I spent 2 weeks doing nothing but support. Here's what I learned about my own product.
Three months into building SupportBridge, I made a decision that felt counterintuitive:
I spent 2 full weeks doing nothing but support.
Not building. Not selling. Not marketing.
Just reading and responding to every single email that came into our support inbox.
Here's what I found:
Week 1 observations:
The questions customers asked weren't random. They clustered into 4 distinct areas:
- Setup questions (how do I connect X?)
- Expectation gaps (I thought it would do Y, but it does Z)
- Edge cases (what happens when...)
- Trust questions (how do I know the AI won't...)
I hadn't anticipated the trust questions at all. They were some of the most important conversations I had.
Customers weren't asking "does it work?" — they were asking "how do I know when it's making a mistake?"
That question completely changed our roadmap. We prioritised transparency features I'd never planned.
Week 2 observations:
I started noticing how customers described their own problems vs how I described solutions.
The language gap was significant.
I used words like "escalation" and "routing." Customers said things like "gets sent to someone" and "doesn't get lost."
We rewrote our entire onboarding copy based on customer language, not our language.
Conversion on the setup flow improved noticeably.
What I'd tell any early-stage founder:
Don't delegate support until you've done at least 4-6 weeks of it yourself.
Not because you'll always do it better than a hire.
Because you'll never understand your customers as well from any other place.
Has anyone else done a "support immersion" like this? What did you learn?
This is a strong early-stage insight because it shows you weren’t just “handling support,” you were mining it for signal.
What stands out most is the shift from feature thinking to confidence thinking. The fact that users weren’t asking “does it work?” but “how do I know when it’s wrong?” is huge—that’s not a UX issue, it’s a trust architecture problem. Most founders miss that entirely until much later.
The language gap insight is equally important. You effectively discovered that you weren’t just translating product → user, you were translating mental models. That alone can reshape onboarding, docs, and even product naming.
The strongest takeaway here: support isn’t a cost center—it’s a live dataset of product truth.
Question:
What was the single most repeated support message that forced you to change something immediately?
Thanks Stephan — really appreciate the thoughtful comment.
You nailed it on the “language gap” part. That was one of the biggest surprises for me. I was writing copy like an engineer (“escalation”, “routing”, “confidence score”), but customers were saying things like “don’t get lost” or “someone actually reads this?”. Completely different mental models.
Answer to your question:
The single most repeated message that forced an immediate change was:
“How do I know the AI won’t make a mistake?”
It came up again and again in Week 1. That one question completely shifted our roadmap. We added the “Max 1-Reply Rule” + very clear escalation transparency because customers weren’t asking “does it work?” — they were asking “how do I trust it won’t embarrass me?”.
That trust question ended up being more important than any feature request.
— Harsh