Commonwealth Bank of Australia cut customer service roles after putting an AI voice bot on its service line, and then asked those people to come back. ABC News reported in August 2025 that the bank restored 45 customer service roles that had been replaced by an AI voice bot. Call volumes went up rather than down, the bank reversed the decision, offered the affected staff the chance to return, and apologised for what it called a misjudgment of staffing requirements.
This is the most useful kind of case, because nothing broke in the technology. Union details showed what followed the go-live: call volumes surged instead of dropping, management added overtime shifts and pulled team leads onto the phones. Chapter 13 of the book names the mechanism. MIT found that despite thirty to forty billion dollars spent, roughly 95 percent of pilots had no effect on profit, and the cause was the learning gap rather than the technology.
What the bank actually bought
A voice bot that answers correctly is still only one of six blocks. The other five carry the result: strategy, data, way of working, governance, and culture and skills. They multiply instead of add, so a single zero makes the whole thing zero.
At the bank, the way of working was the zero. The bot picked up calls, but the work around it stayed as it was, with no redesigned escalation path and no measure of whether the customer's issue was actually closed. So the demand returned through a second call, and the team leads absorbed it.
- Strategy: cost reduction was set before the demand was understood
- Data: deflection was counted, resolution was not
- Way of working: the handover to a human was never designed
- Technology: this block worked
A bot that ends a conversation without ending the problem creates two conversations.
Service pays back fastest, and that is where people stop thinking
Deloitte surveyed more than 1,800 executives and found a payback period of two to four years, three to four times longer than conventional IT. Only about 6 percent reported a return within a year. Customer service is the exception in that study, with around 63 percent landing inside the first year, which is exactly why so many boards start there and treat the number as a given.
It is not a given. Only about 28 percent of use cases fully meet their ROI expectation. The fast payback in service comes from volume that is genuinely repetitive and from a clean handover for everything else. Remove the handover and the volume comes back, with overtime attached.
What the customer feels first
Every AI implementation is a wall or a door, and the customer feels the difference within four messages. A door recognises the question, does the part it can do, and passes the rest to a named person with the context intact. A wall asks the question again.
Your customer does not know which block in your organisation stood at zero. He knows he called twice. Gartner predicted in June 2025 that 50 percent of companies would abandon plans to replace customer service roles with AI by 2027. That correction is the market paying the bill for walls built last year.
The agent is the second reader of everything you make, and the customer is still the first.
What you do on Monday
- Pick your highest-volume service journey and measure resolution instead of deflection, over a two-week window
- Write down who the human owner is at the moment the AI hands over, by name and by shift
- Score the six blocks for that one journey, and fix the lowest one before you widen the scope
- Hold the business case to the Deloitte horizon of two to four years, and treat anything faster as the exception it is
- Put the repeat-contact rate on the same dashboard as the cost per contact, in front of the same person
Teams that run this sequence see roughly four times faster time-to-market and about 68 percent more cost efficiency with the CX+ Method, because the climb happens in the order the organisation can absorb.
You now have a journey, a measure and an owner. The next step is the operating rhythm around them, and that is what CX+ Way of Working × AI lays out: the handover points, the decision rights and the weekly loop that keeps a rollout from becoming a rehiring round. Bekijk de materialen