Last week at CCW Vegas I hosted a think tank on the hybrid workforce. I went in expecting the conversation to be about AI capability, accuracy, deployment timelines, the usual technology questions. Most of what I heard was about something else entirely.
The leaders in the room who are making this work are not the ones with the most sophisticated AI. They are the ones who are most deliberate about workforce design.
That distinction matters. It changes how the work gets sequenced and measured, and what gets asked of the people doing it.
Start with the people, not the technology
Snapfinance shared a story that sums up the point. Before they rolled out AI agents, they spent months specializing their human team. They moved people into areas like analytics, into roles that would still need humans after AI took over the simpler interactions.
By the time AI was handling 40% of their call volume, the people whose work was changing already had somewhere to go. The transition was not a layoff event. It was a planned redesign that happened to involve AI.
That is the opposite of how most organizations approach it. The default playbook is to deploy AI first, watch what happens to call volume, then figure out what to do with the people whose work has shifted. By then the conversation has changed from workforce design to workforce reduction, and trust collapses on both sides.
Snapfinance's approach took longer up front. It cost them whatever they would have saved by moving faster. But they did not have a morale problem, an attrition spike, or a pipeline gap when AI went live. They had a team that was ready for the new shape of the work.
AI is making human agents better
The second thing I kept hearing was that AI is changing what good agents look like, in ways that surprised people.
Anthony from Cresta talked about this directly. When AI gives a human agent full context on a customer, the previous interactions, the history, the likely intent, the agent shows up to the conversation with more confidence. They stop fumbling for context and start solving. The interaction gets better not because AI is doing the work, but because the human is now free to do the part of the work that only a human can do.
One attendee runs operations at an organic mattress company. They started using AI for calls and chats in November. Six months later AI is generating real revenue, and the human team is focused on the conversations where judgment and empathy actually move the needle. He spent more time talking about how he trains his people than how he tunes his AI.
That tracks with something I have been thinking about for a while. The companies that win at hybrid will not be the ones who automate the most. They will be the ones whose humans get more capable as the work gets harder.
We are still measuring the wrong things
The metric question came up repeatedly, and it was the most uncomfortable part of the conversation.
Anthony made the point that containment is not the right thing to optimize. You can hang up the phone and improve your containment rate. That does not mean you solved anything. Task success rate, whether the customer actually got what they came for, is the question that matters. Very few organizations are measuring it well.
One attendee said the hardest part of evaluating AI is that you cannot isolate the variable. If your metrics move, the cause might be AI, or a UX change to the website, or a new IVR script, or a product change. When everything is changing at once, attribution becomes guesswork. The teams getting this right are running smaller controlled experiments, not declaring victory on top-line numbers that are influenced by ten things at once.
Snapfinance shared one number that stood out. Their compliance rate went from 87% to 95% after their AI deployment. Compliance is not glamorous, but in financial services it is what determines whether you can operate at all. The metric mattered because it was tied to a specific outcome the business actually cares about.
The lesson is to pick the metric that maps to the outcome you cannot afford to get wrong, and let everything else be secondary.
The handoff is still the hardest part
One of the most honest comments came from an attendee who said the biggest challenge in her organization is not AI capability but the handoff. The question of when AI escalates to a human, and how that handoff feels to the customer, is unresolved in most operations. Right now, in most cases, it does not feel good.
Cresta's example of conversational design got at this. If a customer asks for an agent and the wait time is five minutes, have the AI take down the information in the meantime. If it can solve the problem during the wait, great. If not, the human gets a fully briefed customer instead of a frustrated one starting over.
That is not an AI capability problem. It is a design problem. And it is the kind of detail that separates operations that feel competent from operations that feel chaotic.
Trust takes deliberate work
The last theme was trust, which sounds soft but is actually the most concrete operational issue I heard.
A few specifics that came up. Snapfinance talked about language nuance. The same Spanish word means different things in different countries, and Latino customers tend to type in multiple short messages and hit enter rather than full sentences. Their AI was getting confused until they trained it on the actual conversational pattern of the people it was serving.
Multiple panelists made the point that you need a human in the middle. Not just to handle escalations, but to evaluate the AI continuously. Someone with operational expertise whose job is to ask whether the AI is actually doing what we think it is doing.
And when customers call back, the AI should know the context of their previous interaction. Nothing erodes trust faster than asking a customer to start over.
What this means for how we plan
I left CCW with a clearer view of what the work actually is.
The hybrid workforce is not a technology rollout. It is a workforce redesign that takes months of deliberate planning, gets the measurement framework right, treats the handoff as a first-class design problem, and earns trust through cultural and operational specifics that most playbooks skip.
The companies making it work are the ones treating it that way. They are not waiting for the AI to be perfect. They are doing the harder work of figuring out what their workforce should look like on the other side of this transition, and then building toward it.
That is the conversation I think we should be having. Not whether AI is ready, but whether we are.
