AgileRadar
AI-Assisted Public Tender Scouting & Proposals
- Role
- Product design and development
- Studio
- Shape Creative Studio ↗
- Platform
- Internal web app
- Data
- Mercado Público · Compra Ágil
01
Project Overview
AgileRadar is an internal intelligent bidding scouting and proposal drafting platform built for Shape. The system interfaces with Chile's public procurement portal (Mercado Público) to evaluate incoming bids and draft commercial proposals in minutes.
02
The Challenge
Scouting public bids on government portals is a tedious process requiring managers to read through extensive PDF or Word documentation (typically 50-100 pages per bid). Evaluating technical fit is manual and consumes days, meaning high-value opportunities can easily be missed. The challenge was to centralize, automatically assess fit using AI, and generate technical-commercial proposal drafts.
The product in use
Screenshots of the platform running with real Mercado Público tenders.
Opportunity board
Shape's accreditation status, pipeline indicators and a kanban by stage, from opportunity to submitted, with each tender's match.
AI evaluation
Each tender opens in a detail panel with its Mercado Público data and the radar analysis: a 0 to 100 score with its justification.
Generated technical proposal
The AI drafts the technical and commercial proposal and the data for the Compra Ágil form; it can be regenerated or downloaded as Markdown or Word.
Supplier accreditation
Shape's status in Mercado Público (enabled, disabled or expired) with the fee expiration date, so bids are never submitted out of term.
Urgency filters
Urgent deadlines (2 days or less), upcoming (5 days or less), region and tender type to prioritize where to focus.
Responsive on tablet
The same dashboard adapted to tablet, with indicators and filters rearranged.
AgileRadar Design System
A light, information-dense interface built with Tailwind CSS v4: violet for action, Shape's warm palette as accent and the system font.
Color Tokens
Primary, secondary, and neutral system tokens built for contrast and consistency.
Primary action, buttons and active states (violet-600)
Brand accent in gradients and highlights
Main app background
Cards and panels
Accent backgrounds and borders
Titles and figures (gray-900)
Metadata and labels (gray-500)
High matches and positive states (emerald-600)
Alerts and discarded opportunities (rose-500)
Typography Scale
Primary Typeface: System font (Tailwind font-sans) (Web app)
UI Component States
Interactive primitives sharing token geometry, borders, and typographic rules.
Buttons
Form Inputs
Value Proposition
We designed an automated pre-qualification workflow combining natural language processing and real-time visualization:
- Real-Time Kanban Board: Opportunities organized across pipeline stages, utilizing WebSockets for instant interface updates.
- AI fit-scoring: each tender is evaluated against Shape's capabilities with a 0 to 100 score and a written justification. Groq (GPT-OSS models) is the primary provider, with Gemini and a local model (Ollama) as fallbacks.
- Multi-Document consolidator: Drag-and-drop document upload (PDF/Word) that extracts text and compiles it into a structured reference context.
- Technical Proposal Draft Engine: Instant generation of tailored proposals addressing critical Chilean bidding criteria in Markdown format.
UX & Design Process
Mapping Shape's Core Stack
Interfaced with Shape's team to define core criteria (UI/UX, React/TS, Design Systems) and structured them as system constraints for the AI scoring model.
Kanban Pipeline Design
Structured the board to display bids based on their compatibility match_score, automatically filtering out low-scoring bids to reduce manager cognitive fatigue.
Consolidating Bidding Documents
Created an intuitive sidebar drawer supporting multiple drag-and-drop document uploads, parsing them on-the-fly to construct the context package.
Interactive Proposal UX
Designed an interface that allows managers to read, refine, and edit proposal AI-generated drafts before downloading them as Markdown or Word.
Estimated results
Estimated reduction in scouting and qualification time
Estimated time to a first technical and commercial draft
Cascading AI layers (Groq, Gemini and local Ollama) so the pipeline never stops
AgileRadar demonstrates how modern web applications built with LLMs can streamline complex B2G procurement pipelines, converting days of manual document parsing into minutes of intuitive, AI-assisted operations.
Similar challenge?
Have complex software that needs clarity?
Let's talk directly—I'll tell you what makes sense to tackle first.