Show HN: BarrierX – AI that finds which lost deals are worth re-engaging now

Hey HN – Kas here, founder of BarrierX. The problem I kept seeing: I spent years watching B2B sales teams treat "Closed Lost" as a graveyard. Thousands of deals sitting in CRM, never touched again. But here's the thing – most of those deals aren't actually dead. They're just badly timed. Budgets get unfrozen. Champions change jobs and land at new companies. Competitors drop the ball. Reorgs happen. The same prospect who said "not now" 8 months ago might be ready today – but nobody's systematically tracking this. Meanwhile, reps burn hours chasing net-new leads from the same generic ZoomInfo lists everyone else has. What BarrierX does: Instead of scraping public data like traditional sales intelligence tools, we analyze your proprietary data:

Ingests your CRM history (won and lost deals) Analyzes patterns: what signals preceded wins vs. losses at your company Monitors lost accounts for trigger events (job changes, funding rounds, leadership changes, tech stack shifts) Surfaces which lost deals are worth re-engaging right now – with reasoning on why

The core insight: the patterns that predict success for your business are specific to you. A "good fit" at Company A looks completely different than at Company B. Generic intent data misses this entirely. Early results: In pilots, teams have resurfaced 15-20% of deals marked "closed-lost" as re-engageable within 90 days. Most of these would have sat untouched forever. On invite-only: We require CRM integration to work properly (the whole point is learning from your data), so we're onboarding gradually to ensure quality. If you're from HN and want to try it – reply with your CRM setup (Salesforce, HubSpot, etc.) and rough deal volume, and I'll get you access this week. What I'd love feedback on:

Has anyone else tried systematically working lost deals? What worked/didn't? For those who've used Gong/6sense/ZoomInfo – what's actually useful vs. noise? Any concerns about the approach I should be thinking about?

Happy to answer questions about the architecture, how we handle data, or anything else.