AWS Innovation Monitor¶

Capability Slug: aws_innovation_monitor Plan Tier: Command Credit Cost: 10 / 15 / 20 / 25 / 35 credits per run (varies by operation type) Category: Intelligence & Attribution Default operation: discover_updates Simulation: Supported Pairs with: Productization Recommendation Engine

Overview¶

Partners and the PDMs assigned to them miss AWS announcements. By the time a Solutions Architect sees a release at re:Invent or in a weekly announcement digest, weeks have passed and a competitor may have already incorporated the new capability into their product. This capability closes that gap by continuously scanning AWS RSS feeds, scoring each release against a partner's product profile, and surfacing the ones that matter — without anyone having to read the firehose.

Innovation Monitor is the ingestion and scoring layer. It extracts AWS service mentions, persists each update as an AwsInnovationUpdate record with a vector embedding, and produces relevance scores using the same weighted algorithm as co-sell matching. The companion capability — Productization Recommendation Engine — sits on top of this corpus and turns surfaced updates into product roadmaps.

The five operation modes¶

Pick the mode based on whether you're scanning, analyzing, or processing feeds.

Operation	Credits	What it does	When to use
`discover_updates` (default)	15	Scan recent updates from configured feeds, score against partner profile, return top N relevant ones	Weekly partner check-in: "what AWS announcements should we discuss?"
`analyze_relevance`	20	Deep Claude Sonnet 4.5 analysis of one specific update's impact on a partner — product impact, integration opportunities, competitive advantage, risk of ignoring	One AWS announcement is interesting; you want a structured impact assessment for the partner
`generate_recommendations`	25	Synthesize multiple relevant updates into productization suggestions (feature enhancements, integrations, pricing, positioning, listing changes)	Quarterly product planning conversation with a partner
`batch_process_feeds`	35	Pull all configured feeds, dedupe against existing records, generate Bedrock embeddings, persist to `aws_innovation_updates`	Initial corpus build or scheduled maintenance refresh
`find_similar`	10	Semantic search across already-ingested updates using vector similarity	"We're working on observability — what AWS announcements relate?"

The relevance scoring algorithm¶

Every update gets a score 0-100 with three weighted dimensions:

Weight	Dimension	What it measures
40%	Product Fit	Does this AWS update align with the partner's product categories?
35%	Market Alignment	Does the target market match the partner's customer base?
25%	Timing / Urgency	How recent is the announcement? How time-sensitive is the opportunity?

Default surfacing threshold (min_relevance) is 50. Updates below that don't appear in discover_updates results unless you explicitly lower it. The threshold and weights are aligned with the Co-Sell Partner Discovery & Matching algorithm so cross-capability scoring stays consistent.

Configured RSS feeds¶

From config/marketplace-feeds.php, in priority order:

Feed	URL	Priority	Category
AWS Marketplace Blog	`https://aws.amazon.com/blogs/awsmarketplace/feed/`	high	marketplace
AWS Partner Network Blog	`https://aws.amazon.com/blogs/apn/feed/`	high	partner
AWS What's New	`https://aws.amazon.com/new/feed/`	medium	announcements
AWS Startups Blog	`https://aws.amazon.com/blogs/startups/feed/`	medium	startups
TechCrunch Cloud (industry)	`https://techcrunch.com/category/cloud/feed/`	low	industry
The New Stack (industry)	`https://thenewstack.io/feed/`	low	industry

Pass feed_sources as an array to scope a run to specific feeds. The default discover_updates set is ['aws_whats_new', 'aws_marketplace_blog', 'aws_apn_blog'].

What `discover_updates` returns¶

{
  "success": true,
  "operation_type": "discover_updates",
  "credits_used": 15,
  "updates": [
    {
      "title": "...",
      "link": "https://aws.amazon.com/about-aws/whats-new/...",
      "published_at": "2026-04-28T15:00:00Z",
      "source": "aws_whats_new",
      "relevance_score": 87,
      "score_breakdown": {
        "product_fit": 92,
        "market_alignment": 85,
        "timing_urgency": 80
      },
      "summary": "...",
      "aws_services": ["Amazon Bedrock", "AWS Lambda"]
    }
  ],
  "total_scanned": 0,
  "relevant_count": 0,
  "partner_name": "...",
  "date_range": { "from": "...", "to": "..." },
  "next_steps": [ /* deterministic action checklist */ ]
}

When the result set is small (≤5 updates), each update is enhanced with quick insights — short AI-generated context suggesting what to do with the announcement. Larger result sets return without the insight enhancement to keep latency low.

What `batch_process_feeds` returns¶

{
  "success": true,
  "credits_used": 35,
  "feeds_scanned": 4,
  "updates_processed": 0,
  "embeddings_generated": 0,
  "errors": [],
  "processed": [ /* per-feed counts */ ]
}

The embeddings are stored on the AwsInnovationUpdate model and used by:

This capability's find_similar mode for semantic search
Productization Recommendation for partner-update matching
Other capabilities that opt into the cross-capability intelligence

batch_process_feeds is idempotent — articles already present (matched by source_url) are skipped. Run it on a schedule (daily or weekly) to keep the corpus fresh.

Required and optional parameters¶

Parameter	Required	Description
`operation_type`	Yes	One of `discover_updates`, `analyze_relevance`, `generate_recommendations`, `batch_process_feeds`, `find_similar`
`partner_id`	Required for relevance modes	Partner record to score against
`feed_sources`	No	Array of feed slugs. Default is the AWS-only set for `discover_updates`; null = all configured feeds for `batch_process_feeds`
`days_back`	No	Lookback window. Default 30 days (`discover_updates`) or 7 days (`batch_process_feeds`)
`limit`	No	Max number of results. Default 10
`min_relevance`	No	Minimum overall score for inclusion. Default 50
`update_id` / `update_data`	One required for `analyze_relevance`	Either the persisted update ID or an inline payload
`update_ids`	Optional for `generate_recommendations`	Specific updates to base recommendations on; if empty, the capability auto-discovers top-5 relevant updates
`focus_area`	Optional for `generate_recommendations`	One of `features`, `integrations`, `pricing`, `positioning`, `all` (default)
`generate_embeddings`	Optional for `batch_process_feeds`	Default `true`

When to use it (vs. its companion)¶

Scenario	Recommended capability + mode
"Scan feeds and surface what matters this week"	`aws_innovation_monitor` → `discover_updates`
"Analyze this one announcement's impact on a partner"	`aws_innovation_monitor` → `analyze_relevance`
"Build the partner-update corpus for the first time"	`aws_innovation_monitor` → `batch_process_feeds`
"Find updates similar to a topic we're already exploring"	`aws_innovation_monitor` → `find_similar`
"Match all relevant updates to a partner's products"	Productization Recommendation → `match_updates`
"Build a 12-month product roadmap"	Productization Recommendation → `generate_roadmap`
"What's the competitive risk of ignoring these?"	Productization Recommendation → `competitive_analysis`
"Generate a step-by-step implementation plan"	Productization Recommendation → `implementation_plan`
"Match updates across all partners (PDM batch)"	Productization Recommendation → `bulk_partner_match`

API access¶

The capability is exposed through the agent runtime as aws_innovation_monitor. Partner API keys scoped with manage_content and view_marketing_analytics plus the agent capability slug can run all five operation modes. The simulation interface is available on the same key. See Partner API Capabilities.