From my perspective, integrations and actions are becoming just as important as the model behind the agent. An AI can give a perfect answer, but if it can’t access the customer’s order, update something, trigger a workflow, or hand off the conversation with the right context, the human team still ends up doing most of the work. I’m also with you on looking beyond demos.

With live chat, you’re often waiting for someone to click “chat” and ask a question. An AI agent can be more proactive, helping compare products, explaining pricing, or qualifying someone while they’re still exploring the site. But I wouldn’t look at conversion rate alone. I would want to know whether the leads are actually good and whether the AI is helping people reach a decision faster.

One newer SaaS product I’ve been paying attention to is Chatbase. What stood out to me isn’t just the AI chatbot side, but how much they’re trying to build around the actual customer support workflow. The combination of AI agents, procedures, testing, analytics, and human handoff makes it feel less like “add an AI widget to your website” and more like building an actual support system.

I have noticed the same shift. What makes it interesting to me isn’t just the number of features Chatbase has added, but the move from answering questions to actually handling support workflows. Actions and Procedures are probably the biggest indicators of that change. Once an AI can take an action, follow business rules, and then hand off to a human with the conversation context intact, it starts looking less like a chatbot and more like part of the support operation.

One I would add is Better Stack. It doesn’t get mentioned as much as some of the bigger SaaS names, but it solves a very real operational problem around monitoring, incident management, and keeping teams informed when something goes wrong. What I like about tools like this is that the value is pretty easy to measure. If a tool helps you detect an issue faster, communicate it clearly, and reduce the time spent jumping between different systems, it’s doing its job.

For a small business, the biggest benefit of AI support isn’t necessarily replacing people, it’s making 24/7 responsiveness possible without needing to hire around the clock. What really matters, though, is whether the AI actually gets customers to a resolution. Handling FAQs is easy, knowing when to stop, escalate, and give a human the full context is where the real value is.

The regret usually comes from treating AI as a feature instead of changing the workflow around it. Customers might love the idea of an AI assistant, but that doesn’t automatically mean they’ll trust it with important tasks. I’ve found that the boring use cases can actually create more value, answering common questions, finding information, checking order status, or handling simple requests.

This is a good point because I think pricing consistency can sometimes create a false sense of simplicity. The price may be the same, but the cost of getting that sale isn’t. I would be more interested in looking at profit per successful sale by channel than just comparing the listed price.

I have had that experience where a tool looks great initially, but the real opinion forms after a few weeks of using it in your normal workflow. The small things you initially overlook, alerting, incident context, setup friction, and how much manual work is still required, can end up mattering more than the headline features.

I have seen AI genuinely help when it handles repetitive work, pulls the right context, or moves a task toward completion instead of just generating another response. But adding an AI teammate to a product doesn’t automatically make the product better. For me, reliability, ease of use, and whether it actually solves the business problem still matter more.

An agent is only really useful if it removes steps from a workflow rather than giving me another interface to manage. If I still have to check its work, correct its answers, move information between systems, and eventually do the task myself, then I’m not sure we’ve gained much. The most valuable AI use cases I’ve seen are the ones where the agent can actually move something forward, retrieve the right information, take an action, and escalate when it reaches its limits.