Dashboard/Scout/wiki/use-cases/index.md
Scout

wiki/use-cases/index.md

Scout - Use-Case Catalog

Last Updated: 2026-05-14

How To Read This

Each use case defines:

  • What the user asks for.
  • What Scout should produce.
  • Provider strategy.
  • Test scenarios.
  • Example websites for validation.
  • Risks and pass/fail criteria.

This catalog should become the test roadmap. A use case is not "real" until it has fixtures, live smoke tests, schemas, and documentation.

UC-01 - Product Catalog To Algolia Records

User Story

As an ecommerce/search user, I want Scout to extract products from retailer websites and prepare Algolia-ready product records that can power search, category pages, and PDP experiences.

Example prompt:

Find the top categories for Estee Lauder and get 10 products each with all attributes, price, color, review, everything. Then get top 10 best sellers as well.

Output Records

ProductRecord

Important fields:

  • objectID
  • brand
  • name
  • subtitle
  • url
  • category
  • subcategory
  • product_type
  • price
  • sale_price
  • currency
  • rating
  • review_count
  • colors
  • sizes
  • variants
  • badges
  • description
  • benefits
  • ingredients
  • images
  • availability
  • source
  • completeness_score

Provider Strategy

Skill mode:

  1. WebSearch for category or PDP discovery.
  2. WebFetch category/listing pages.
  3. WebFetch PDPs for enrichment.
  4. In-app browser/session if WebFetch is incomplete or blocked.
  5. Crawl4AI fallback for normal sites.

Standalone mode:

  1. Explicit URLs or seed config.
  2. Crawl4AI.
  3. CDP/profile.
  4. Saved HTML.
  5. Proxy/provider later.

Test Scenarios

  1. Known PDP URL: extract one complete product from a single page.
  2. Category page: extract first 10 product cards from a category listing.
  3. Listing + PDP enrichment: merge listing fields with detail page fields.
  4. Best sellers: extract top 10 best sellers from a known best-sellers page.
  5. Hard-site fallback: local Crawl4AI blocked, WebFetch or browser/session succeeds, blocked artifact records the path.

Example Websites

Site Example URL Why it matters
Estee Lauder https://www.esteelauder.com/products/681/product-catalog/skin-care Hard-site beauty retailer with rich product attributes and blocking behavior.
Lacoste https://www.lacoste.com/us/lacoste/men/clothing/button-down-shirts/ Apparel category with sizes/colors and product cards.
Nike https://www.nike.com/w/mens-running-shoes-37v7jz8ucbuznik1zy7ok Large retailer, dynamic category filtering, product variants.
Sephora https://www.sephora.com/shop/skincare Marketplace-style beauty catalog with brands, ratings, prices.
Patagonia https://www.patagonia.com/shop/mens-shirts Apparel catalog, sustainability/product metadata.

Pass Criteria

  • At least 10 records per target category when visible.
  • No category/navigation links misclassified as products.
  • PDP enrichment improves completeness score.
  • Output can be indexed into Algolia without hand-editing.

UC-02 - Company About And Leadership Intelligence

User Story

As a prospect researcher, I want Scout to fetch company about and leadership pages and extract company facts, executive names, titles, bios, and social/profile links.

Example prompt:

Get company intel for Nike, including about page, leadership, executives, bios, and LinkedIn URLs if available.

Output Records

CompanyPageRecord, ExecutiveRecord

Important fields:

  • company name
  • website
  • page type
  • executive name
  • title
  • department/function
  • bio
  • headshot URL
  • LinkedIn URL or social/profile URL when available
  • board membership flag
  • source URL
  • confidence

Provider Strategy

  1. WebSearch for official company/about/leadership URLs.
  2. WebFetch official pages.
  3. Crawl4AI map/scrape for site sections.
  4. Browser/session only if JS or blocking prevents content.
  5. Optional enrichment from third-party profile providers, marked as non-official source.

Test Scenarios

  1. Official leadership page with visible executive cards.
  2. About page with company mission and history.
  3. Leadership page where bios are behind modal or "View bio" links.
  4. Company with no official leadership page, requiring fallback to investor filings or newsroom.
  5. Extract source provenance and distinguish official from third-party sources.

Example Websites

Site Example URL Why it matters
Salesforce https://www.salesforce.com/company/leadership/ Official leadership page with executive and board sections.
Adobe https://www.adobe.com/about-adobe/leaders.html Official executive profiles.
Nike https://about.nike.com/ Corporate about surface, leadership may need investor filings/newsroom enrichment.
Shopify https://www.shopify.com/about Company overview and executive/board data may require mixed sources.
Microsoft https://www.microsoft.com/en-us/about/ Large public company with multiple official company pages.

Pass Criteria

  • Extracts executive name and title with high precision.
  • Does not invent LinkedIn URLs.
  • Marks missing fields as missing instead of hallucinating.
  • Separates official source facts from third-party enrichment.

UC-03 - Investor Relations And Financial Document Intelligence

User Story

As an investor/prospect researcher, I want Scout to discover investor pages, annual reports, 10-K/10-Q filings, earnings releases, decks, and financial metrics.

Example prompt:

Get investor intelligence for Nike: annual reports, 10-K/10-Q, investor presentations, revenue numbers, risk factors, and executive quotes.

Output Records

InvestorDocumentRecord, FinancialMetricRecord, ExecutiveQuoteRecord, RiskFactorRecord

Important fields:

  • company
  • ticker
  • document type
  • document URL
  • filing date
  • fiscal period
  • revenue
  • segment revenue
  • operating income
  • margin
  • guidance/outlook
  • risk factors
  • strategy themes
  • executive quotes
  • source page/PDF

Provider Strategy

  1. WebSearch for official investor relations pages.
  2. WebFetch IR pages.
  3. PDF/document parser for annual reports, 10-K, 10-Q, presentations.
  4. SEC/EDGAR provider for public companies.
  5. Financial API/provider enrichment where available.

Test Scenarios

  1. Find annual report from official IR page.
  2. Extract document inventory from IR financial documents page.
  3. Parse a PDF annual report and extract risk factors/MD&A/revenue.
  4. Extract investor presentation deck metadata and key claims.
  5. Cross-check official IR page against SEC filing source.

Example Websites

Site Example URL Why it matters
Nike IR https://investors.nike.com/ Retail/apparel public company, annual reports and SEC filings.
Adobe IR https://www.adobe.com/investor-relations.html Software company with financial documents and presentations.
Salesforce IR https://investor.salesforce.com/financials/annual-reports/default.aspx SaaS public company with annual reports.
Apple IR https://investor.apple.com/sec-filings/ High-quality official SEC filing surface.
Microsoft IR https://www.microsoft.com/en-us/investor/annual-reports Official annual reports and long-form financial docs.

Pass Criteria

  • Document URLs are official or clearly labeled third-party.
  • Extracted numbers include period, units, and source.
  • PDF extraction is deterministic on saved fixtures.
  • Scout never reports naked numbers without provenance.

UC-04 - Careers And Job Posting Extraction

User Story

As a job seeker or job-applicator tool, I want Scout to find jobs at target companies for a role, normalize job postings, and return apply URLs.

Scout should also support a discovery-first job hunter flow. In that flow, the user enters the kinds of jobs they want, salary range, target titles, location or remote preference, seniority, and role keywords. Scout first finds companies in that area, returns candidate company names and careers pages for approval, and then monitors those companies for matching open roles.

Example prompt:

Find job openings at Adobe and Salesforce for product marketing manager roles.

Discovery-first prompt:

I am looking for AI product marketing, developer advocate, or solutions engineering roles, remote or New York, salary above $160k. Find companies hiring in this area, then monitor those companies daily for matching jobs.

Output Records

JobPostingRecord

Additional records:

  • JobSearchProfile
  • TargetCompanyRecord
  • JobDeltaRecord

Important fields:

  • search profile:
  • target titles
  • role families
  • keywords
  • salary range
  • location/remote preference
  • seniority
  • target industries
  • required skills
  • excluded roles/terms
  • target company:
  • company
  • website
  • careers URL
  • LinkedIn URL when discoverable
  • industry/category
  • reason matched
  • confidence
  • company
  • title
  • job ID
  • location
  • remote/hybrid/onsite
  • department
  • employment type
  • salary range
  • description
  • responsibilities
  • qualifications
  • apply URL
  • ATS platform
  • posted date when available
  • source

Provider Strategy

  1. WebSearch for company careers page.
  2. Discovery-first mode: WebSearch for companies matching the role/industry criteria.
  3. Return reviewable target company candidates.
  4. Detect ATS provider for approved companies.
  5. Use ATS-specific provider when possible.
  6. Crawl4AI/browser fallback for custom career pages.
  7. Save blocked/ATS-required evidence if extraction cannot proceed.

Test Scenarios

  1. Official careers landing page discovery.
  2. Job intake dialogue creates a reusable JobSearchProfile.
  3. Company discovery returns a candidate target-company list with reason matched.
  4. User-approved target companies are searched for matching jobs.
  5. Search within a company jobs page for a target role.
  6. ATS detection: Workday, Greenhouse, Lever, Ashby, SmartRecruiters.
  7. Job detail enrichment from apply page.
  8. Multi-company comparison for the same role query.
  9. Daily run detects new, changed, and closed jobs.

Example Websites

Site Example URL Why it matters
Adobe Careers https://careers.adobe.com/us/en/ Large enterprise careers page with searchable jobs.
Salesforce Careers https://www.salesforce.com/company/careers.html Official careers page, job search flow.
Nike Careers https://careers.nike.com/jobs Large retail/apparel employer with many roles.
Workday Careers https://www.workday.com/careers Important ATS/vendor and employer example.
Target Careers https://corporate.target.com/careers/ Retail employer, Workday-style flows.

Pass Criteria

  • Role query returns relevant postings only.
  • Apply URLs are direct and official.
  • ATS platform is detected.
  • Scout does not attempt to submit applications.
  • Missing salary or dates are marked missing.

UC-05 - Algolia Prospect Research Evidence Collection

User Story

As an Algolia seller or researcher, I want Scout to collect evidence about a prospect's company, ecommerce/search experience, investor priorities, hiring signals, and product catalog.

Example prompt:

Get prospect research for Nike: about, leadership, investor, hiring, product catalog, ecommerce signals, and recent news.

Output Records

Composite evidence package:

  • CompanyPageRecord
  • ExecutiveRecord
  • InvestorDocumentRecord
  • JobPostingRecord
  • ProductRecord
  • NewsArticleRecord
  • SearchExperienceRecord
  • CompanySocialRecord
  • EvidenceSummary

Required discovery targets:

  • canonical company website
  • company LinkedIn page
  • official company socials such as X/Twitter, YouTube, Instagram, TikTok, or Facebook when relevant
  • about/company overview page
  • leadership/executive page
  • executive names, titles, bios
  • executive LinkedIn/profile URLs when available
  • investor relations page
  • careers page
  • newsroom/press page
  • ecommerce/product catalog pages when relevant

Provider Strategy

  1. WebSearch for official sources.
  2. WebFetch known pages.
  3. Crawl4AI for site mapping and normal pages.
  4. PDF provider for investor docs.
  5. ATS provider for jobs.
  6. Product provider for ecommerce.
  7. Social/news provider where available.

Test Scenarios

  1. Build a source map for one prospect.
  2. Discover canonical website and official company LinkedIn/social profiles.
  3. Discover executives and executive LinkedIn/profile URLs where available.
  4. Extract at least one record from each evidence class.
  5. Detect ecommerce/product catalog surfaces.
  6. Detect hiring signals for search, ecommerce, data, AI, personalization, merchandising.
  7. Produce a provenance-backed evidence bundle for downstream PRISM synthesis.

Example Websites

Prospect Example URL Why it matters
Nike https://www.nike.com/, https://investors.nike.com/, https://careers.nike.com/jobs Large retail/ecommerce public company.
Sephora https://www.sephora.com/, https://www.sephora.com/beauty/careers Beauty ecommerce and search-heavy experience.
Lululemon https://shop.lululemon.com/, https://corporate.lululemon.com/ Retail product catalog plus corporate/investor pages.
Best Buy https://www.bestbuy.com/, https://investors.bestbuy.com/ Electronics retailer with catalog/search complexity.
Williams-Sonoma https://www.williams-sonoma.com/, https://ir.williams-sonomainc.com/ Multi-brand commerce and investor context.

Pass Criteria

  • Evidence bundle separates raw records from analysis.
  • Every claim can link back to source records.
  • No final sales narrative is generated without record provenance.

UC-06 - Documentation And Knowledge Base Extraction

User Story

As a developer or support tool builder, I want Scout to extract documentation pages into searchable records.

Example prompt:

Index the Algolia Search API docs into clean records for an internal assistant.

Output Records

DocumentationPageRecord

Important fields:

  • product
  • doc title
  • section
  • URL
  • headings
  • summary
  • code snippets
  • API methods
  • parameters
  • version
  • last updated when available
  • source

Provider Strategy

  1. Sitemap/llms.txt when available.
  2. WebFetch known docs pages.
  3. Crawl4AI for docs site traversal.
  4. Markdown extraction and code block preservation.
  5. Optional docs-specific provider such as Context7 when host supports it.

Test Scenarios

  1. Known docs URL extraction.
  2. Multi-page docs crawl with dedupe.
  3. Code block preservation.
  4. API parameter extraction.
  5. Algolia-ready docs record output.

Example Websites

Site Example URL Why it matters
Algolia Docs https://www.algolia.com/doc/api-reference Direct relevance to Algolia and API records.
Stripe Docs https://docs.stripe.com/api API reference with rich structured docs.
Shopify Dev Docs https://shopify.dev/docs/api Commerce/developer docs and versioned APIs.
Vercel Docs https://vercel.com/docs Modern docs site with framework/platform sections.
OpenAI Docs https://platform.openai.com/docs AI platform docs with frequent changes.

Pass Criteria

  • Code examples are preserved.
  • Docs pages are chunked without losing hierarchy.
  • URLs remain canonical.
  • Records are useful for search and RAG.

UC-07 - Newsroom And Press Monitoring

User Story

As a researcher, I want Scout to collect recent official company news, press releases, product launches, leadership changes, and strategic announcements.

Example prompt:

Get Nike's recent official news and product launch signals.

Output Records

NewsArticleRecord, CompanySignalRecord

Important fields:

  • title
  • date
  • author/source
  • company
  • article URL
  • tags/topics
  • mentioned products
  • mentioned executives
  • signal type
  • summary
  • source

Provider Strategy

  1. WebSearch for official newsroom.
  2. WebFetch newsroom index and articles.
  3. RSS feed if present.
  4. Crawl4AI for archive pagination.
  5. Date filtering and duplicate detection.

Test Scenarios

  1. Discover official newsroom page.
  2. Extract article list with dates.
  3. Fetch and summarize individual articles.
  4. Classify signal type: product launch, leadership, financial, partnership, sustainability.
  5. Filter to last 30/60/90 days.

Example Websites

Site Example URL Why it matters
Nike Stories https://about.nike.com/en/newsroom Official brand/product/corporate news.
Adobe News https://news.adobe.com/ Product and leadership announcements.
Salesforce News https://www.salesforce.com/news/ Enterprise software news and earnings releases.
Shopify News https://www.shopify.com/news Commerce platform announcements.
Microsoft Blog https://blogs.microsoft.com/ Large company strategic announcements.

Pass Criteria

  • Only recent articles when date filter is requested.
  • Article dates are parsed and normalized.
  • Signal classification is explainable and sourced.

UC-08 - Social Signal Normalization

User Story

As a prospect researcher, I want Scout to normalize social posts and company social signals into structured records that can be analyzed downstream.

Example prompt:

Get recent LinkedIn and X posts for Nike about ecommerce, AI, personalization, and product launches.

Output Records

SocialSignalRecord

Important fields:

  • platform
  • account/company
  • post URL
  • author
  • timestamp
  • text
  • media URLs
  • engagement metrics when available
  • topics
  • signal type
  • source provider

Provider Strategy

Scout should not rely on raw Crawl4AI for LinkedIn/X.

Preferred providers:

  • Official APIs where available.
  • Host connectors.
  • Apify-like actors/providers.
  • Browser/session capture only when user explicitly provides access.
  • Saved exports.

Test Scenarios

  1. Ingest saved LinkedIn/X export into records.
  2. Normalize provider output from a social scraper.
  3. Classify posts by topic and signal type.
  4. Deduplicate reposts or syndicated posts.
  5. Preserve source platform and provider provenance.

Example Targets

Target Example URL Why it matters
Nike LinkedIn https://www.linkedin.com/company/nike/ Company social posts, hiring, brand and product signals.
Adobe LinkedIn https://www.linkedin.com/company/adobe/ Enterprise software and AI/product announcements.
Salesforce LinkedIn https://www.linkedin.com/company/salesforce/ Enterprise buyer and platform messaging.
Nike X https://x.com/Nike Brand/product signal surface.
Salesforce X https://x.com/salesforce Enterprise news and campaigns.

Pass Criteria

  • Scout labels provider and access method clearly.
  • Scout does not claim direct LinkedIn/X crawling support unless validated.
  • Social records include enough provenance for audit.

UC-09 - Store Locator And Location Data

User Story

As a local search or commerce user, I want Scout to extract store/location data from locator pages.

Example prompt:

Get Nike stores in New York with address, hours, phone, and services.

Output Records

LocationRecord

Important fields:

  • brand
  • store name
  • address
  • city
  • state/region
  • postal code
  • country
  • latitude/longitude when available
  • phone
  • hours
  • services
  • store URL
  • source

Provider Strategy

  1. WebSearch for locator page.
  2. Browser/session or Crawl4AI for dynamic locator pages.
  3. Network/API capture if locator loads JSON.
  4. Manual seed city/zip inputs.

Test Scenarios

  1. Known store locator URL.
  2. Search by city/zip.
  3. Extract first 10 store records.
  4. Parse hours and services.
  5. Handle map-based dynamic pages.

Example Websites

Site Example URL Why it matters
Nike Stores https://www.nike.com/retail Retail locator with brand/store metadata.
Sephora Stores https://www.sephora.com/happening/stores/sephora-near-me Beauty retailer locator.
Apple Stores https://www.apple.com/retail/ Structured retail location pages.
Starbucks Store Locator https://www.starbucks.com/store-locator Dynamic map/location flow.
Target Store Locator https://www.target.com/store-locator/find-stores Retail locator with services.

Pass Criteria

  • Addresses are parsed into components.
  • Dynamic locator limitations are explicit.
  • Records do not confuse corporate pages with store records.

UC-10 - Generic Page To Structured Facts

User Story

As an analyst, I want Scout to fetch any useful web page and extract structured facts according to a task-specific schema.

Example prompt:

Fetch this about page and extract company facts, products, markets, and leadership mentions.

Output Records

GenericPageFactRecord

Important fields:

  • page URL
  • page title
  • page type
  • fact type
  • fact value
  • evidence text
  • confidence
  • source provider

Provider Strategy

  1. Explicit URL.
  2. WebFetch in skill mode.
  3. Crawl4AI in standalone mode.
  4. Browser/session for blocked or dynamic pages.
  5. LLM/schema extraction when deterministic parsing is not enough.

Test Scenarios

  1. Known about page.
  2. Known investor page.
  3. Known careers page.
  4. Known FAQ page.
  5. Arbitrary page with user-provided schema.

Example Websites

Site Example URL Why it matters
Nike About https://about.nike.com/ Corporate overview.
Adobe About https://www.adobe.com/about-adobe.html Company facts and mission.
Salesforce About https://www.salesforce.com/company/ Enterprise company positioning.
Shopify About https://www.shopify.com/about Company overview and product/business model.
Algolia About https://www.algolia.com/about/ Relevant internal/prospect tooling example.

Pass Criteria

  • User-provided schema drives extraction.
  • Evidence text is retained.
  • Scout marks low-confidence or missing fields.