We learned that an AI customer agent shouldn't try to answer everything

Hey Indie Hackers! 👋

After launching MySampark, we've been thinking a lot about one question:

What should an AI customer agent actually do?

At first, the obvious answer seems to be: answer every customer message.

But the more we looked at real business conversations, the more we realised that's not really the goal.

Businesses get a lot of repetitive questions:

“What's the price?”
“Is this available?”
“Do you deliver?”
“How long does delivery take?”
“What are your opening hours?”

These questions don't necessarily need a person sitting there waiting to answer every few minutes.

But then you get a completely different type of conversation:

“I received the wrong product. What should I do?”

“I have a problem with my order.”

“Can I speak to someone?”

That's where we think the human still matters.

So with MySampark, we're working toward a simple principle:

🤖 AI handles the routine.
👤 Humans handle what needs a human.

The AI Agent can respond to customer conversations using the business's information, while the team can step in when a conversation needs personal attention.

We're also trying to avoid the old-school approach of only responding when someone types an exact keyword.

A customer shouldn't have to say exactly “price” for the system to understand that they're asking about the price.

We're still improving this and learning from real conversations.

For founders building AI products:

What's one thing you've learned about where AI should stop and a human should take over?

I'd genuinely love to hear how others are approaching this.

— Hemangi
MySampark