.png)
Picture this: your AI startup has 14 months of runway, a founding team of six, and a backlog of 200+ feature ideas from beta users. You can't build everything. You can't even build 10% of everything. You need a systematic way to screen possibilities and find the ones worth pursuing — the exact challenge that drug discovery natural products research has been solving for decades.
Company Context
NovaMind (name changed), a Series A AI-native startup building intelligent document processing tools, faced this scenario in early 2025. Their UX research lead, Dana, had a background in pharmaceutical research before transitioning to tech. She recognized that the team's chaotic feature prioritization mirrored the inefficiency of unstructured compound screening in early-stage drug research.
NovaMind had 3,200 active users, a burn rate of $180K/month, and a product that worked well for legal documents but showed weak retention in financial services use cases. The question was not "what should we build?" but "how do we systematically find what matters?"
The Challenge
The team had tried conventional prioritization: voting, RICE scoring, and stakeholder debates. Each method produced a different top-five list. Engineering was frustrated by constant reprioritization. Sales pushed for enterprise features. Support flagged usability gaps. Without a structured screening process, every sprint planning session became a negotiation rather than a decision.
The Drug Discovery Natural Products Approach
Dana proposed adapting the natural products screening funnel — a methodology refined over 40 years of pharmaceutical research — to product discovery. The framework has three stages:
Stage 1: Extract Library Creation (Week 1-2)
In pharma, researchers build libraries of natural extracts before testing anything. Dana's adaptation: catalog every feature request, support ticket, churned-user interview, and sales objection into a structured "extract library" using Airtable. Each entry included the source, the stated need, the inferred job-to-be-done, and the user segment. Final count: 247 discrete entries.
Stage 2: Primary Screening (Week 3)
Pharmaceutical primary screening eliminates compounds that fail basic activity thresholds. Dana designed a three-filter screen: (1) Does this address a need expressed by 5+ users in the target financial services segment? (2) Can it be prototyped in under two weeks? (3) Does it align with the core AI processing capability? Of 247 entries, 31 passed all three filters.
Stage 3: Secondary Assay and Validation (Week 4-6)
Surviving candidates undergo rigorous testing in drug research. Dana's team ran rapid prototype tests with 8 financial services users per concept. They used Maze for unmoderated testing and Dovetail for interview synthesis. Each concept received a composite score based on task completion rate, stated willingness-to-pay, and frequency of the underlying need.
Results
Three concepts emerged as clear winners, each scoring above 80% on the composite metric. The top candidate — automated compliance annotation — showed 92% task completion and had 6 of 8 testers say they would pay extra for it.
NovaMind shipped compliance annotation in 5 weeks. Within 90 days, financial services retention improved from 23% to 41%. The feature became their primary differentiator in that vertical and directly contributed to closing two enterprise contracts worth $340K ARR.
Lessons Learned
Structured screening beats intuition. The compliance annotation feature ranked 47th on the original stakeholder voting list. Without systematic filtering, it would have been buried for months.
Stage gates prevent waste. By eliminating 87% of candidates in the primary screen, the team invested prototype-testing resources only where they mattered most.
Cross-disciplinary methods transfer. The drug discovery natural products funnel worked because the underlying logic — systematic reduction of a large candidate pool through increasingly rigorous filters — is domain-agnostic.
Speed requires constraints. The six-week timeline forced decisions. Without a deadline, the screening process would have expanded to fill available time.
For UX researchers at AI-native startups, drug discovery natural products methodology offers a battle-tested structure for turning chaotic backlogs into focused, evidence-based product bets.
Learn more about systematic discovery approaches in our guides on how to do product discovery and product discovery phases. For the broader framework behind this methodology, see natural products and drug discovery and natural products drug discovery. Also explore natural products in drug discovery and drug discovery from natural products for complementary perspectives.