The biggest lesson for me has been that AI chatbot can mean very different things. Some are basically smarter FAQ bots, while others can actually take actions and work with your existing systems. That difference matters a lot once you’re in production.

For a small business, I’d probably avoid choosing based purely on the number of features. If the team can’t easily update the bot, review failed answers, and improve it without needing a developer, most of those features won’t matter much.

• This reply was modified 1 month ago by  IWUJI.

‎I’d say automation gets you to try an AI agent, but reliability is what determines whether you trust it. If I know the agent handles a particular workflow properly every time, I’m much more comfortable giving it more responsibility.<div>‎</div><div>‎</div>

What’s interesting is that this could completely change how SaaS products are designed. Instead of adding another button or dashboard, the product could simply understand the goal and figure out the steps. That feels like a much bigger shift than just adding AI features.<div>‎</div><div>‎</div>

I’ve been curious about this too. The approval part actually stands out to me because I’d want the convenience of natural language without giving the agent freedom to change important settings on its own. That balance makes Backstage interesting.

I’d probably start with one agent and keep the scope controlled. Multiple agents sound powerful, but if the user has to figure out which agent to talk to, the experience can get confusing pretty quickly.

‎I’d probably build around a boring problem that businesses already pay people to solve. AI is great, but the real value is when it takes a repetitive task off someone’s plate completely.

For me, the biggest lesson is that the first users usually come from actually talking to people, not just posting everywhere and hoping they sign up. Getting a few real users, listening to what they struggle with, and improving the product around that feedback can snowball into referrals pretty quickly. It’s not always the fastest looking approach, but it feels more sustainable.

Honestly, I think I’d lean toward the AI native approach, but only if the AI actually makes the support team’s job easier. I’ve seen plenty of tools where the AI sounds impressive in the demo but creates more work once real customer conversations start coming in. For me, the big test would be how well it handles handoffs, context, and the messy edge cases that automation usually struggles with.

I think it can simplify things, but only if the shared setup is actually well managed. Having one agent across channels sounds great until email, WhatsApp, and voice all need slightly different handling. For me, the real win would be keeping the context and core logic shared while still allowing each channel to behave naturally.

I think the sweet spot is avoiding customization just for the sake of it. I’d rather customize the parts that actually improve the customer experience and keep everything else simple. It’s easy to build something that works great today and becomes a headache to maintain six months later.

Honestly, I think launching without a waitlist makes sense, especially if the product was already usable. A waitlist can build some hype, but nothing really replaces having actual people use the product and tell you what’s confusing or missing. I’d probably launch early too and let the feedback shape the product.

For me, it’s AI being added just for the sake of saying a product has AI. If it creates more steps, interrupts what I’m already doing, or gives me suggestions I didn’t ask for, I usually stop using it pretty quickly. The useful AI features are the ones that quietly save me time without getting in the way.

I’ve wondered about this too. PocketBase feels like one of those tools that can take you surprisingly far if your product doesn’t have overly complex requirements. The part I’d be most curious about is what happens after the user base and data start growing, because that’s usually where the “great for an MVP” tools reveal their limits.

The live Shopify data is honestly the part that interests me most. A chatbot telling you where to find your order details isn’t that impressive, but actually pulling up the order and giving you the status is different. I’d love to hear from someone using it with real customers though, especially whether it’s actually cutting down those repetitive. Where’s my order?