AI Vendor Intelligence Agent

Track every pricing change, model launch, outage and security incident across the major AI vendors — without spending hours on Twitter.

You select the AI vendors (OpenAI, Anthropic, Google, Azure AI, AWS Bedrock, Mistral, Cohere, xAI, or any other name) and a lookback window. The actor returns one consolidated vendor intelligence report per vendor — with sourced citations — covering:

• Pricing changes — $/M tokens, plan changes, free-tier shifts
• Model launches — new models, GA flips, deprecations
• Feature updates — prompt caching, function calling, vision, batch API, fine-tuning
• Incidents & outages — live status incidents and recent disruptions
• Security & compliance — disclosures, CVEs, SOC 2 / HIPAA / GDPR notes
• Competitive moves — acquisitions, partnerships, enterprise wins, funding

Output is one structured JSON record per vendor, ready to drop into a Slack alert, a weekly CIO digest, or your internal vendor scorecard.

---

Who this is for

• CIOs and IT leaders evaluating which AI vendor to standardize on
• Procurement teams tracking pricing changes mid-contract
• AI / tech consultants advising 5+ clients across vendors
• SMB owners dependent on a single AI tool who can't afford to miss an outage
• Builders who want a single weekly digest of "what changed across all major AI vendors"

---

What you get back (per vendor)

{
  "vendor": "OpenAI",
  "vendorSlug": "openai",
  "summary": "OpenAI: 2 pricing signals, 1 model launch, 3 feature updates, 0 incidents, 0 security notes, 1 competitive move in the last 30 days.",
  "lookbackDays": 30,
  "pricingChanges":   [{ "title": "...", "url": "...", "date": "...", "evidence": "..." }],
  "modelLaunches":    [{ ... }],
  "featureUpdates":   [{ ... }],
  "incidents":        [{ ... }],
  "securityNotes":    [{ ... }],
  "competitiveMoves": [{ ... }],
  "sourceUrls": [
    "https://openai.com/news/",
    "https://platform.openai.com/docs/changelog",
    "https://openai.com/api/pricing/",
    "https://status.openai.com/"
  ],
  "scrapedAt": "2026-05-16T01:23:45.000Z"
}

Each item inside a category carries the source url, a date, and a short evidence snippet so you can audit and forward without re-checking.

---

Pricing — pay per report

Event	Price	What you get
vendor-report	$2.50 per report	One complete intelligence report for one vendor, covering all six categories with sourced citations

A typical CIO weekly digest covering 5 vendors costs $12.50. A consultant tracking 10 vendors across 4 clients monthly = $100/month vs ~20 hours of manual research.

Pricing is set at the Apify Store level via Pay-Per-Event. You only pay for reports successfully produced.

---

Example use cases

1. "Alert me when any major AI vendor changes pricing or launches a new model." Schedule the actor weekly across [openai, anthropic, google, azure, mistral]; pipe the dataset into Slack / email when pricingChanges or modelLaunches is non-empty.

2. "Have we seen any incidents this week?" Run with intelligenceTypes: ["incidents"] against your production vendors and surface only the incidents array.

3. CIO weekly digest. Schedule Mondays 7am, monitor your top 5 vendors, render the summary line and modelLaunches + pricingChanges into an internal newsletter.

4. Procurement renewal prep. Run with lookbackDays: 90 two weeks before a contract renewal — get all pricing signals and competitive moves in one PDF-ready report.

5. Vendor diligence for an investment. Run with lookbackDays: 180 across the candidate's stack to see how often their AI vendors change pricing or have outages.

---

Input

{
  "vendors": ["openai", "anthropic", "google", "mistral"],
  "intelligenceTypes": [
    "pricing", "model_launches", "feature_updates",
    "incidents", "security", "competitive_moves"
  ],
  "lookbackDays": 30,
  "maxItemsPerCategory": 5,
  "outputFormat": "report"
}

Field	Type	Default	Description
vendors	string[]	— (required)	Vendor slugs or freeform names. Built-in slugs: openai, anthropic, google, azure, aws, mistral, cohere, xai. Unknown names fall back to a web-search-driven harvester.
intelligenceTypes	string[]	all 6 categories	Filter which categories to include in the report.
lookbackDays	int	30	How far back to consider changelog / blog entries.
maxItemsPerCategory	int	5	Cap per category per vendor.
outputFormat	string	report	report = full structured report. compact = headline summary + counts only.

---

Built-in vendors

Slug	Vendor	Sources
openai	OpenAI	News, pricing, platform changelog, status
anthropic	Anthropic	News, pricing, API release notes, status
google	Google AI / Gemini / Vertex	Blog, Gemini API changelog, Cloud pricing, Cloud status
azure	Microsoft Azure OpenAI / AI	Pricing, blog, "what's new", Azure status
aws	AWS Bedrock	Pricing, what's new, doc history, AWS health
mistral	Mistral AI	News, pricing, changelog, status
cohere	Cohere	Blog, pricing, changelog, status
xai	xAI (Grok)	News, API page, docs changelog, status

Pass a freeform name (e.g. "Together AI" or "Replicate") and the actor falls back to a web-search-driven harvester that still produces a categorized report — handy for less common vendors.

---

How the report is built

1. For each known vendor, fetches the vendor's news / changelog / blog, pricing pages, and status page.
2. Extracts article titles, dates, and snippets from each page (resilient to most blog / changelog layouts).
3. Classifies every snippet into one of six categories with a heuristic pattern set.
4. Deduplicates within category, caps to maxItemsPerCategory, attaches source URLs and dates.
5. Renders a one-line summary so the report is glanceable.
6. Emits one PPE vendor-report event so usage is tracked per real report produced.

For unknown vendors, the actor runs 5 targeted Google queries (pricing change, model launch, outage, security, partnership) and classifies the results the same way.

---

Competitive comparison

	This actor	Generic AI news scrapers	Vendor-specific changelog scrapers	Manual research
Categorized output (pricing / launches / incidents / security / competitive)	✅	❌	❌	✅ (slow)
Multi-vendor in one run	✅	partial	❌ (one per vendor)	❌
Incident / status detection	✅	❌	❌	❌
Pay-per-result pricing	✅ ($2.50/report)	usually per-run	usually per-run	n/a
Built-in vendor registry	✅ (8 vendors out of the box)	❌	❌	n/a
Web-search fallback for unknown vendors	✅	❌	❌	✅

---

Operational notes

• Built on apify/actor-node-playwright-chrome:20, headless Chromium.
• Resilient to most cookie / consent overlays (Accept / Got it / Dismiss).
• Each vendor processed independently — one vendor failure does not abort the run.
• Default run: 4 GB memory, 15 min timeout — enough for 8–10 vendors with full reports.
• The actor emits one PPE event per successfully produced report; failed vendor records do not charge.

---

Roadmap

• Diff mode (only show what changed vs the last run)
• Slack / webhook output template
• Daily digest combining all selected vendors into a single Markdown report
• Subscriptions to specific change types (pricing-only, incidents-only) at a discounted rate

---

Built by Johnson AI Consulting — part of the Apify Revenue Initiative.