Internal tool · 2025 · Slack + Claude API
Persona Focus Group
The hardest part of designing for a platform with five distinct user types isn't the complexity — it's the time. Getting quality, stakeholder-specific feedback meant scheduling five separate conversations and hoping the questions still held up by the time the answers arrived. I built a tool to collapse that lag to thirty seconds.
Role
Product Designer & Builder · End-to-end
Stack
Slack app · Claude API · Vercel
The problem
Five user types. Five opposing mental models.
VentureCrowd serves five fundamentally different user types. They don't just have different needs — they have directly opposing priorities. A feature that builds trust with a Retail Investor can actively undermine confidence with a Wholesale Investor. A UX decision that excites a Founder can be a dealbreaker for a Fund Manager. Getting honest, specific feedback from each group required scheduling that rarely happened before a direction was already committed to.
Founder
Raising capital
Syndicate Lead
Running syndicates
Fund Manager
Managing funds
Wholesale Investor
Sophisticated capital
Retail Investor
Individual investors
The solution
A panel discussion in Slack — on demand.
The bot runs in Slack — ask a question, get responses in a thread. Ask all five personas at once and watch where they split, or direct a question at one for a focused deep dive. Each persona is a carefully constructed system prompt: specific background, stated fears, a communication style, and hard limits on what they'll accept — grounded in real VentureCrowd archetypes, not generic investor stereotypes.
A single-persona deep dive in under thirty seconds. The response is specific, opinionated, and grounded in how this user type actually thinks — not a generic UX principle about information hierarchy.
Example output
The value is in the tension.
The most useful outputs aren't the ones where everyone agrees. A question about showing the full cap table on every deal page produced directly opposing answers from the Founder and the Retail Investor. Neither answer alone would have shaped a useful design direction. The conflict between them did.
Same question · Opposite answers · Both valid
This thread didn't produce an answer — it produced a design brief. The right solution isn't full public disclosure or full opacity; it's a tiered visibility model that gives each persona what they actually need.
What this enables
Research happens before the commit, not after.
The practical value isn't just speed, though thirty-second research sessions instead of five scheduled interviews is genuinely useful. The deeper shift is that it changes when research happens. Because the cost of a gut-check is zero, I can pressure-test a product idea before investing any design time — not after a direction has already been handed to a developer. Copywriting decisions, feature scope, onboarding flows, pricing framing — things that would previously run on instinct now have a fast feedback loop.
Built end-to-end
The engineering is the scaffold. The value is in the prompts.
I built the bot end-to-end — Slack app wiring, Claude API integration, deployed on Vercel. The persona prompts were written and refined by me, drawing on direct knowledge of VentureCrowd's user base built over four-plus years. Each prompt includes a specific background, first-person voice, stated fears, and constraints that generate realistic tension between responses.