I run a company built on human agents. Here’s why I stopped worrying about AI replacing them.

Sometime last year, the conversation shifted. 

For nearly two years, every conversation with a client, board member, or peer in the contact center world centered on the same question: should we be putting AI agents in front of customers, and how quickly can we do it? Then, almost without anyone noticing, that question went away. Everybody had done it. The new question was quieter and a lot more awkward: okay, now what? 

Simply Contact put out a white paper in April with a title I keep coming back to: What Changes When AI Owns Outcomes. Their read of the market is that nearly everyone is past the pilot stage, but hardly anyone has actually rebuilt their operation around what they deployed. The AI sits on top of a support org designed for a world without it—automating interactions without necessarily addressing the underlying issues that caused customers to need help in the first place. And leadership is done being impressed by demos. They want resolution times, cost, and stability. They want it to work. 

I’ve spent most of 2026 living in that gap, and I’ve arrived somewhere I didn’t expect to. I no longer think the hybrid model, meaning AI agents and human agents working in the same queue, is a stopover on the way to full automation. I think it’s where we’re going to live. But hybrid alone isn’t the answer. The real advantage comes from orchestrating the two as one operating system.  

And I think the operators who accept that early are going to build something that the ones who are still waiting for humans to disappear never will. 

Let me try to explain why, using other people’s research rather than just my own opinion. 

The work is getting re-sorted, not removed. CX Today’s trends piece from the spring describes AI moving from “assisting with tasks” to owning whole workflows: pulling records, verifying identity, documenting the call, and resolving the routine stuff from end to end. Fine. But look at what’s left on the human side of that line: the negotiation, the judgment call, the customer who is really calling because they’re scared or angry and the billing question is a pretext. Nobody has scripted that successfully in forty years of trying. That’s not a smaller job. It’s a harder one, and it deserves better people and better pay than the industry has historically given it. 

The handoff is the whole game. Parloa came back from CCW Berlin in March talking about what they called the Great AI Divide. On one side, companies that can move a conversation from an AI agent to a human with the full context intact. On the other side, there are companies that can’t. Juniper Research’s Human + AI paper makes a related point from the technology side. The standardized agent frameworks now emerging matter mostly because they decide whether AI agents can actually talk to the humans and systems around them. I’d put it more simply. If your customer has to repeat themselves after an escalation, you don’t have a hybrid model. You have two broken ones stapled together. 

Nobody knows how to charge for this yet. No Jitter’s midyear review cites a Talkdesk survey where nearly every CX leader says the hybrid workforce is delivering value, and about one in five now think of AI agents as labor rather than software. Sit with that for a second. If AI agents are labor, then the seat-based and minute-based pricing models this whole industry grew up on are finished. You can’t price digital labor by the seat and you shouldn’t price human judgment by the minute. The market is going to move increasingly toward outcomes. I don’t know exactly what that looks like yet, but I know it’s coming, and I’d rather be early than right. 

Now the uncomfortable part, at least for people in my business. 

The classic BPO pitch of scale, labor arbitrage, and seats is precisely the layer that AI is eating. If the only thing you can say to an AI agent is “we have a cheaper human,” you’ve already lost. I say this as someone who has spent decades in an industry built around those economics. 

But here’s what I’ve come to believe. Every AI resolution engine throws off two things: a resolved interaction, and a huge amount of exhaust. What worked. What didn’t. Where the model hesitated. Where the customer’s tone turned. Which handoffs fell apart. Over time, those signals become something even more valuable: operational intelligence about where AI works, where it doesn’t, and how the entire customer operation should continuously improve. Most enterprises are sitting on a mountain of that data and reading almost none of it, because the people who could interpret it are busy taking calls. 

That’s the opportunity. The humans in a hybrid model aren’t just the escalation path. Done right, they’re the sensing layer. The agent handling the hard 15% is also the person best placed to explain why the other 85% went the way it did, and to push that back into how the AI agents, the routing rules, and the knowledge base get better next week. Human expertise becomes an input into the system’s continuous improvement, not simply a fallback when automation fails. That’s where the value moved. It’s judgment work, and it needs people who understand the business, not just the script. 

The mistake I see most often is treating all this as a staffing ratio. Some share AI, some share humans, and turn the dial as the models improve. That’s an org chart. It isn’t an operating model. 

A real hybrid operation, in my experience, comes down to three questions you have to be able to answer for every interaction type.  

• Who owns the outcome: the AI, the human, or some defined combination?  

• What does the handoff carry, and how do you know when it fails?  

• And what does each tier learn from the other, on a weekly rhythm rather than a quarterly review? 

Get those right, and the ratio takes care of itself. Skip them, and you’ll spend 2027 explaining to your board why the pilot that looked so good never touched a P&L line. 

I don’t think the contact center that pairs humans with AI is a compromise on the way to something purer. The future of CX is an intelligent operating model where AI and human expertise continuously learn from one another, and where the intelligence generated by that system helps the entire operation continuously improve.  

I think it’s the best version of customer operations this industry has ever had a shot at building. I’d rather build it than wait it out.