Back
Case study

When every RFP on the page looks the same, how does a vendor know which one is theirs to win?

A 4-screen RFP marketplace, designed from a timed brief that allowed AI-assisted execution on one condition: it couldn't look like AI had done it.

Role · Solo design Screens · Login, Listing, Detail, Dashboard Constraint · 3-hour brief, AI-assisted
Scan
The problem
In a marketplace with hundreds of live RFPs, near-identical titles from different organizations are common. A vendor who can't tell two listings apart without opening both either wastes time or misses the right opportunity. That's a trust problem for the marketplace, not just a UI annoyance.
What I designed
A public listing, an RFP detail page, a vendor dashboard, and login, plus the token and component system underneath all four.
What changed
A generic AI-typical first pass became a system with real authored decisions in it, most visibly in how the listing solves its one hard problem, and in a mid-project reframe of what the dashboard was actually for.

Everything below this line is the scrutiny layer: reasoning, wrong turns, and the system. Skip it if the scan told you what you needed.

Collect

The brief was short, but three lines in it did all the real work.

"Some RFPs may have very similar titles and categories. Design the listing so vendors can confidently identify and differentiate the right RFP without opening every result." That's a discoverability problem wearing a listing-page costume.
"Think about how a vendor can quickly understand whether an RFP is relevant to them." Buried in the dashboard requirement, this line quietly redefines its job as relevance, not record-keeping.
"AI use is allowed, but it shouldn't look like a design an AI created." Not a feature requirement. A constraint on every visual decision that followed.

Build: four decisions

Nothing else in the brief needed real invention. Login, a detail page, a dashboard shell are well-understood patterns on their own. Where I spent design time tracked directly to these decisions, not to the screen count.

01 / 04

Make the card answer the question, not the detail page

Problem
"IT Infrastructure Upgrade Phase 1" and "IT Infrastructure Upgrade Phase 2" sit back to back in a list of 128. A title-and-category card, which is what the first AI-generated pass produced, makes them look like duplicates.
Discovered
The information that actually tells them apart (organization, budget scale, deadline, RFP ID) already exists. Convention just hides it behind a click, on the assumption that cards should stay minimal. That assumption is exactly what breaks this brief.
Decision
Promote organization, RFP ID, budget range, and deadline into the card itself, in a consistent scannable position.
Result
Delhi Municipal Corporation, ₹42L to ₹58L, closing in 6 days. Pune Smart City Ltd., ₹1.1Cr to ₹1.4Cr, closing in 14 days. Obviously different opportunities, no need to open either one.
Public RFP Listing screen, showing two similarly-titled IT Infrastructure Upgrade cards distinguished by organization, budget, and deadline
Public RFP Listing. The differentiation problem, solved at the card level.
02 / 04

Reframe the dashboard from "my activity" to "what matters today"

Problem
A vendor dashboard's default job, in almost every reference pattern, is tracking your own submissions: status, dates, outcomes.
Discovered
The brief's actual ask was different: help the vendor know if an RFP is relevant to them, using visualizations. My first pass was a pure proposals table before I caught this.
Decision
Lead with relevance instead: matched RFPs with a visible match score, plus a category-breakdown chart. Move the proposals table down to second position.
Result
The dashboard now answers "what should I look at today" before it answers "what did I already do." That's the order the brief actually asked for, once I read it closely enough to notice.
Vendor Dashboard screen, leading with Recommended for You matched RFPs before the proposals table
Vendor Dashboard. Relevance first, tracking second.
03 / 04

Treat the detail page as a live decision, not a static document

Problem
An RFP detail page is conventionally a wall of text: scope, eligibility, terms, read once and left.
Discovered
Everything on this page is time-pressured and decision-relevant: a real deadline, evaluation weightings that determine where to spend proposal effort, a status that moves through stages. Static prose throws that away.
Decision
Split the page. Narrative content on the left. A persistent panel on the right that behaves like a small dashboard: a countdown with a progress bar, evaluation criteria as a weighted bar chart, and the primary action always in view.
Result
A vendor sees, in the same glance, how much time is left and where evaluators will actually weight their decision.
RFP Detail Page screen, showing a deadline countdown and evaluation criteria bar chart in the right-hand panel
RFP Detail Page. Narrative on the left, live decision support on the right.
04 / 04

Make "don't look AI-made" a system decision, not a styling pass

Problem
The first AI-generated draft was functionally complete and visually generic: the default blue-purple gradient, uniform type weight, no real hierarchy.
Discovered
The tell isn't messiness. AI output is usually clean. The tell is uniformity: no bespoke color decision, no typographic point of view, nothing distinguishing carried through consistently.
Decision
Rebuild the token system by hand before touching any screen again. An emerald accent instead of the reflexive gradient. A genuine 8px grid. IBM Plex Mono reserved for numbers only. Every button, input, and nav item got real Default, Hover, Pressed, and Disabled states.
Result
The system-level decisions carry across all four screens instead of being re-argued on each one. It's also why login, the simplest screen, still reads as part of the same product.
Login screen, showing the emerald brand panel and vendor/admin role toggle
Login. The plainest screen, carrying the same system as the rest.

Assemble

The system underneath the four screens, shown together.

Full Figma canvas showing all four screens alongside the component library: buttons, badges, inputs, nav items, and links in their various states
The full file. Screens and component library, side by side.

Color

Type

Inter, interface and narrative Numbers get their own voice. IBM Plex Mono: ₹42L · 06 days · RFP-2026-0143

Components, built once

Under scrutiny

What I'd tell you if you asked twice.

Intervention

Three hours was never the real budget. Getting from a generic AI-assisted first pass to a system that holds together (consistent shell, real icons, real interaction states, plus a layout bug where two screens had silently doubled in width) took several more passes after the timer ran out.

I assumed the dashboard was a tracking tool. It wasn't.

That's the single biggest structural change in the project, and it came from rereading the brief, not from user research I don't actually have on a timed exercise.

The real skill in a timed brief isn't finishing everything.

It's triaging correctly what can stay rough and what can't. I spent the least effort on login, the most on the listing card, because that's where the brief's actual difficulty lived.

The brief's two-sided marketplace only required vendor-facing screens.

Organizations creating and managing RFPs are real in the brief's premise but out of scope for the four required deliverables. Worth knowing if you're wondering why there's no admin view here.