A Worked Example

Meridian Bank is a mid-sized regional bank, forty branches, built over decades on a straightforward franchise: take deposits, make loans, and serve customers who value a local branch. None of that has stopped working. Meridian is not one of the businesses Thorndike would tell a CEO to exit — its lending book is sound, its deposits are sticky, and its competitive position is holding, maybe even improving as smaller rivals struggle. Worth saying plainly, because what follows is not a turnaround story. It’s what reengineering looks like when a company is fundamentally healthy and simply can’t afford to keep doing one particular thing the old way anymore.

Strategic Diagnostic

Nothing at Meridian is in crisis. What put this on the board’s agenda is quieter than that: a handful of competitors have started offering round-the-clock self-service withdrawals, and Meridian’s own cost structure, forty branches’ worth of teller staffing, has started to look heavy next to banks running leaner. The board and CEO’s diagnostic conversation doesn’t produce a dramatic mandate. It produces something more useful: an agreed statement that Meridian competes on relationship banking and lending judgment, not on how a customer gets twenty dollars out of their account, and that the second one deserves to be reengineered precisely because it isn’t where the bank’s advantage lives.

Selecting and Mapping the Process

Withdrawal processing is an easy pick once the diagnostic is done: high headcount, meaningful cost, and no customer has ever chosen Meridian because of how well its tellers count cash. Mapping the current process, deliberately without attaching any employee’s name to any box, produces this:

1. Customer travels to a branch, during business hours only
2. Waits in teller line
3. Presents withdrawal slip / ID to teller
4. Teller manually verifies signature against signature card
5. Teller manually checks account balance (ledger card / branch terminal)Large withdrawal → branch supervisor approval
6. Teller manually counts out cash
7. Teller records transaction in ledger / terminal
8. Teller reconciles cash drawer at end of day
9. Back-office posts transactions to general ledger overnight
10. Updated balance available next business day

Every step in that chain exists because a person once had to be physically present to verify, calculate, or record something. That’s the whole diagnosis in one sentence.

The mapping isn’t entirely friction-free. More than one branch manager, asked to walk through the process, describes it as more complicated than it actually is: approval steps that turn out not to exist, exceptions that turn out to be rare edge cases treated as routine. None of it is exactly a lie. It’s what happens when the person describing a process knows, without anyone saying so, that the process is under review.

Designing an Ideal State

The “what if” questions are almost too easy to ask once the current-state map is on the table. What’s possible now that wasn’t five years ago? Electronic authentication, real-time ledgers, machines that count cash faster and more accurately than any teller. What do customers actually want? Not a relationship with a teller for a withdrawal — convenience, speed, and hours a branch can’t offer. The redesigned process:

1. Customer approaches any ATM, 24/7, many locations
2. Authenticates: card + PIN / chipRare fault/exception → centralized monitoring and support
3. System checks balance in real time
4. Customer requests amount
5. Machine dispenses cash automaticallySchedule → cash-logistics team replenishes machine
6. Transaction posted instantly to central ledger
7. Balance updated in real time

Meridian installs 150 ATMs across its footprint — roughly matching prior withdrawal volume, but with far greater reach than forty branch lobbies ever had.

It’s worth naming what this is a small, contained version of: the same force this book named earlier as the reason it exists at all. ATMs were the electronic-authentication-and-real-time-ledger wave of their moment. Artificial intelligence is doing the equivalent to a much wider set of tasks today, and the questions Meridian’s board asked about a teller window are exactly the questions worth asking now about the work nobody has gotten around to reengineering yet.

The Transformation Path

Meridian doesn’t flip a switch. It pilots ten machines in three branches for two quarters, uses that window to retrain staff and work out fault-handling procedures, then rolls out the rest over the following year, branch by branch, with review checkpoints at each wave rather than one company-wide cutover.

Who the Redesign Actually Touches

This is where the process above turns into people, and where the three questions from the Introduction — what’s no longer needed, what can be automated, what can be picked up by someone else — get answered role by role rather than in the abstract.

Role	Before	After	Disposition
Tellers	120 (3/branch)	40 (1/branch)	55 eliminated — withdrawal processing, their primary task, is fully automated. 25 reassigned into the new Digital Operations team. The 40 who remain handle deposits, account service, and everything a machine still can’t.
Back-office posting staff	25	5	20 eliminated — real-time posting removes the batch-processing task entirely. 5 reassigned into Digital Operations for reconciliation and anomaly monitoring.
Branch supervisors	40	40	Unchanged headcount. Their large-withdrawal approval duty moves to centralized monitoring, freeing roughly a quarter of their time for the staff-management and customer-escalation work that was always the harder part of the job anyway.
Cash logistics	8	10	Grows slightly. More machines to service than branches ever had drawers, and the routes are different, but the function survives essentially intact.
Digital Operations (new)	—	30	Absorbs 25 former tellers and 5 former back-office staff. Real-time transaction monitoring, exception handling, and — for the first two years — helping customers who are still learning to trust a machine with their money.

Total headcount tied to this process moves from 193 to 120: a 38% reduction, net of the one function that actually grows. That number is defensible in a way a flat “20% synergy” assumption never is, because it’s built from five separate, specific decisions, not one company-wide guess.

It’s worth sitting with what didn’t happen here. Nobody’s job was eliminated because a spreadsheet said “cut 38%.” The org chart got thinner because the process underneath it changed first, and thirty of the people whose old task disappeared had somewhere real to go, not because Meridian was generous, but because the redesigned process genuinely needed monitoring and support work it hadn’t needed before. That’s the difference this book keeps insisting on. It’s also, not coincidentally, the harder version of the job. The easier version stops at the 38% and calls it done.