Skip to content
Page 8 of 10

Challenge 1: Federal Vendor Intelligence Tool

What You're Starting With

From the lesson, you've got:

  • The Three Pillars (Scope, Intent, Structure): use them every time you prompt
  • User stories with acceptance criteria: your plan for each piece of the vendor tool
  • The Explore, Plan, Implement, Verify workflow: your process for building iteratively
  • Right-sizing prompts: small asks in plain English, medium asks as user stories, big asks broken into smaller pieces

Some of you already started building during the lesson. In the mob session, your team created a small component (a vendor profile card, an agency breakdown, or a business type indicator). You can choose to build from there or start fresh.

Your AI chat tool already knows a lot about federal contracting, USAspending, SAM.gov, and acquisition terminology. Ask it. Explore. Let it help you understand what should go into a vendor intelligence tool. That is the Explore step of your workflow.

The Challenge

Build a Federal Vendor Intelligence Tool: a polished, interactive web page that consolidates the data acquisition professionals need when researching a federal contractor.

You are building this entirely by talking to an AI chat tool. No code editor, no terminal. Just you, your team, and a conversation with AI. Everything you learned in the lesson (the Three Pillars, user stories, the Explore, Plan, Implement, Verify workflow) is where you put it to work.

Who You're Building For

Acquisition professionals researching a potential vendor. They need to assess whether this vendor is someone the government can buy from, should buy from, and how. The questions stack up fast: How much federal work has this company done, and with which agencies? What NAICS codes do they operate in, and are they small under those codes? Do they have an existing contract vehicle, or would this be open-market? Are they registered in SAM? Any exclusions or debarments? What socioeconomic designations do they hold for set-aside eligibility? Is there enough competition in this space to support a set-aside?

Today, answering those questions means checking USAspending.gov for award history, SAM.gov for registration and eligibility, the exclusions database for debarment records, and often GSA eLibrary for contract vehicle information. At least three or four separate systems, each with its own search interface, data format, and conventions. Your tool puts that data in one place: clear, structured, and actionable.

What to Build

Items are listed in priority order. If time is tight, focus on the items near the top first.

For this challenge, you are building with fake sample data (not pulling from a live data feed yet). Your job is to get the structure, layout, and logic right so that when real data comes in later, the tool just works.

  • A vendor profile card showing sample data: company name, total federal award amount, number of transactions, and business type (small business, large business, 8(a), etc.)
  • An agency breakdown: top 3-5 awarding agencies with dollar amounts, showing which parts of government the vendor works with
  • A NAICS code display: what industries the vendor operates in, with codes and descriptions
  • Visual design that feels like an actual intelligence dashboard, not raw text; something with layout, color coding, and structure you would trust for a vendor assessment

These are options for teams that finish the baseline capabilities. Your team can also define your own stretch goals based on what interests you.

  • Add a vendor search bar so the user can look up different companies
  • Create a SAM registration status card showing sample UEI, CAGE code, registration status, and expiration date
  • Add an exclusion check indicator: green "No Active Exclusions" or red "EXCLUDED" badge
  • Add a spending trend section showing sample year-over-year award amounts
  • Include a set-aside eligibility summary: which designations does the vendor hold (8(a), HUBZone, SDVOSB, WOSB) and are they current

In Your AI Assistant

If you didn't get a visual prototype: AI tools respond differently each time, even from the same prompt. Try these in order:

  1. Look for a Preview, Play, or Run button in your tool's interface. Claude.ai, ChatGPT, and Gemini can all render live previews, but you may need to click a button to switch from code view to preview. Some tools require you to be logged in.
  2. Look for Canvas mode. Some tools have a side-by-side view where you chat on one side and see your prototype on the other. In Gemini, click the + button below the prompt box and select Canvas. In ChatGPT, look for a Preview or Canvas option.
  3. Did AI give you code to install, a download link, or setup instructions instead of a visual page? Ask: "I don't want to download or install anything. Can you turn this into an interactive prototype I can see and use right here in our conversation?"
  4. Got HTML code but no live preview? Click Download code above the code block, or copy the code into a plain text editor (like Notepad or TextEdit), save it with a .html extension, and open it in your browser.
  5. Still stuck? Start a new conversation, or switch to a teammate's machine. Don't spend too much time troubleshooting.

Tips

  • Start with one section. Do not try to build the whole tool in your first prompt. Pick the vendor profile card or the agency breakdown and get that right first. Then add the next piece.
  • Use the workflow. Explore, Plan, Implement, Verify. Write a user story for each section before you ask AI to build it. Check the result against your acceptance criteria before moving on.
  • Quick user story template: "As an acquisition professional researching a vendor, I want [what you're building] so that [why it helps]. Given [a specific condition], when [something happens], then [what the user should see]." For example: "As a contracting officer, I want a profile card showing total federal awards so I can assess the vendor's scale. Given the data shows $49.9B in total awards, when I view the card, it displays '$49.9B' with 'Large Business' as the business type."
  • Verify criterion by criterion, then be specific. Do not eyeball the whole page and say "looks good" or "make it better." Go through each acceptance criterion. Pass or fail. When one fails, say exactly what is wrong: "The award amount shows as a raw number. Format it as currency with B/M suffixes." That specific feedback is what gets you a fix, not a re-roll.
  • If your conversation gets long, start fresh. Remember the oxygen tank. Context windows fill up. If AI's responses start feeling off after many exchanges, open a new conversation and paste in what you want to keep building from. You will be surprised how much better it works.

Go build. That is the brief. Spend the rest of this session block working on your challenge with your team. Your Facilitator will let you know when it is time for the Reflection.