Proladex is a niche B2B fintech platform that acts as a curated marketplace and diligence layer between investment firms and third-party service providers. It connects funds, family offices, and allocators with vetted consultants, tech vendors, and specialist firms. Providers go through an application-based onboarding with verification before their profiles go live. The platform streamlines the historically painful "find, vet, reference-check, then engage a vendor" workflow that investment teams deal with on every transaction.

Investment professionals rely on personal networks and cold market searches to find service providers for due diligence work. This process is slow, inconsistent, and heavily relationship-dependent. Teams default to whoever they already know rather than whoever is the best fit. Meanwhile, qualified service providers struggle with visibility and inbound lead generation despite having strong credentials.

Proladex already solves the discovery and vetting side. The question became: how do we take this further? How might we use AI to make the matching smarter, the engagement faster, and the RFP process nearly effortless — so investment teams can go from need to shortlist to engagement in minutes?

Our founder worked directly with the Proladex team to understand both sides of the marketplace — the investment professionals searching for providers and the service providers seeking visibility. Those insights were then relayed to me to inform the design direction.

Research inputs included detailed briefs from our founder based on their discussions with Proladex about the vendor lifecycle from discovery through engagement, a review of existing user feedback and provider application data shared by the Proladex team, competitive analysis of platforms like Clutch, G2, Gartner, and expert network tools, and a full breakdown of the pain points in the current vendor sourcing workflow as described by investment professionals through Proladex.

Key findings:
• Investment teams reuse the same handful of providers regardless of deal fit because searching for new ones takes too long
• Budget and timeline are the two hardest constraints to match against, and getting either wrong wastes weeks
• Decision-makers don't have time to read through lengthy vendor profiles — they need ranked, explained, scannable results
• RFP creation is tedious and repetitive — roughly 80% of every RFP is the same boilerplate, but the 20% that's deal-specific actually determines the quality of vendor responses
• Service providers want more inbound visibility but have no way to signal credibility beyond their own marketing

Based on the insights relayed from our founder's discussions with the Proladex team, I developed two primary personas representing the core user types.

The first is a senior investment professional — a managing director type in their late 30s who runs multiple deals simultaneously and needs service providers across categories like Quality of Earnings, legal advisory, IT diligence, and management consulting. Their main frustration is always going with whoever they already know rather than finding the best fit. They want to shortlist 3 vetted vendors in under 5 minutes. They're highly technical and live in tools like Bloomberg, Slack, and Excel.

The second is a mid-level associate — a VP type in their late 20s who executes vendor sourcing on behalf of deal leads. They draft RFPs, manage vendor communication, and coordinate timelines. Their main frustration is spending hours writing RFPs that are almost entirely the same every time. They want to generate a professional, deal-specific RFP in minutes. They're SaaS-native and adopt new tools quickly.

From the research handoff, I identified the core workflow that every deal team follows: describe the need, find matching vendors, compare options, build a shortlist, then send an RFP. This five-step loop became the product's backbone.

Feature prioritization:
• Must-haves: AI copilot for vendor matching, vendor directory with rich profiles, AI-powered RFP generation, and shortlist management
• Should-haves: Vendor comparison view, session history, dashboard analytics, and dark mode support
• Could-haves: Team collaboration, direct RFP email delivery, vendor messaging, and a mobile app

Information architecture:
• Dashboard — home base with KPIs and recent activity
• Copilot — the AI chat and core feature
• Vendors — browsable, filterable directory
• RFPs — list view and AI-powered builder
• Shortlists — saved vendors organized by category
• History — past copilot sessions for reference

This is the AI-powered chat where users describe what they need in plain English. The system returns ranked vendor matches with scores, explanations, and actionable next steps. I chose a chat-first approach over a structured form to reduce friction and mirror how deal teams actually talk about vendor needs. During AI processing, animated streaming steps show what's happening instead of a generic spinner. Each match card surfaces the score, vendor category, and a brief explanation at a glance, with three actions: View Profile, Shortlist, and Draft RFP.

Each result is designed for fast decision-making. A color-coded match score gives instant signal quality — green for strong matches above 80%, blue for solid matches in the 70–80% range, and amber for moderate fits. Every score is paired with a one-line explanation of why this vendor was matched. Users can give inline thumbs up/down feedback and directly shortlist or start drafting an RFP without leaving the conversation.

Users provide minimal context — the vendor name, deal type, budget range, and timeline. The AI generates a complete 7-section RFP covering Introduction, Scope of Work, Timeline and Milestones, Budget and Fee Structure, Team Requirements, Evaluation Criteria, and Submission Instructions.

The critical design decision here was section-level editing. Users can regenerate any individual section with specific feedback without losing edits to the rest. The builder auto-saves every 30 seconds, supports a toggle between editing and formatted preview, and tracks each RFP through Draft, Sent, and Responded stages.

A command center showing four KPI cards with 30-day sparkline trends, recent copilot sessions, saved shortlists, and a summary of top-rated vendors across categories.

The browsable database supports two view modes — card view for visual scanning and table view for side-by-side data comparison. Users can filter across the seven service categories that Proladex covers: Quality of Earnings, Legal Advisory, Management Consulting, IT Diligence, Insurance, HR and Recruiting, and ESG. Each category carries a distinct color that stays consistent everywhere in the app.

The aesthetic direction is refined and utilitarian. This is a tool for finance professionals making high-stakes decisions, so the interface avoids anything decorative or trendy.

The typeface is FK Grotesk in two weights — regular and medium. It's geometric, professional, and distinct without being distracting. The singular accent color is a decisive blue (#0047FF) used exclusively for primary actions. Dark mode uses a neutral scale that reduces eye strain during the long working hours these users have. Light mode uses OKLch color space for perceptual uniformity across the palette.

Components are built on shadcn/ui with the New York style variant, giving us accessible, themeable primitives from day one. The seven vendor categories each have a unique color used consistently across match cards, the directory, shortlists, and dashboard charts so users build category recognition over time.

The primary flow starts at login, which lands on the Dashboard. The user clicks "Start a new search" and enters the Copilot. They type something like "I need a QoE provider for a healthcare roll-up, budget around $75k." The AI returns three ranked matches. From any match card, the user can shortlist the vendor, view their full profile, or draft an RFP. Choosing Draft RFP opens a minimal form, and the AI generates a full 7-section document in about 8 seconds. The user edits what they need, previews the formatted version, and marks it as sent.

AI response latency (3–8 seconds): A blank screen feels broken. I designed streaming progress steps with staggered animations that narrate what the AI is doing. This made the wait feel purposeful rather than uncertain.

Match scores lack context on their own: I paired every score with a color-coded badge for instant signal and a one-sentence explanation of why this vendor fits the stated requirements.

Users want AI-generated content but need control: Section-level editing solved this — users can regenerate any of the 7 RFP sections with specific feedback without losing edits elsewhere.

Balancing information density with speed: I used progressive disclosure throughout. The dashboard shows KPIs at a glance, vendor cards show the essentials, profiles expand to full detail, and card-to-table view toggles let users choose their preferred density.

• Query to vendor shortlist takes roughly 30 seconds
• Full 7-section RFP generation takes about 8 seconds
• 8 major screens designed end-to-end
• Over 50 UI primitives and 15+ custom components
• Any feature reachable in 2 clicks or fewer from the sidebar
• Full dark mode coverage across every screen

Trust is the real product. For AI tools in financial services, showing why a recommendation was made matters more than the recommendation itself. Match explanations, vendor data transparency, and feedback mechanisms aren't nice-to-haves — they're the core of the experience. This aligns directly with Proladex's value proposition as a trust and vetting layer.

Streaming needs its own UX language. Traditional loading spinners don't work for AI interactions. Progressive loading states that describe what's happening in real time made users noticeably more patient and confident in the output.

Density and clarity can coexist. Progressive disclosure, view mode toggles, and strong typography hierarchy let us serve both the power user scanning 20 vendors and the director who just needs 3 names in 30 seconds.

Knowing the codebase changes how you design. As a designer who understands the technical implementation, I was able to design with real component constraints, propose feasible streaming UX patterns, and communicate more effectively with the engineering team during handoff.

• Multi-user collaboration with shared shortlists and co-edited RFPs
• Vendor performance analytics tracked across engagements
• Direct RFP delivery and vendor response tracking through the platform
• Deeper integration with Proladex's existing vetting and onboarding pipeline
• Mobile-optimized experience for on-the-go deal teams