Chapter 11 — The Changing Job Market and New Enterprise Roles

Chapter 11 walkthrough from The AI Contact Center Handbook by Sho Shimoda. Available on Amazon.

The chart on the easel

It is printed on foam-core, mounted on an easel, and leaned against the wall of the sixth-floor executive briefing room of a global insurance company in Zurich. It has been sitting there since the strategy offsite three weeks ago and nobody has taken it down. The chart is from McKinsey's 2024 report on the future of work. The line that everyone keeps looking at is the one marked Customer service representatives. It bends downward starting around 2027 and by 2030 it has lost roughly half its mass.

The CEO is standing in front of it. Around the table are the COO, the CHRO, the CIO, the head of customer experience, and — on video from Manila — the head of the outsource partner that runs the company's 4,200-seat contact center operation. The CEO points at the chart with the end of his pen.

"So the question," he says, "is what we tell the four thousand people whose jobs this line represents. And also what we do about the fact that we haven't hired for any of the jobs the line implies we'll need instead."

There is a long pause.

The head of CX, who has been through three variants of this conversation in the last six months, finally speaks. "The problem," she says, "is that the chart is right about the direction and wrong about the timing. It's not that half those jobs disappear by 2030. It's that half of those jobs change into something else by 2028, and if we're not the ones defining what they change into, someone else will define it for us."

The workforce transition is not a headcount problem. It is a role-definition problem, and the companies that get it right are the ones that start naming the new roles before the old roles finish shrinking.

11.1 The projection: which jobs shrink, which grow

Start with the number that gets quoted most often, and be specific about where it came from. The McKinsey Global Institute's 2024 report modeled task-level automation across roughly 850 occupations. For customer service representatives in developed economies, the model projected that between 40 and 55 percent of the tasks currently performed could be automated by 2030, midpoint around 50 percent. Task-level automation of 50 percent does not translate one-to-one into headcount reduction of 50 percent — tasks recombine, new tasks emerge — but it does translate into significant reshaping.

Deloitte's 2024 Global Contact Center Survey put its own version at roughly 30–40 percent net reduction in traditional agent headcount by 2030 among enterprises that have completed an AI-first transformation. Gartner's 2025 forecast is more aggressive on automation and more conservative on headcount: they expect 80 percent of customer service interactions to have some AI involvement by 2029, but they expect the human workforce to shrink by only about 25 percent because AI creates demand for interactions that previously would not have happened at all.

The three numbers — 50 percent, 30–40 percent, and 25 percent — sound like they contradict each other, but they measure different things. McKinsey measures task-level automation potential. Deloitte measures observed headcount in transformed enterprises. Gartner measures net industry employment. All three point in the same direction.

Roles that shrink dramatically: Tier 1 agents handling routine, script-followable inquiries. By 2030 in most industries this role will exist at perhaps 30–40 percent of its 2024 headcount, with remaining humans handling exceptions and cases where the AI has failed or the customer has explicitly requested a human. Roles that shrink moderately: traditional QA analysts. The QA function does not disappear — it shifts from evaluation to calibration. Roles that stay roughly flat: team leads and supervisors. The span-of-control math changes, but coaching and coordination do not automate. Roles that grow: everything on the AI side of the stack.

11.2 The new roles being born

Some of the new roles have crisp job descriptions. Others are still being invented in real time. Here are the ones that have hardened enough by 2026 that you can actually hire for them.

AI Agent Builder

The person who designs, configures, tests, and deploys the automated agents that handle customer conversations. In practical terms, this means working inside a platform like Amazon Connect's agent builder, Salesforce's Agentforce Studio, or Google's Conversational Agents Console — dragging intents together, wiring up backend actions, connecting to CRM data, and testing the resulting flows against a library of sample conversations.

In plain English: An AI Agent Builder is essentially a conversation-flow engineer. On any given day they might sketch a decision tree for how the AI should handle a refund request, wire the AI up to pull real account balances from the CRM, and then run two dozen mock conversations to see where it breaks — the way a QA engineer would stress-test any other piece of production software.

The role sits at the intersection of three older disciplines: chatbot designer (from the 2015–2020 era), conversation designer (from the 2018–2023 UX-writing era), and RPA developer (from the process automation era). The best AI Agent Builders in 2026 come from three feeder pools: senior support engineers who understand the CRM and the backend; former UX designers who understand conversation flow; and — in an increasing share of hires — former Tier 2 agents who spent five years learning exactly which customer requests are hard and which are easy. That third group has a specific advantage: they know, from lived experience, where the exceptions live.

CX Optimization Specialist

Once you have automated agents running, someone has to watch how they perform, tune them, and continuously improve them. The role is analytical: reading transcripts, looking at containment rates, measuring customer satisfaction scores by conversation type, running A/B tests on prompt variations, and reporting on which model versions are improving and which are regressing. If the AI Agent Builder is the person who ships the model to production, the CX Optimization Specialist owns it once it is there.

AI Security Administrator

New attack surface, new specialty. Monitors the contact center's AI systems for prompt injection attacks, model jailbreaks, deepfake voice attacks, and adversarial attacks that target vector databases and retrieval systems. Also owns the Guardian Agent stack from Chapter 10 — configuring it, tuning thresholds, investigating flagged events. The role did not exist in 2023. By 2026 it exists in every enterprise contact center that has deployed generative AI at scale.

Conversational Designer

Writes the AI's voice — literally. The system prompt, the fallback responses, the tone-of-voice guidelines, the specific phrases the AI uses to greet, apologize, escalate, and close. This is the role most often underestimated by companies making their first agentic deployment, which is why so many first-generation AI agents sound like they were written by an insurance policy. Good conversational designers come from copywriting, UX writing, and — increasingly — playwriting and screenwriting backgrounds.

Prompt Engineer for CX

Related to but distinct from the Conversational Designer. Writes and maintains the technical prompts that shape the AI's behavior across many conversations — the system prompt that defines the AI's role and constraints, the few-shot examples that teach it how to handle specific situations, the retrieval prompts that pull the right knowledge base article. The Prompt Engineer thinks about the model as a system to be tuned; the Conversational Designer thinks about the model as a character to be voiced. Different specialties.

Guardian Agent Operator

Monitors the Guardian Agent stack in real time. When the Guardian flags a suspected deepfake, a suspicious refund authorization, or an AI response that violated a compliance rule, the operator looks at the flag, decides whether it was true or false positive, and either releases the block or confirms it. Over time the operator's decisions become training data for the Guardian itself. This role frequently emerges out of existing QA and fraud teams — one of the most direct transition paths from a shrinking role to a growing one.

11.3 How today's agents can transition

The headline projections make the workforce transition sound abstract, as if 40 percent of people in the industry will simply disappear. What happens in practice is smaller, messier, and much more human — people move sideways, up, or out one at a time, and the pattern of which direction they move depends heavily on whether their employer bothers to build a bridge for them.

The transitions share a pattern. In each case, the person's existing knowledge — of customers, of edge cases, of what good service actually looks like — turned out to be the scarce and valuable input, not the specific tasks their old role required. The technical skill (how to configure the AI, how to read the analytics dashboard, how to tune a Guardian threshold) was teachable in weeks to months. The judgment was not.

The implication for anyone leading a contact center: the workforce you already have is the highest-signal talent pool for the new roles. Losing them to a layoff and then hiring fresh externally is almost always more expensive, slower, and produces worse outcomes than reskilling. The math favors internal transition in something like 80 percent of cases. But it only works if the reskilling program actually exists.

One more transition worth documenting because it happens quietly and often. The team lead who became a full-time trainer for the AI. In several organizations the operations leader most respected by peers — the one everyone consults when a call goes sideways — has been formally moved into a role whose sole purpose is to design the training conversations that shape how the AI handles hard cases. The role has different names at different companies (AI Curriculum Lead, Model Trainer, Conversation Architect). The pattern is the same: take the person whose judgment the frontline already trusts, and use them to encode that judgment into the AI. The result is an AI that behaves the way your best agent would behave, at scale.

11.4 What HR needs to do now

Every conversation about workforce transformation eventually lands on HR, and HR is often the least prepared function in the enterprise. Not for lack of goodwill — the specific muscles required (skills mapping, individualized development, cross-role bridging) are not the muscles most HR functions have historically built.

One. Map the roles. For every role in the current contact center — agent, senior agent, team lead, QA analyst, workforce planner, trainer, quality manager — list the specific skills, tools, and experience. Map the same for the target roles. The gap between the two maps is the reskilling plan. Most HR functions have never done this map at the level of granularity that would let them build individualized transition paths.

Two. Rebuild the learning function. The old model — three-week initial cohorts in a room with folding tables — does not survive contact with the new roles. Reskilling into AI Agent Builder is roughly three to six months of part-time learning, and it needs hands-on time with the actual AI tooling. Companies that get it right partner with vendors like Salesforce Trailhead, Google Cloud Skills Boost, AWS Skill Builder, or focused CX-training providers like ICMI. Companies that get it wrong send people to generic online courses and wonder why nothing changes.

Three. Hire for the new roles now. Even if the transition plan is to reskill internally, some new roles need external senior hires immediately to build the muscle and mentor internal transitions. Waiting until 2028 to hire an AI CX leader is choosing to be two years behind competitors who hired one in 2025.

Four. Change the incentive system. The old bonus structures — average handle time, calls per hour, adherence to schedule — actively punish the behaviors the new workforce needs. If the human agent is now handling only the hardest calls, penalizing them for longer handle time is exactly wrong. If a senior agent spends 20 percent of her week helping train the AI on edge cases, the incentive plan needs to reward that time explicitly.

Five. Be honest about who will not transition. Some fraction of the existing workforce will not make the jump. This is not a moral failing on their part or the company's — it is a normal distribution across a large population. The better approach is a transparent tiered program: aggressive investment in the people who show aptitude and interest, generous severance and outplacement for the people for whom transition is not the right path, and open communication about which is which.

Six. Talk to the workforce. Most companies communicate about workforce transformation in ways designed to minimize legal risk rather than maximize trust. The result is a workforce that hears rumors, believes the worst, and disengages. Companies that get this right commit publicly to reskilling investment, publish the target composition of the workforce, and hold quarterly all-hands meetings where the CEO takes questions. It is uncomfortable. It also works.

11.5 The industry-level implications

Zoom out one more level. The contact center industry, globally, employs somewhere between 12 and 17 million people, depending on which count you trust and how you define the boundary. It is concentrated in specific economies — the Philippines has about 1.4 million BPO workers, India has close to 1.2 million, the US has around 3 million, the UK and Ireland together have around 850,000, and a growing share sits in Latin America (Colombia, Mexico, and Costa Rica are the fastest-growing hubs). A workforce transition that reshapes 30–50 percent of those roles in the next five years is not just a corporate HR issue. It is an industrial-policy issue in at least a dozen countries.

The one-line takeaway: When a single contact-center-adjacent sector represents 8 percent of a country's GDP, its exposure to AI automation is not a corporate matter — it is a national one, and the BPO economies are the most exposed of all.

One more industry-level observation. The traditional career ladder in a contact center — agent to senior agent to team lead to supervisor to manager to director — was long, slow, and stable. Ten to fifteen years from the headset to the C-suite adjacent role. That ladder is now bending. The new roles that pay the most and command the most influence are not further up the traditional ladder — they are on a parallel track that did not exist five years ago. An AI Agent Builder with three years of experience can now make more than a team lead with fifteen, and the promotion path from the builder role runs into product management, AI operations, and eventually a Chief AI Officer role that reports directly to the CEO. The consequences for how a company designs its early-career development pipeline are significant. The consequences for how it retains talent are more significant still.

The forecast for the industry, then, is not AI takes 50 percent of the jobs. It is: over the next five years, the composition of the industry's workforce changes substantially; the countries and companies that invest in the transition end up on the growing side; the countries and companies that do not end up on the shrinking side; and the pace at which the transition happens depends much more on the pace of enterprise reskilling programs than on the pace of model capability improvements. The technology is running ahead of the humans. Whether the humans catch up is a policy choice at every level from the individual agent to the multinational government.

What to take with you from Chapter 11

The workforce projection everyone quotes — roughly half of contact center roles reshaped by 2030 — is directionally right, definitionally fuzzy, and less useful than the more granular question: which specific roles shrink, which grow, and how does an existing employee move from the first list to the second.

The new roles are already real. AI Agent Builder, CX Optimization Specialist, Conversational Designer, Prompt Engineer for CX, AI Security Administrator, Guardian Agent Operator — every one appears on job boards in 2026 with real salaries and real hiring pipelines. The scarce skill in all of them is not the technical configuration work, which is teachable in weeks to months. The scarce skill is the customer judgment that a tenured agent, a veteran QA analyst, or an experienced team lead has spent years developing. That judgment is the reason internal transitions almost always beat external hiring for the growing roles.

HR functions that have not started the skills-mapping, curriculum-building, and incentive-redesign work in 2026 are going to spend 2027 and 2028 losing internal talent to competitors who did start. And industries and governments in the BPO economies are running their own version of the same race — with retraining investment on one side, protective policy on the other, and worker representation increasingly in the middle.

The next chapter is about the other half of the transition: the vendor decision. If Chapter 11 was about the people who will run the new contact center, Chapter 12 is about the platform they will run it on. Between them they define the shape of the operation five years from now.