain. It also provides a natural testing ground across retail, wealth management, global banking, and global markets, letting this organization see what actually works before committing to a single blueprint. "Our proliferation obviously comes at cost and complexity, but gives us some latitude to hedge the bet, and see which one has efficacy and probably will become a more standard blueprint, and which one just does not work."
Another panelist identifies Genesys and RingCentral as the only two platforms positioned to deliver a truly integrated, AI-enabled experience across differences, and consolidation is the deciding factor. "The key word here is everything around integration—the conversational experience that any of the CCaaS vendors in that space can provide." At the most constrained end of the spectrum, a DoD-affiliated architect is bypassing vendor AI entirely, building conversational understanding models on AWS Bedrock to process survey data from non-native English speakers within a fully contained security perimeter. "The ins-and-outs of these transactions stay within our data center, or the data center in the cloud."
Panelists see agentic AI as significant in contact center but draw a sharp distinction between back-office automation and customer-facing deployment, where there is no room for error. The first productive use cases for agentic AI are internal: workforce management, monitoring, quality assurance, and other operational workflows. "When you are scheduling, managing workforce, monitoring things on the quality side, that's ripe for it. I'm more excited about doing it there than on the customer side." Those are ripe use cases for agentic AI precisely because the stakes are lower. Panelists also distinguish between agentic automation and AI-powered knowledge management, the latter being less about "do it for me" and more about equipping human agents with deeper product understanding and runbook generation. Another enterprise architect describes agentic solutions as the new "front door" before human escalation, handling call routing, auto-resolution, and deflection, but imagines that headcount reduction will be gradual and that impact will vary by line of business. Wealth management and inbound versus outbound operations each present different automation thresholds.
Pricing models may begin to shift as AI capabilities change what a CCaaS "seat" actually means. Historically, costs have been mostly structured around per-user licenses. "Contact center in general is associated with the number of people you have, the number of agents you have in the space, the furniture, the software, so on and so forth." But they acknowledge that if CCaaS providers embed sophisticated AI natively, consumption pricing could accelerate. Another executive has already had conversations with providers about emerging hybrid models that combine token-based pricing with per-agent licenses. "The CCaaS provider, these agentic solutions, these agents that we are going to be deploying, I will welcome a license like a human agent, but it's not pure consumption based on the quantity or volume of data that you are going to be doing, or the amount of calls that we are going to be receiving and taking. It's more really a pricing model based on the tokens that these agents are going to be consuming."
In the Observatory survey, ETR asked respondents which pricing model they prefer for CCaaS products. Per-user licensing remains the clear preference, followed by site licenses and consumption-based pricing. Per-device licenses trail.
In the most security-constrained environments, the challenge is not what AI agents can do but whether their actions can be audited and attributed to the standards required by federal oversight. "We need to identify them by an identity and access management control, which gets very hairy as these agents travel from system, to system, to another system." The core issue is traceability: when an agent retrieves data on behalf of a human, both the agent's action and the human's authorization must be logged to a specific security clearance level. "If somebody wants to see a set of data, the agent goes and grabs it. We need to log that, which person has reviewed that documentation, and what kind of security clearance this person has. The laws have not caught up with the technology. It's not that we cannot do it, but then how do you certify an agent just like a human being?"
Looking ahead, panelists describe a market with too many vendors offering too little differentiation, where AI adoption could either define competitive advantage or accelerate obsolescence. One sees consolidation as both likely and healthy. "There's a lot of feature parity. There are differences that suit different use cases better than others, but I think the space is very fractured." Running RFPs in this environment is needlessly complicated. "There are not big differences one player to the other that helps make the RFP easier, that suits a particular industry or particular use case. If that consolidation happened, I think that would accelerate innovation, specifically around AI enablement."
Another frames the future in binary terms: CCaaS providers will either leverage AI as an advantage or lose seats to it. One financial services leader expects on-premises CCaaS deployments to largely disappear over time and self-service to become the dominant theme. "The whole CCaaS moves to the cloud wholesale, I don't imagine anything left on-prem." But they believe open architecture will persist over bundled platforms. "I don't think anyone will be able to play an end-to-end platform bundle solution." A third argues that current CCaaS vendors, and all of us, are fundamentally misunderstanding what AI portends. "I think in the future, you will see a groundbreaking platform coming up that's not based on previous buildups. It's not like how RingCentral 2.0 came from 1.0 with more features. We just don't understand the magnitude of agent and the AI implications. It needs to be rebuilt, rethought, and understand the disruption very clearly."
Observatory reports don't usually leave the ETR Platform. This one is the exception. For a limited time, the full Observatory for CCaaS report is available to download: complete vendor rankings, spending data, ROI expectations, and feature-level analysis across 13 platforms, straight to your inbox. Fill out the form to get it.
Findings from ETR Insights 467: Observatory for CCaaS — Data Feedback Panel. March 2026. Panelists represent senior technology leadership across financial services, higher education, business services, and defense-adjacent industries. Vendor mentions reflect panelist commentary and do not constitute ETR endorsement.