Chapter 1 walkthrough from The AI Contact Center Handbook by Sho Shimoda. Available on Amazon.
The 7:14 a.m. ritual
At 7:14 on a Wednesday morning, Marcus Delaney sits down at Workstation 34B in the third-largest utility company in Ohio, adjusts the headset that leaves a small red mark on his ear by 5 PM every day, and begins the ritual of opening his applications. There are eight of them.
At 7:22, before any of that finishes loading, the first call arrives. The caller is a woman named Debra whose gas has been off for three days. Marcus asks her to hold for a moment while he pulls up her account. The CRM is still loading. He can hear her breathing on the line.
The legacy contact center is not a technology problem. It is an accumulated tax that the business pays every day, in cash and in trust, without ever seeing the invoice.
What "legacy" actually means (in plain English)
The word "legacy" gets thrown around loosely. In most industries it just means "old." In the contact center world it has a more precise meaning. A legacy contact center is a system in which the routing of an interaction — who takes the call, from what queue — is decided by a piece of infrastructure that lives, or once lived, on your physical premises. That infrastructure is called an ACD, or Automatic Call Distributor. In the classic setup, you own that hardware. It was sold to you by Avaya, Cisco, Nortel, Mitel, or Siemens. You paid for it up front. You depreciate it. If you want to add a hundred seats for the holiday season, you call the vendor, wait six weeks, and pay per port.
But the phone system is only the innermost layer. Wrapped around it is the CRM. Wrapped around the CRM is the workforce management platform. Wrapped around that is the QA function. Wrapped around that is the training program. Wrapped around that is the real estate footprint. Every layer is defensible on its own. Together they are the operating model — and the thing that makes it "legacy" is not any single piece, but the way the pieces reinforce each other.
In plain English: An ACD is the piece of software that decides which agent's phone rings. In a modern cloud system it's a service running on a server farm somewhere. In a legacy system it is a physical box in a room in your building, humming quietly, keeping the whole operation alive.
The hidden tax on the agent
The industry has a term for what Marcus is doing. It is called toggling. The 2023 Aberdeen Strategy Research survey put the average number of applications a contact center agent uses per interaction at 6.4, with the top decile at 11. CCW's benchmark says agents spend as much as 22% of their handle time not searching for information, but navigating between the systems that hold it. That is roughly one full working day per week, per agent, spent on nothing except moving a mouse from one window to another.
| Metric | Legacy contact center reality | Source |
|---|---|---|
| Apps per interaction (avg) | 6.4 — top decile hits 11 | Aberdeen 2023 |
| Handle time spent navigating between apps | Up to 22% — a full day per week | CCW benchmark |
| Annual agent turnover (US) | 30–45% — outsourcers 60%+ | Industry benchmarks |
| All-in cost to replace one agent | $10K – $20K (higher in regulated verticals) | Recruiting + training + ramp |
| 500-seat center, 40% turnover | $2M – $4M / year in replacement cost | Derived |
There is a wonderful piece of accidental honesty in the industry. Almost every job posting for a contact center agent contains some version of the phrase "must be able to multitask." That is a marker. That is the industry admitting, in the recruitment ad, that the fundamental design of the work is broken. Nobody puts "must be able to multitask" on a job posting for a well-designed job. What the industry is really saying is: "our systems will make your day impossible, and we are hiring for the personality type that can tolerate that."
The one-line takeaway: Agent turnover in a legacy center is not a workforce problem — it is an operating-model tax paid in human capital.
The hidden tax on the customer
The customer's version of the tax is different, but it comes out of the same pocket. Debra, still on the line, has already called twice before. Marcus is about to ask her to re-explain her situation. He knows he is about to do it. Debra knows he is about to do it. This is the point in a call where a certain kind of customer starts crying, and a certain other kind of customer starts yelling — both reactions are the same information: the system just failed them again.
| The tax | What actually happens | Industry benchmark |
|---|---|---|
| 1. Context loss on transfer | The receiving agent sees the account number and a two-line note. Everything else — the tone, the history, the previous attempts — starts over. | ~60% "explain again" rate on cross-queue transfers |
| 2. Dropped calls on handoff | Legacy switch drops the call during transfer. Customer calls back, gets a new agent, starts over. The system logs it as a normal disconnect. | 3–8% of transfers |
| 3. Abandonment at the queue | Bad routing, misdirecting IVR, batch-scheduled WFM. Customer hangs up before an agent answers. | Abandonment climbs sharply past 90 seconds; legacy peaks run 4–8 min |
| 4. The 5-level IVR | Every business unit added its own branch. Patience drops sharply after the third level. Customers just press zero. | DTMF research on menu depth |
| 5. The callback loop | Incomplete answer → customer calls again next day. In ops reports it's two contacts. In the customer's experience it's one problem that took twice as long. | FCR averages ~70%; legacy worse |
The three obstacles to modernizing your CRM
Suppose you decide, sensibly, that the way out of the trap starts with the CRM. You will run into three obstacles, in this order, and it is worth knowing them by name before you start.
| Obstacle | What it looks like | When it kills the project |
|---|---|---|
| 1. Data quality | 15 years of duplicates, wrong addresses, phone numbers in six formats, open cases from 2018 nobody ever closed. The migration ingests the mess unchanged. | During the technical migration — because "we'll clean up after go-live" almost never happens. |
| 2. Integration lock-in | A spider web of nightly file drops, database triggers, and screen-scraping bots. Every custom field was built by someone who left in 2018 and didn't document why. | At go-live — the vendor's connectors handle 80% and the last mile breaks. |
| 3. Human resistance | Ten years of muscle memory and undocumented workarounds. Tenured agents can't unlearn in three weeks of training. | Three months after go-live — when veterans quietly quit and productivity dips. |
Projects that budget only for the first obstacle — that scope the effort as "a CRM migration" and hire a systems integrator to run the data conversion — routinely spend eighteen months and their entire budget solving obstacle 1, and then run into obstacles 2 and 3 with no money and no political capital left.
The boardroom moment
It usually happens in a Q3 review. The CX director presents the quarterly numbers — NPS down four points, abandonment up 12%, first-call resolution flat, cost per contact up 8% on flat volume. The CFO listens, takes off his reading glasses, and asks the question nobody in the room has an answer to.
"And what does it cost us to not do that?"
That is the moment. That is when the trap becomes visible as a trap and not as a set of individual, tolerable annoyances. Every contact center leader in the industry has, in some form, had that moment. What matters is not the number — the number is unreliable and everyone in the room knows it. What matters is the reframing. The cost of change becomes a specific dollar figure. The cost of not changing becomes a growing, cumulative liability with no bottom in sight. Once that reframing lands, the argument changes shape. It is no longer "should we modernize?" It is "how fast can we do this without breaking things?"
What to take with you from Chapter 1
A legacy contact center is not a legacy technology. It is a legacy operating model held in place by three reinforcing structures — the on-premises phone infrastructure, the pre-cloud CRM, and the workforce that has adapted to both. Each is defensible on its own. The trap is the way they hold each other up. The costs are real, they are large, and they hide inside metrics that nobody reports to the board. Agent turnover eats millions of dollars a year. Customer effort moves the loyalty numbers that eventually move the revenue numbers. The cost of modernizing the CRM is not primarily a software cost — it is a data cost, an integration cost, and a human cost, in that order.
Chapter 2 is about the architecture that replaces the trap. It introduces three acronyms — UCaaS, CCaaS, and CPaaS — and argues that the interesting story is the way they are converging. If Chapter 1 named the trap, Chapter 2 opens the door.