Customer service has fewer new ideas than the market suggests, so the same ideas return every 7 to 10 years with new software attached. Each wave promises a lower cost per contact. Companies adopt it further than customers will tolerate. Reaching a person turns into a selling point, the selling point turns into a cost, and the next wave arrives.
Service forecasters should watch three lines at once. One is how easy it is to reach a person, another is where the people who answer are located, and the third is which metric the budget is argued in. The lines move against each other, and most of their turning points are public admissions by the companies that pushed a wave hardest.
Every automation wave has ended the same way so far. Deflection outruns what the technology can resolve, volume flows back to the human queue, and the company that announced the savings announces the rehiring. The mass tier then automates again with better tools, and human service moves up the price list, much as handmade goods became luxury once factories took over the rest.
The dashed segments are forecasts taken from analyst predictions published in 2026. They point to an odd reversal, in which the offshore human agent returns as the cheaper option once AI vendors stop subsidising inference.