I think reliability is probably the biggest thing here, businesses can be open to AI, but if the system makes mistakes or breaks an important workflow, the cost can be much higher than just paying for traditional software, AI-native tools definitely have potential, but having a good balance between automation and human oversight seems really important.

I really like the point about measuring the problems AI actually solves rather than just how well it talks, the part about AI potentially creating new problems is important too, if automation makes support faster but leaves customers frustrated, then the numbers alone don’t tell the whole story.

I agree that the real value of a forum like this comes from people sharing what actually worked for them, including the problems they ran into, that kind of practical experience can be much more useful than just reading feature lists, especially when you’re trying to figure out how AI support performs in real situations.

Yeah, I think the real value of a forum like this is seeing how people deal with problems that you don’t always find in the official docs, even a simple tip from someone who has already tried something can save a lot of time when you’re setting up an AI agent yourself.

I like the point about not trying to automate everything, it’s easy to focus on how many conversations AI can handle, but if it takes over the simple stuff and gives human agents more time for the difficult cases, that already seems like a pretty useful outcome.

For me, the biggest thing would be how well the platform handles the stuff that happens after the demo, a tool can look great on paper, but if the AI gives wrong answers or needs a human for every small issue, the benefits disappear pretty quickly, i’d probably care more about real resolution rates and how easily it fits into the existing workflow than having a huge list of features.

The part about support conversations feeding back into product improvements really stood out to me, if you keep seeing the same questions or problems, that’s useful information beyond just resolving the tickets, it can show you what needs to be fixed or explained better in the product itself.

I think the human handoff is just as important as the AI actually taking action, even if it can handle most of the steps itself, there will always be cases where it needs to stop and ask for help, if that handoff includes all the context and the customer doesn’t have to explain everything again, that’s a pretty big win.

I agree, especially if you’re coming from something like Zendesk, sometimes you learn more from seeing how other people actually solved a problem than from going through a bunch of documentation, the real world examples are probably the most useful part of a forum like this.

I think the hardest part would actually be moving over, not choosing the AI-native platform itself, if a team has years of workflows and customer history in Zendesk or Intercom, starting fresh can create its own headaches, but if the new system can make that transition smooth while actually reducing the amount of manual work agents do, I can definitely see the appeal.

I think the all-in-one approach makes a lot of sense, especially for smaller teams that don’t want to manage a bunch of separate tools, the real test will be whether all those features work well together and are actually reliable enough to replace parts of an existing support stack.

I agree with testing it against real customer questions, a platform might look great on paper, but what really matters is whether it can handle the situations customers actually bring to support, if it can solve routine issues without adding extra steps and leave the harder cases to humans, that’s where the value really shows.

I agree that integrations and human handoff are probably two of the biggest things to look at, a chatbot can answer questions, but being able to actually take action and then bring in a human when needed is what makes it much more useful in real support situations.

I agree with this dashboards are useful for spotting that something changed, but looking at the actual conversations gives you a much better idea of what needs fixing, recurring topics and negative feedback especially seem like good signals for improving the AI over time.

I think having one agent across multiple channels could definitely make things easier to manage, the main challenge would be keeping the experience natural on each channel, since people communicate very differently over WhatsApp, email, chat, and voice, if the agent can keep the right context while adapting to each one, that’s where the real benefit is.