I think I will probably look at Coolify less from the “how easy is it to deploy?” angle and more from “how much work will it still be giving me months later?” If I am constantly dealing with backups, updates, monitoring, failed deployments, or server issues, then the cheaper setup may not really feel cheaper anymore. For me, I would only choose it again if the control and savings still made sense after counting the time spent maintaining everything myself.
Speaking from my own point of view though.

A lot of these “AI agent” launches still feel like SaaS companies putting a smarter front door on the same old product. I look at it like the real move starts when the agent actually removes the need to learn the software itself when you can just explain what you want done and it handles the menus, forms, and workflows in the background. At that point, the best SaaS tool might not be the one with the most features, but the one you barely have to think about using.

Things I will most probably change if I relaunch my SaaS today:

1. I will make onboarding much simpler, because even a good product can lose people if they don’t understand the value quickly.

2. I will focus the launch around one strong use case instead of trying to show everything at once. I think proving one thing really well first would make it easier to build trust and grow from there.

I’d probably look at it a bit differently from the way it is. The real surprise after launch might be how much the tool changes the team’s normal way of working, because even a powerful platform can become annoying if everyone has to constantly adjust their process just to make the AI fit in.

I personally think that Actions become really useful when the customer doesn’t have to worry about what is happening behind the scenes. If the AI can understand the request, get the information it needs, trigger the right action, and come back with the result in the same conversation, the whole thing feels a lot smoother and less like you’re dealing with separate systems.

A good demo can also make an AI agent look really impressive, but the real test is what then happens when things stop being perfect or going well. Can it still handle failed APIs, limited permissions, confusing conversations? and know when to bring in a human, then I’d say that’s when you know the system is actually reliable and dependable.

I think AI is going to make online buying more competitive in a way that affects what businesses are actually competing for. In the past the main battle was to get the customer’s attention, but if AI agents become the middleman, brands may have to compete to be the option that the AI sees as most trustworthy, relevant, and worth recommending, and that could reward smaller businesses with genuinely better products while exposing bigger brands that rely mostly on visibility.

I think Chatbase becomes more useful when it helps reduce what we don’t know, not just response time. If it can explain why a product fits my needs, flag possible drawbacks, and still know when a human should step in, then it feels more like real shopping assistance than just another chatbot.

For me, the bigger change would be an AI that can tell when I’m about to make a bad purchase, not just help me make one faster. If it can point out that a product is overpriced, poorly reviewed, or unnecessary based on what I already have, then it’s actually improving the shopping decision instead of just speeding it up.

Another mistake, in my opinion, is assuming an AI agent should respond to every query. Knowing when not to push another feature, follow-up, or sales message can be just as important as knowing what to say, especially when a prospect simply needs time to think.

The way AI is gradually changing the meaning of “convenience” in online shopping intrigues me. Convenience used to mean having thousands of products available online, but now it’s becoming more about not having to sort through those thousands of options yourself.

What I like about this is that AI doesn’t necessarily take the salesperson out of the picture; it just changes when they need to step in. There’s really no point having a human answer the same basic questions all day when they could be handling the conversations where trust and actual decision-making matter.
So basically, salesperson still has his/her job secured.

I believe that we occasionally gauge the effectiveness of AI sales representatives by how well they guide clients along a funnel, even when the client may not yet be prepared to do so. Even a quicker path to the incorrect result is a poor customer experience.
AI sales agent should ultimately be able to perform its tasks judiciously and still be understanding to its clients and provide what they need.

I think the interesting thing is that we might not look for items at all, rather than asking “where can I buy this? or even get this?” We’ll tell an AI what we need, our budget and what’s important to us and let it do the hunting.

I feel like when the Human steps in, it make things wayy easier. Resolving hat needs to be resolved