When we saw a video of Jimmy Hosang passionately proclaiming that CMS was dead, we had to invite him for a chat.
What followed was the most open and honest conversation we've had about AI since it became mainstream.
In this interview, we cover:
⚰ Why CCaaS might be on the way out
♟️ How BPOs can adapt to the rise of AI agents
🤝 What enterprise buyers will still need from BPO
🐍 The cause and effect of AI "snake oil" salespeople
💡 And much more.
---
Q: Before this chat, you told me that BPOs shouldn’t make any CCaaS decisions in the next two years because of technological progress. Can you elaborate on that and offer up some alternatives?
A: Sure. CCaaS was originally a way to avoid massive engineering costs, especially in managing telephony platforms. But with agentic AI—which refers to autonomous systems that can act on their own accord—we’ll soon have AI agents that can listen to calls, understand customer intent, and route it to another appropriate machine or human. You’ll also be able to build up the complexity of what a voice-based, AI-powered IVR agent can do. First, it’ll be able to handle basic IVR functions, followed by simple requests. Eventually, it will be able to handle complete requests with automation, both responding to information and taking action on behalf of customers.
With that in mind, clients are asking BPOs to manage multiple CCaaS vendors, but as the tech improves, it doesn’t make sense to develop separate AI assistants for each one. BPOs should instead work on developing generalized AI that plugs into all the CCaaS platforms. That way, you only have to train one model once, then tune it, and it becomes a unique value proposition for the business.
Q: What’s stopping enterprise buyers from internally developing agentic AI that can automate any business process, including their contact centers?
A: Mainly, the complexity of the processes, operations, and decision-making. Buyers often become obsessed with shiny new SaaS solutions, thinking they’ll solve everything. However, cross-functional decisions are required to scale, and each department often wants to make its own choice.
In large operations, you might also have several BPOs, each using different CCaaS platforms. Building autonomous agents that can operate across all these systems is complex and overwhelming, so enterprises will want to look to BPOs for help.
Q: Many argue that current AI solutions are unreliable, ineffective, and overhyped. What would you say to those who have been burned and are now skeptical?
A: I’m actually sympathetic to that viewpoint. There’s a lot of snake oil being sold. We’ve seen it with speech analytics, live chat, and chatbots—they never reduced the need for human agents. If you throw too much money at AI upfront and it fails, you’ll be left with scars—and that’s what’s happened to many companies in the market—my advice: don’t make any long-term decisions around a single system right now. Focus on implementing AI that you own, control, or license, which gives you the flexibility to use it in various ways, like hiring a human who can be highly productive. Start small, experiment, learn, fail, and then try again.
Plus, people have built up this framework in their minds about how things should work, even technologists who’ve been steeped in SaaS. They still see the world as humans interacting with data and websites through the same old user interfaces, but that’s about to change as agentic AI does most of the interactions for us—that’s what people are struggling to wrap their heads around.
Q: What’s another common misconception holding people back from successful AI implementation?
A: Over the last decade, we’ve been led to believe that broken processes require a massive digital transformation—upgrading CRM, CCaaS, WFM, or even consolidating everything into a single platform. Yet, if you observe an agent’s workflow, they’re still constantly toggling between multiple screens, so perhaps that approach was always flawed.
What many leaders don’t realize is that you don’t need to fix broken processes before implementing AI since it can manage them more efficiently than human agents ever could. We've already developed agent assist tools that can navigate through flawed systems and enter data across multiple CRMs. So, while there's a prevailing belief that everything must be fixed before adopting AI, the reality is AI can handle your existing processes—even if they’re imperfect.
—
Connect with Jimmy on LinkedIn here to continue the discussion.