Customer experience was, until recently, a cost centre that occasionally answered the telephone. Tickets closed, handle time, contacts avoided: the vocabulary of a liability being managed downward. Pact and MaryRuth’s have moved the line item quietly, and the move is the story. A Gladly-sponsored case study on Modern Retail is the evidence.

Lauren Inman-Semerau at Pact launched an AI chat in October 2025 that now resolves 57% of conversations and converts chat-driven journeys at 17%. The standard e-commerce benchmark is two to five. The discipline is not the AI; it is the refusal to ask the AI to deflect.

MaryRuth’s consolidated four platforms into a single customer thread and watched efficiency rise 35% in two months — mechanical, expected, uninteresting. The interesting number came next. Jim Rodden’s team now saves one in five cancellation conversations and generates three times the cost-per-contact revenue. The mechanism is triage, not retention: diagnose the complaint, then prescribe a different product or cadence.

This is the pattern every dashboard missed. Volume was never the variable worth watching; lifetime value was, and nobody could see it.

AI did not rescue CX from the cost bucket. It cleared the queue, which freed the data that had always named CX a revenue function and had been ignored because no one had time to read it. Rodden’s team fed internal QA, customer feedback, and human evaluation back into the model, and in doing so pointed a listening instrument at the product catalogue.

The finance team now has a reason to care.

CX budgets built on cost per contact and average handle time optimise for the conversation ending. Pact and MaryRuth’s optimise for the conversation mattering — and have attribution through to the next order to prove it. One is the cheaper number. The other is the larger one.

This is why the “AI will gut support” thesis has read wrong from the start. The jobs were never the point; the view was. The backlog had always obscured the cohort, and the C-suite returned the favour by budgeting CX as overhead. Clear the queue and the cohort appears. Once it appears, the CFO begins asking different questions.

The technology came second. What matters, what has always mattered, is which office CX reports to. A team placed under a growth officer builds a growth function. Place the same team under the COO and it is a queue again inside two years. Org charts are policy documents wearing a uniform; the metrics at Pact and MaryRuth’s suggest the chart underneath has changed.