How do you keep users from abandoning a slow AI task?
by•
One workflow in my app can take about a minute on the first run because it reads several pages before producing a report. The work is happening, but a spinner makes it feel frozen.
I added clear progress stages and cached repeat runs. That helped, but the first run is still slow.
If your product has a long-running AI or research task, what helped users feel comfortable waiting? Did clearer progress make a difference, or did you have to redesign the task?
37 views
Replies
Could you show the pages found as they come in, then let people leave and return to the finished report? A clear estimate helps too, as long as it reflects the actual wait.
A song in my app takes about 75 seconds, so I live with the same problem. What helped, in order. Make leaving safe: the run is one record on the server, so closing the screen loses nothing, the library row flips when it finishes, and a push says it is ready. The push permission is offered right as the wait starts, which is when it makes the most sense. Honest progress: a bar that eases toward about 92 percent over the typical time and only snaps full when the result really arrives, so it never sits at 100 waiting. Something to do: a small tap game on the progress screen, so the minute is a pastime rather than a spinner. And the payoff: when the song is ready it starts playing by itself, because the data showed many people never pressed play after waiting.