Part 2 overview from The AI Contact Center Handbook by Sho Shimoda. Available on Amazon.
Why Part 2 exists
Part 1 was the map. Part 2 is the ground. When the map says "agentic AI is going to handle the routine work," Part 2 is where you find out what actually changes on the ground — for the human on the headset and for the technology stack that has to make decisions in the seconds a customer is waiting.
The two chapters attack the same shift from opposite sides. Chapter 4 starts with the person: Priya Menon at 5:30 PM in Bangalore, taking three calls in an hour instead of thirty, and having a completely different day because of it. Chapter 5 starts with the machine: a bank at 11:47 PM on a Tuesday, deciding in three seconds whether the caller reaching for Maria Alvarez's password is Maria — or a fraudster who swapped her SIM card three hours earlier. Same shift. Two very different things reorganizing around it.
The two chapters, in one paragraph each
The widely-cited Gartner projection — 80% of routine issues resolved autonomously by 2029 — is not a headcount cut. It is a rebalancing. AI takes the low-complexity, high-volume work (password resets, order status, balance inquiries). What reaches a human is a different mix: the emotionally complex, the policy edge cases, the escalations where an apology from a person is the actual resolution. Chapter 4 walks through what that shift does to average handle time (up, and that's good), what the "in-ear copilot" is and why it works (it never talks to the customer), the four skills that suddenly matter (judgment, systems thinking, prioritization, emotional labor), and the honest headcount conversation — including what it means for the Philippines BPO industry.
Read the Chapter 4 walkthrough →The Customer Data Platform was built for a use case where the answer can be a day old. The customer service moment can't wait. Chapter 5 introduces the XOP — the "air traffic control" layer that reads every live signal (SIM swap events, sentiment scores, journey state, fraud flags), fuses them with context from every system of record, and makes a routing decision in 200-500 milliseconds. It walks through the three components (event bus, decision engine, orchestration layer), the CAMARA APIs that finally let enterprises query the telecom network as a first-class data source, and the vendor scrum currently reshaping the CCaaS + CRM boundary.
Read the Chapter 5 walkthrough →Two vignettes that anchor the two chapters
| The scene | What it teaches |
|---|---|
| 5:30 PM, Bangalore Priya Menon takes three calls in her first hour — a widow closing her husband's account, a $340 roaming dispute, and a customer who explicitly asked to speak to a person |
These are "escalations by design" — the routine calls never reach Priya at all. The widow thanks her at the end of the call, which is a kind of gratitude Priya rarely got in her old job, because in her old job she was too far downstream in a chain of failures to be seen as the person who fixed anything. |
| 11:47 PM, Phoenix bank A call arrives for Maria Alvarez's password reset. Voice pattern-matches at 91%. But her SIM was swapped 3 hours ago at a T-Mobile store the fraud system has flagged twice this month |
Four systems have to be consulted (carrier SIM-swap API, auth history, CRM, fraud analytics) — in three seconds. Only one kind of system can fuse them fast enough: an XOP. Route to human fraud specialist? Route to routine reset queue? Hang up and alert the real Maria on her registered backup channel? The answer needs to be current to the second. |
Figure 2 — Two vignettes, one shift. The human is doing higher-value work. The platform is doing higher-speed reasoning. Neither works without the other.
What Part 2 buys you
| If you are… | Part 2 gives you… |
|---|---|
| A CX or contact center operations leader planning the workforce | A clear-eyed read on which roles shrink, which roles grow, and what the new skill profile looks like. Plus the honest cost/compensation math nobody puts on the earnings-call slide. |
| A CIO or CTO defining the platform strategy | The architectural vocabulary — event bus, decision engine, orchestration layer, CAMARA — and a defensible position on where the XOP sits relative to your existing CCaaS, CRM, and CDP investments. |
| A frontline manager or workforce coach | A concrete picture of what the "in-ear copilot" is, which vendors ship it, and what the coaching and QA process should look like when 40% of an agent's real-time suggestions come from a machine. |
| A vendor evaluator sitting through the CCaaS + CRM pitch cycle | A framework for cutting through the "we do XOP too" claims that every incumbent is starting to make. The question to ask is whether they built a decision engine or bolted a streaming pipeline onto a batch database. |
What comes next
Part 3 — Quality & Measurement — takes the two front-line elements from Part 2 (the newly-empowered human and the newly-orchestrated platform) and asks the harder second-order question: how do you know it's working? Every AI-mediated conversation creates a new set of QA problems that the legacy scorecard was never built to catch. Part 3 is about rebuilding the measurement layer for a world where the agent doing the talking is not the human you're evaluating.
Continue to Part 3 — Quality & Measurement →