Why I'm building Midevela
Over the last few months, I've been building Midevela, an AI sales employee for online stores.
The idea started with a simple observation.
Most AI chatbots can answer questions.
Very few can actually help someone buy.
If you've ever shopped online, you've probably asked questions like:
"Which one is better for oily skin?"
"I'm looking for a laptop under $800."
"What's the difference between these two products?"
A good salesperson understands what you're trying to achieve, asks the right questions, remembers the conversation, and guides you toward the best choice.
Most chatbots don't.
They answer one question at a time, lose context when the conversation changes, and often require merchants to spend hours configuring rules before they're useful.
That's the problem I'm trying to solve.
Instead of building another support chatbot, I'm building an AI that behaves more like a knowledgeable sales employee. It learns a merchant's product catalog and business information, recommends products naturally, keeps track of customer intent, and knows when it's better to hand the conversation over to a human.
I'm still in beta, and I'm sharing the journey here because I want to build this in the open and learn from other founders.
I'm curious:
If you run an online business, what's the biggest limitation you've experienced with today's AI chatbots?
I'd love to hear your experiences.
The distinction between answering questions and actually helping someone buy is what stood out.
Since you're already in beta, have merchants been using Midevela in the way you expected, or has anything about their actual usage surprised you so far?
That's a great question. We're still in the early beta stage, so one of the biggest things we're validating is exactly that whether merchants use Midevela the way we expect or in ways we haven't anticipated. One hypothesis we're testing is that merchants don't just want an AI that answers questions; they want one that can confidently guide shoppers toward the right product without sacrificing trust. I'm expecting the biggest surprises to come from watching real customer conversations, and we're building the product to learn from those interactions rather than assuming we already have all the answers.
That’s helpful context. The distinction between answering questions and actually influencing a buying decision is an interesting one to validate.
I’d like to continue the conversation outside the thread. What’s the best email to reach you on?
midevelaos @gmail .com
Thanks! I’ve just sent it over.
Looking forward to hearing your thoughts whenever you have a chance.