The CX-trend predictions that held up (and the ones that didn't)
A grade-the-homework look at the 2024-2025 CX-trend forecasts from Zendesk, Gartner, and Forrester: which held up, which missed, and what the pattern says about reading next year's.
Open any customer-support strategy deck from late 2023 and you will find the same confident slide: a hockey-stick curve, a round number, and a near-term date. AI would resolve most tickets. Chatbots would become the front door. Headcount would fall. Two years on, the homework can be graded — and the honest scorecard is less flattering than the forecasts were.
What held up
Start with the win, because the analysts earned one. The safest call of the cycle was that generative AI would go from novelty to default, and fast.
It happened. Every major help desk shipped an AI agent, copilots landed in agent consoles, and "do you have an AI option" became a standard line in RFPs. If the prediction was "AI will be everywhere," it held comfortably. The trouble is that "everywhere" was never the interesting claim.
What missed
The interesting claim was about outcomes, and that is where the decks aged badly. Zendesk's 2025 CX Trends report put a number on the ambition.
"A few years" is now roughly here, and reality is nowhere close. Gartner's own field data found that
— no agent, no callback. The gap between "touched by AI" and "actually finished by AI" is enormous, and it is where most deflection theater lives. We pulled that thread in what a 70% resolution rate actually means; the short version is that vendor-reported autonomous resolution and customer-verified resolution are different animals.
Forrester made the boldest optimistic call and missed it cleanly. Its Predictions 2024 report forecast that the global average customer experience would improve for the first time in three years. Instead,
Then US scores fell again in 2025, a fourth consecutive year. One cause Forrester itself named was "disappointing implementations of potentially game-changing technology," AI included. The forecast and the postmortem came from the same building.
The predictions that held up were directional and dull. The ones that missed came with a round number and a date you could drop into a budget.
The reversals
The most telling corrections were the analysts revising themselves. The 2023-2024 subtext — that AI would let you cut the team — quietly inverted. In June 2025 Gartner predicted that
, and that half of the firms that did cut would rehire under different titles. Only about a fifth of leaders had actually reduced staffing because of AI in the first place. Weeks earlier, Gartner had hedged its own headline agentic-AI forecast, warning that
When the same firm predicts 80% autonomous resolution by 2029 and 40% project failure by 2027, it is quietly telling you the ceiling is aspirational. The staffing story is worth reading in full in the org chart after AI.
Jury still out
Two calls remain genuinely open: whether autonomous resolution reaches its promised ceiling on any real timeline, and whether chatbots become anyone's primary channel rather than their most-complained-about one. Both are possible; neither is settled. And "agent washing" — Gartner counted only about 130 real agentic vendors among the thousands claiming the label — makes the honest number hard to see through the marketing.
How to read the next batch
Notice the shape of the scorecard. The calls that survived contact with reality were vague and directional ("AI adoption will surge"). The calls that shattered were specific, optimistic, and dated ("80% resolved, headcount down, by a quarter you can put in a budget"). A forecast with a round number and a near date is usually a vendor's revenue target wearing an analyst's font. Read them for direction, discount the dates, and run your own pilot without lying to yourself rather than someone else's slide. The safest bet in CX forecasting, it turns out, was pessimism.