We Rebuilt Our Product Twice to Learn One Thing

The journey from a Slackbot to a Full Fledged UI to a Slackbot again.

During our YC batch, we were strictly in Slack.

We believed traditional UI was not the interface to surface intelligence or agent output.

But we had every sign of feedback from customers telling us that Slack isn't the right interface.

Having a Slack-only bot for alerts put extreme pressure on getting every alert 100% accurate and valuable - anything else and it was spam.

Reasoning models weren't so strong back then and our ICP didn't really think in terms of alerts.

So we moved to a fully fledged self-serve UI.

It helped us speak our ICP's language - showing 1,000s of accounts upfront with high quality insights was an easy buying decision for customers.

They instantly saw the value and bought it.

But operationalizing? Became absolute hell. Endless UI/UX optimization. For a lean team it takes focus away from core value - debating on why someone isn't clicking a button is useless.

But then something started really shifting.

We launched our MCP / API and Webhooks.

Claude Code is picking up and GTM teams are thinking in terms of push/alert-based actions vs static data dump.

It turns out post the initial sell 2 weeks into the product usage, almost everybody prefers MCP / API / Webhooks / Slack

vs pre-sales where they prefer seeing a UI with big numbers to get an idea of the possibilities and an understanding what data exists behind the chat interface.