Why This Match.

AI should never be a black box — especially not in finance. Every PRISM match in TECH comes with a human-readable rationale that decomposes why the engine surfaced this partner, how each of the three PRISM layers contributed to the final score, which specific overlaps drove the recommendation, and which concerns lowered the score before you commit time to a partner call.

What the card contains

Every Why-This-Match card is structured into four sections, produced at match-generation time by PRISM Layer 2 (the Claude scorer) and persisted alongside the match row.

Narrative summary

Three to five sentences written by Claude describing the specific strategic overlap between your organization and the partner. The model is prompted to reference concrete, named attributes — the partner's sector, their stated partnership objectives, their stage, your own target criteria — rather than generic "both sides are innovative" filler. A representative example:

"Strong strategic alignment with Ramp. Their corporate spend platform complements your treasury-management roadmap cleanly — Ramp's existing integrations with Stripe Issuing and Marqeta show a technical architecture ready for mid-market bank partnerships. Their recent $150M Series D signals runway for a 12-18 month pilot, and their East-Coast engineering presence eases coordination with your NY partnership team. Their Corporate Spend sector is adjacent-but-not-overlapping with your Partnership Intelligence positioning, which makes this a complementary rather than competitive fit."

Layer breakdown

A side-by-side view of the three PRISM scoring layers with the contribution each made to the final score. This is unique to TECH — other AI matching platforms show only an opaque final number.

Shared objectives

A list of concrete partnership objectives that both organizations have in their profile — not inferred, but directly extracted from what both sides told TECH they were looking for. If one side says "Compliance automation" and the other says "Compliance-first infrastructure," that surfaces here.

Concerns

Factors that lowered the score or that a reasonable user should consider before reaching out. Typical concerns: stage mismatch (a Series A reaching out to a 150-year-old bank without a pre-existing connection), regulatory gap (partner operates in a state where you do not hold licensure), geography (time-zone or in-person proximity barriers), or capacity signal (partner's LinkedIn activity suggests a hiring freeze).

How the rationale is generated

The Why card is not a post-hoc explanation of a separate scoring decision — it is produced by the same Claude call that generates the LLM fit score. One prompt, one response, one consistent piece of reasoning. This is deliberate: a rationale generated separately from the score can drift from what actually drove the score.

The prompt structure explicitly requires the model to:

• Cite at least two named attributes from each organization's profile.
• Reference the shared objective field(s) if any overlap exists.
• Identify and list concerns even when the overall score is high (ceiling 8/10 narrative when concerns are material).
• Frame projections with confidence language ("likely", "suggests") rather than definitive claims.
• Never reference the user's name, email, phone, or other PII — the prompt excludes all PII from the input.

What the Why card does NOT do

Clarity about limits is as important as clarity about capabilities:

• It does not predict success. A high score means high compatibility, not high probability of closing. Many high-fit partnerships fail for execution reasons the AI cannot see.
• It does not replace due diligence. The rationale is a starting point for deeper investigation using Intelligence tools (Fit Analysis, ROI Projection, Due Diligence) and human judgment.
• It does not know what it does not know. The model cannot access private pipeline data, unannounced leadership changes, or confidential financial signals. If material facts are non-public, the rationale will miss them.
• It does not substitute for your compliance team. Regulatory observations in the rationale are cues, not clearance.

Response schema

{
  "summary": "3-5 sentence narrative",
  "layers": {
    "vector_similarity": {
      "score": 0.87,
      "weight": 0.3,
      "contribution": 0.26
    },
    "llm_fit": {
      "score": 9,
      "weight": 0.5,
      "contribution": 0.45,
      "dimensions": {
        "problem_alignment": 9,
        "technical_compatibility": 8,
        "regulatory_alignment": 9,
        "cultural_fit": 8,
        "strategic_value": 10
      }
    },
    "feedback_adjustment": {
      "value": 0.02,
      "reason": "user has accepted 4 of 5 Corporate Spend partners"
    }
  },
  "final_score": 0.94,
  "factors": {
    "shared_objectives": ["..."],
    "concerns": ["..."]
  },
  "ai_meta": {
    "model": "claude-sonnet-4-5",
    "model_version": "...",
    "prompt_template_version": "prism-v3",
    "generated_at": "iso8601"
  }
}

API surface