AI conversations inside the product don’t just help customers get work done. They reveal what customers are trying to achieve while the work is happening - and that may change how we build products.
For years, product teams have learned from behavior.
We watched which screens customers opened.
Which buttons they clicked.
Which reports they exported.
Which workflows they abandoned.
From those signals, we tried to infer what customers actually needed.
As more products add AI chat, most discussions focus on whether chat will replace traditional interfaces.
I think something much more interesting is already happening.
For the first time, we can observe customers while they're still figuring out the problem they're trying to solve.
Not afterwards.
While the work is happening.
To me, that's far more interesting than whether chat becomes the next product interface.
Behavior tells us what happened. Conversations tell us why.
Imagine opening a navigation app and saying:
"I need to get home, but I also have to stop at a pharmacy and pick someone up before six."
You didn't ask for a route.
You described a goal.
The conversation exposed the constraints, priorities, and outcome you actually cared about.
That conversation happened inside the product - and it taught the product something it couldn’t learn from clicks alone.
The same thing happens in enterprise products.
A customer might ask:
"Help me prepare for tomorrow's customer meeting."
At first, it sounds like they need a meeting summary.
A few follow-up questions later, something else emerges.
They're not trying to prepare for the meeting.
They're trying to answer one question:
"Is there anything that could surprise me tomorrow?"
That's a completely different product problem.
Without the conversation, we might build a better meeting summary.
With the conversation, we discover that what customers actually need is a risk review.
The follow-up is often more valuable than the original request.
Conversations create a new product signal
Traditional analytics tell us what users did.
Conversations reveal:
- what they were trying to accomplish
- what they were uncertain about
- which assumptions they corrected
- what they expected to happen next
Those signals are difficult to discover through clicks, analytics, feature requests, or user interviews alone.
Not because those methods aren't useful.
Because conversations happen while customers are doing the work - not after.
Repetition is where products are born
One conversation is interesting.
A hundred similar conversations are a roadmap.
When customers repeatedly ask the same follow-up question...
When they consistently need the same context before making a decision...
When the same reasoning keeps appearing...
The next improvement probably isn't a better prompt.
It's a better product.
Product recognizes that the capability deserves to exist.
Engineering turns the stable reasoning into deterministic logic, workflows, and code.
The conversation revealed the need.
Repetition revealed the opportunity.
Chat and dedicated experiences serve different needs
I don’t think every conversation should become a feature.
Some requests will always be unique.
Chat works well for exploration, investigation, and highly specific needs.
Dedicated experiences work better when the same job keeps coming back.
The goal isn’t to replace one with the other.
It’s to use each where it works best.
A different way to think about AI products
We already measure response quality.
Task completion.
User engagement.
I'd add one more question:
What did our customers teach us today that our product still doesn't know how to do?
Because maybe the biggest opportunity AI gives product teams isn't a new interface.
It's a new way to learn.
Traditional products showed us what customers did.
AI conversations let us understand what they were actually trying to accomplish.
And your next feature might already be hiding there.
Top comments (0)