Productization Recommendation Engine¶

Capability Slug: productization_recommendation Plan Tier: Command Credit Cost: 15 / 20 / 25 / 30 / 40 credits per run (varies by operation type) Category: Intelligence & Attribution Default operation: match_updates Simulation: Supported Pairs with: AWS Innovation Monitor

Overview¶

Solutions Architects keep up with new AWS releases, but only for the largest few partners on their book. PDMs and partners themselves often miss announcements from the prior month or quarter — and by the time they catch up, a competitor has already shipped against the new capability. The Productization Recommendation Engine takes the AWS update corpus that AWS Innovation Monitor maintains and turns it into something every partner can consume: matched recommendations, prioritized roadmaps, competitive risk analyses, and implementation plans.

This is the matching and recommendation layer. Innovation Monitor provides the corpus; this capability does the partner-specific reasoning over it.

The five operation modes¶

Operation	Credits	What it does	When to use
`match_updates` (default)	15	Match recent AWS updates to a partner's product profile, return top recommendations sorted by score with quick actions	Weekly partner check-in: "what should we discuss?"
`generate_roadmap`	25	Turn matched updates into an immediate / short / medium / long-term roadmap with quick-wins and high-impact initiatives	Quarterly product planning conversation
`competitive_analysis`	20	Analyze the risk of not acting on these updates, with SA-bias alerts and competitor-landscape context	Defending an investment decision; pushing back on "we don't need that"
`implementation_plan`	30	Generate detailed implementation guidance for a specific recommendation — phases, dependencies, resources, success metrics	Engineering kickoff after a recommendation is approved
`bulk_partner_match`	40	Run `match_updates` across every partner in your portfolio	PDM portfolio review: "which partners should I prioritize this week?"

The matching algorithm¶

Matches AWS updates to partner products on three weighted dimensions, aligned with Co-Sell Partner Discovery & Matching:

Weight	Dimension	What it measures
40%	Product Fit	How well the AWS update aligns with the partner's product categories
35%	Market Alignment	Target market overlap
25%	Implementation Feasibility	Technical complexity and resource requirements (not just "does it fit", but "can they ship it")

Threshold:

RECOMMENDATION_THRESHOLD = 55 — minimum score to surface a recommendation
MAX_RECOMMENDATIONS_PER_BATCH = 10 — capped per run

What `match_updates` returns¶

{
  "success": true,
  "credits_used": 15,
  "matches": [
    {
      "update_id": 1234,
      "title": "...",
      "match_score": 82,
      "score_breakdown": {
        "product_fit": 88,
        "market_alignment": 80,
        "feasibility": 75
      },
      "aws_services": ["Amazon Bedrock", "AWS Lambda"],
      "quick_actions": [ /* deterministic suggestions */ ],
      "rationale": "..."
    }
  ],
  "next_steps": [ /* checklist tied to the matches */ ],
  "competitive_alert": {
    "urgency": "high|medium|low",
    "summary": "..."
  }
}

The competitive_alert is generated only when at least one match has a high score — it surfaces the urgency to act before competitors do.

What `generate_roadmap` returns¶

A four-bucket time-horizon roadmap (always returned, regardless of which time_horizon you pass):

Bucket	Window
`immediate`	0-30 days
`short_term`	30-90 days
`medium_term`	90-180 days
`long_term`	180-365 days

Plus three cross-cutting lists:

quick_wins — high-impact, low-effort items pulled across buckets
high_impact_initiatives — items with the highest expected business impact regardless of effort
resource_requirements, dependencies, success_metrics — context for execution

What `competitive_analysis` returns¶

{
  "credits_used": 20,
  "updates_analyzed": 0,
  "competitive_risk_summary": "...",
  "urgency_level": "high|medium|low",
  "by_update": [ /* per-update risk assessment */ ],
  "competitor_landscape": [ /* competitors known to be moving on this */ ],
  "market_timing": [ /* timing-related risk factors */ ],
  "recommendation": "...",
  "sa_bias_alert": { /* nullable — flagged when SA recommendation pattern looks suspect */ }
}

The SA bias alert is the distinctive output. Sometimes an SA correctly recommends against adopting a new AWS service. Sometimes the recommendation is driven by unfamiliarity rather than fit. The alert flags that possibility so you can seek a second opinion — it's not a presumption of bias, just a flag worth checking.

Required and optional parameters¶

Parameter	Required	Description
`operation_type`	Yes	One of `match_updates`, `generate_roadmap`, `competitive_analysis`, `implementation_plan`, `bulk_partner_match`
`partner_id`	Required for partner-scoped modes	The partner to match updates against
`time_horizon`	Optional for `generate_roadmap`	`3_months`, `6_months`, or `12_months` (default)
`focus_areas`	Optional for `generate_roadmap`	Array of areas (e.g. `["ai", "security", "data"]`)
`update_ids`	Optional for `competitive_analysis` and `implementation_plan`	Specific updates to focus on; defaults to top matched updates
`min_score`	Optional	Override the 55-default `RECOMMENDATION_THRESHOLD`
`days_back`	Optional	Lookback window for fresh matches

Prerequisites¶

AwsInnovationUpdate corpus must exist. Run AWS Innovation Monitor's batch_process_feeds mode first to build the table. The two capabilities are designed to work together; this one assumes the corpus is populated.
Partner.product_categories must be set for matching to produce meaningful scores.
Embeddings present on updates. batch_process_feeds generates them by default (generate_embeddings: true).

When to use which capability (Innovation Monitor vs. Productization Recommendation)¶

Scenario	Recommended capability + mode
"Scan AWS feeds and surface what matters"	Innovation Monitor → `discover_updates`
"Build / refresh the AWS update corpus"	Innovation Monitor → `batch_process_feeds`
"Deep analysis of one specific announcement"	Innovation Monitor → `analyze_relevance`
"Find updates semantically similar to a topic"	Innovation Monitor → `find_similar`
"Match the corpus to a single partner's products"	`productization_recommendation` → `match_updates`
"Build a 12-month product roadmap"	`productization_recommendation` → `generate_roadmap`
"Risk analysis if we don't adopt these"	`productization_recommendation` → `competitive_analysis`
"Step-by-step implementation guidance"	`productization_recommendation` → `implementation_plan`
"PDM portfolio review across all partners"	`productization_recommendation` → `bulk_partner_match`

API access¶

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