What does AI do to first contact resolution rates?

AI lifts first contact resolution on the contacts it handles, because it answers in the same session by definition. But cross-channel FCR (did the customer come back about the same issue, anywhere) is the number that matters, and it moves less than the AI-channel number suggests. The AI channel can look excellent while the overall picture is unchanged.

A team reports FCR of 91% on its AI chat channel. Last year, human chat ran at 78%. Headlines write themselves. Eight weeks later, voice volume is up; customers who got an answer from the AI are calling to confirm or to dispute. Voice FCR is down. The AI helped its own slice. The cross-channel number told a different story.

What people in the field are saying

CX Decoded has a piece, "The AI metrics mirage: why your contact...", that names the per-channel-FCR flattery directly: the AI's own number is the easiest to make look good, and the easiest to mistake for the whole picture.

Why is AI-channel FCR naturally high?

Because the AI does not hold conversations open across multiple visits. If the AI answers, the session closes. The customer is either satisfied or they leave (which counts as resolved on the channel's books) or they switch channels (which counts as resolved on the AI's books too). The structure favours the metric.

How does cross-channel FCR catch what AI-FCR misses?

It counts the customer, not the conversation. If the same customer contacts any channel about the same problem within a window (seven days is common), the original contact is marked unresolved. The AI's channel cannot hide a customer who left and called the phone line three hours later.

What can AI legitimately improve about FCR?

The repetitive routine contacts where the customer just wanted the answer. These were already high-FCR on the human channel; AI maintains the rate and absorbs the volume at lower cost. The question is whether the harder contacts (where FCR was always the challenge) are getting better, the same, or worse under AI.

What usually happens to overall FCR under AI?

Mixed. The AI handles its slice well, the residual human team handles a harder mix, the human-channel FCR drops a few points, the cross-channel re-contact rate stays roughly flat or drifts up by one or two points. Headlines pick the channel that won; the honest read is the cross-channel number.

What should you measure?

Cross-channel FCR with the customer as the unit of analysis. The same metric, applied to the same customer base, before and after AI launch. Anything else is comparing the AI's slice to its own old self, which proves nothing.