Learning progress and AI Leads results stay put

Losing progress you already made is one of the more frustrating bugs a product can have. This batch fixes three ways that was happening.

What changed

Reloading the page during a Learning session or while viewing AI Leads results could reset progress you had already made, whether that was your position in a learning flow or leads you had already reviewed. A page reload should never lose work, but in some cases it did, which meant something as ordinary as an accidental refresh, a browser tab crash, or simply navigating away and back could undo real effort you had already put in.

Separately, swiping through AI Leads (approving or rejecting a suggested lead) could result in a duplicate charge for the same swipe. If a swipe registered more than once due to a timing issue, you could be charged twice for what should have been a single action. This is a billing integrity issue as much as a usability one: the number of swipes you made and the number you were charged for should always match exactly, and in some cases they did not.

And when a lead-finding run got stuck internally, it would previously retry indefinitely without ever resolving, rather than either recovering on its own or stopping and flagging itself for you to review. A run that keeps retrying forever without making progress just burns time without telling you anything is wrong, and from your side it can be hard to tell the difference between a run that is legitimately taking a while and one that has actually gotten stuck and will never finish on its own.

All three are fixed. Reloading no longer resets Learning progress or AI Leads results. Swipes are charged once, matching exactly the number of swipe actions you actually took. And stuck runs now either recover automatically or pause so you can see and address them, instead of looping forever with no visible end.

These three issues, though separate in cause, share a common thread: each one was a place where the state you saw on screen or the charge on your account did not reliably match the work you had actually done. Fixing them means Learning and AI Leads are now consistent with your actual activity in all three of these respects.

How to use it

No changes to how you interact with Learning or AI Leads. Use both the way you always have; the fixes are in reliability, not workflow, so there is no new setting to enable or new step to remember.

If a run does get stuck going forward, you should now see it paused for review rather than silently spinning, so check your AI Leads run status if a batch seems to be taking unusually long. A paused run gives you a clear signal and a clear next action, rather than leaving you to guess whether it is still working.

If you have historical concerns about a duplicate charge from before this fix, your billing history in account settings shows your swipe-related charges, and support can help reconcile any specific case you are unsure about.

If you had previously avoided reloading mid-session because you were worried about losing progress, that caution is no longer necessary. You can navigate away and back, refresh, or switch tabs during a Learning session or while reviewing AI Leads results without your progress resetting.

Why it matters

Progress that disappears on reload makes a feature feel unreliable even when the underlying logic is fine, because you cannot trust that your work will still be there. That kind of unreliability discourages you from actually using a feature the way it is meant to be used, since you end up hesitant to navigate away or take a break partway through, worried that stepping away will cost you the work you already did.

Duplicate charges are a direct billing integrity issue: you should be charged exactly once per swipe, no more. This is not a matter of degree; a system that occasionally double-charges for the same action is a system you cannot fully trust with your billing, regardless of how rare the occurrence is.

And a run that retries forever without resolving wastes your time waiting on something that was never going to finish on its own. If you are relying on AI Leads to source a batch of contacts before starting a campaign, a stuck run with no clear end state means the rest of your workflow is blocked behind something you have no visibility into.

Fixing all three means Learning and AI Leads behave the way you would expect: your progress persists, your charges match your actions, and stuck work either fixes itself or tells you it needs attention. Those three properties, taken together, are what it actually means for a feature to be dependable enough to build a routine around.

This matters most for anyone using AI Leads regularly as part of how they source new contacts. A feature you use often is one where small reliability gaps compound over time, since you hit the edge cases more frequently just by using the feature more. Closing these three gaps means the routine use case, reviewing leads day after day, holds up the way it should.

Stuck runs also have a downstream cost beyond the run itself. If AI Leads is the first step in a pipeline that eventually feeds a campaign, a run that never resolves means everything after it stays blocked too, without any clear signal pointing back to the actual cause. Now that stuck runs surface as paused rather than looping indefinitely, you can identify and clear that blockage quickly instead of spending time troubleshooting further downstream in the pipeline for a problem that actually started earlier.

Taken together, these fixes are about closing the distance between what you did and what the product recorded. Progress you made should stay made. A single swipe should be a single charge. A run that cannot finish on its own should say so rather than pretending it is still working. None of these are dramatic changes to how Learning or AI Leads function day to day, but each one removes a specific way the product could quietly disagree with your own record of what happened.

Availability

These fixes apply to Learning and AI Leads for all users.