AI isn’t just taking work out of the agent queue. It’s changing what’s left, and exposing an operating model built for work that is disappearing.
For years, CX leaders have focused on what AI can take on: which contacts can be automated, which journeys can become self-service and where technology can improve efficiency. But as those capabilities move from pilots into the standard operating model, there is a bigger question: What happens if the AI strategy actually works?
Predictable contacts disappear. Proactive service resolves more issues before customers ever need to reach out. What increasingly reaches the agent is the customer whose self-service journey failed, the exception the workflow cannot resolve, the emotionally charged situation, or the problem spanning multiple systems and departments.
The standard answer is that agents will handle the harder contacts. That answer is too small. The agent organization cannot simply become whatever is left after AI has automated the rest.
We’re already seeing signals of what that means in practice.
A team of agents in the Philippines supporting a communications client had been trained on the customer experience surrounding an in-home field technician visit. But most had never lived or worked in the market they supported, so their understanding of that experience was limited to classroom training.
Field technician visits were filmed and documented, critical moments were mapped, and agents experienced the journey through virtual reality. Within days, first-contact resolution increased from 22% to 65%.
The technology made the experience possible, but the lesson wasn’t really about VR. The agents didn’t need more information. They needed context. They needed to understand what happened beyond their part of the customer journey so they could make better decisions within it.
This is a small example of a much bigger shift now underway in CX. Context, judgment, and the ability to take ownership when the expected path breaks becomes more important.
The question is no longer simply how technology can make agents better at the jobs we’ve designed for them. It’s whether we’re designing the right jobs, and the right operating model for what comes next.
That starts with five moves:
- Map the work AI leaves behind. Understand which interactions will still require an agent and how that work is changing.
- Define agent ownership before headcount. Decide what agents should own before determining how many are needed.
- Build the talent profile around the work. Hire and develop for the skills the new workload actually demands.
- Train for where predictability ends. Prepare agents to navigate exceptions, ambiguity and moments that require judgment.
- Measure and invest for the model you’re building. Align KPIs, technology and investment to the future operating model, not the one automation is making obsolete.
Start with what AI leaves behind
Most AI dashboards tell leaders what the technology handled: containment, automation, resolution, adoption and cost. Those numbers matter, but they only tell half the workforce strategy. The other half is what AI consistently hands back to the agents.
When AI resolves predictable interactions, what remains becomes more concentrated around exceptions, ambiguity and failed journeys. That work deserves the same rigor organizations have applied to automation: where does the standard process break, what judgment is required, and which outcomes require human intervention?
The answers can change how performance should be interpreted. If AI disproportionately removes short, predictable contacts, average handle time can rise because the average agent interaction is harder. Escalation patterns can change. Performance variance can widen as judgment matters more than adherence. Some KPIs can appear to deteriorate precisely because the AI strategy is working.
Traditional metrics do not suddenly become obsolete. However, their context has changed. Your automation rate tells you what AI took away. It doesn’t tell you what it left behind. Build a residual-work map alongside the automation roadmap: the exceptions, failed journeys, cross-functional issues, judgment calls and recovery moments that remain. Use it to forecast the skills, capacity and authority the remaining work will require.
The automation roadmap tells you where technology is going. The residual-work map tells you what your workforce needs to become.
Agentic AI raises the stakes further. If ownership is unclear, data is fragmented or the underlying process is broken, giving AI more autonomy won’t fix the operation; it can accelerate its weaknesses. The question isn’t simply how agentic the technology can become, but whether the operation around it is ready for that autonomy.
Redesign ownership before you redesign headcount
Once the remaining work is visible, the natural question is capacity: How many agents will we need? But capacity should not be the first question. Ownership should be: What areas are most important to prioritize agent ownership?
If an AI strategy reduces contacts while the workforce strategy simply reduces headcount, the organization risks automating the old operating model without designing the next one.
Consider the traditional tier structure. Tier 1 handles common issues, while increasingly complicated problems move through specialized levels of support. Moving agents higher up that ladder as AI absorbs Tier 1 work still assumes agent work should be organized around contacts, queues, and escalations.
AI gives us a reason to question the ladder itself. When technology can identify intent, retrieve information, complete actions and resolve predictable needs, agents can be designed around ownership of the outcomes that require judgement.
A contact has a beginning and an end. An outcome can cross channels, systems, departments, and multiple interactions. A contact can be transferred. Ownership stays accountable to the result. Managing a contact asks how efficiently the interaction was handled. Owning an outcome asks whether the customer’s actual need was resolved.
But ownership without authority is just another label. If an agent cannot deviate from a workflow, make a judgment call or coordinate across the functions required to resolve an issue, they do not really own the outcome. They own the next step.
Before redesigning capacity, run an ownership test against the highest-value work AI leaves behind: Who owns the outcome? Do they have the authority to resolve it? Can they cross the organizational boundaries required to do so? Define ownership first, align authority to it, then design capacity around the work.
Build tomorrow’s talent profile from tomorrow’s work
Changing ownership changes what high performance looks like. Contact centers have spent decades creating repeatability: know the process, find the answer, follow the workflow, be consistent, and reduce unnecessary time. We spent years teaching agents to behave more like machines. Now we finally have machines that are exceptionally good at being machines.
When information can be surfaced instantly, knowledge retention becomes less differentiating than knowing how to apply it. When predictable workflows can be automated, speed through a defined process becomes less valuable than knowing what to do when the process no longer fits.
As routine tasks are automated, agent performance shifts towards investigation, judgment, ownership, recovery, and contextual decision-making. That means today’s highest performers will not automatically be tomorrow’s most valuable talent.
The signals are already visible in agents who investigate rather than immediately escalate, recognize when the prescribed answer does not fit, connect information across systems, recover a relationship after something goes wrong, and take ownership when the process stops providing an obvious next step.
Even the language we use needs to catch up. Cortney Jonas Burnos, VP of AI & Digital Solutions at Transcom, has questioned whether “agent” still accurately reflects the role, describing today’s frontline professionals as “high empathy, high skill CX experts working alongside AI.” The language matters because it shapes the roles we design, the capabilities we reward, and the talent we recruit.
Build tomorrow’s talent profile from the emerging work, not yesterday’s top-performer scorecard. Identify the capabilities that the work requires, find the agents already demonstrating them and compare those behaviors with what current recruiting profiles, scorecards, coaching, and career paths reward. The gaps tell you where the talent strategy needs to change.
Train for where predictability ends.
The field technician example illustrates the difference: better performance didn’t come from giving agents more information. It came from giving them the context to apply what they already knew.
That becomes increasingly more important as AI absorbs more of the predictable path. Agents enter where predictability ends. They need experience applying knowledge when information is incomplete, the prescribed answer does not fit, an AI recommendation needs to be challenged or the customer’s problem crosses the boundaries of the workflow.
The lesson is not simply to use more technology in training. It is to change what agents are being prepared for. Train the exception, not just the process. Use simulation to expose agents to incomplete information, failed recommendations, emotional situations, competing priorities, and cross-functional problems before the customer becomes the test.
There is something counterintuitive about that. One of AI’s most valuable roles in the future of CX will be helping humans strengthen the capabilities that matter most when automation reaches its limits.
Measuring the work you are creating, not the work that is disappearing
Changing the role without changing the scorecard leaves agents doing tomorrow’s work while being evaluated against yesterday’s job.
Leaders increasingly need two views of performance. The first measures what AI handles: automation, containment, adoption, resolution, and efficiency. The second measures what agents now own: complex resolution, recovery, judgment, ownership, and customer outcomes. Looking at one without the other creates an incomplete picture.
The investment strategy needs to evolve too. The hidden cost of a successful AI strategy may be everything an organization failed to redesign around it. Funding technology without investing in the data integrity, governance, systems integration, training, role design, and frontline authority it depends on can create an AI-enabled version of yesterday’s operating model.
The workforce plan and the AI plan can no longer be separate exercises. Every successful automation decision changes the work agents receive, which changes the capabilities, authority, training, and measures the organization needs around them.
Customer connection deserves the same precision. Customers do not need an agent inserted into every journey simply to prove one is available. When AI can resolve a need accurately, immediately, and with less effort, that is a better experience. But access to an agent still matters when the situation requires judgment, recovery, trust, meaningful value or risk. Fewer agent interactions make those moments more consequential, not less.
As Cortney Jonas Burnos puts it, “We spent years designing contact centers around managing routine contacts. AI now handles those routine interactions effortlessly. The real opportunity is empowering people with true authority and full ownership over meaningful outcomes.”
For years, organizations have asked what AI can do inside the contact center. The more consequential question for the next three to five years is what the contact center must become because of it.
The organizations that stop at automation will have a more efficient version of the past. The ones that redesign the work around it will build what comes next.
Continue the conversation
The AI conversation is moving fast. The operating model around it needs to move just as deliberately.
Transcom’s AI, but Make It Human series on Leading Voices goes beyond the AI hype to explore what this shift means in practice, from the changing role of frontline professionals to where AI creates real value and where judgment, empathy and expertise still matter most.
Explore AI, but Make It Human on Transcom Leading Voices
Guest post written by Transcom.
