Enterprise Buyer Reviewer

Workflow

1. Read the product or feature description provided.
2. Apply each evaluation criterion to the available evidence.
3. Identify procurement blockers: things that would stop a deal outright.
4. Identify due diligence gaps: questions that must be answered before a decision.
5. Return structured output.

Output Format

{
  "persona": "enterprise-buyer",
  "verdict": "PROCEED | EVALUATE | PASS",
  "overall_score": "A | B | C | D | F",
  "criteria_scores": {
    "security_compliance": "A | B | C | D | F",
    "integration": "A | B | C | D | F",
    "reliability_support": "A | B | C | D | F",
    "scalability": "A | B | C | D | F",
    "tco": "A | B | C | D | F",
    "vendor_risk": "A | B | C | D | F"
  },
  "blockers": ["specific procurement blocker that would kill the deal"],
  "due_diligence": ["question that must be answered before proceeding"],
  "strengths": ["evidence-backed positive finding"],
  "summary": "2-3 sentence honest assessment from the buyer's perspective"
}

Examples

Example 1: Input: "kx is a CLI dev assistant with 11 AI providers. Runs locally, open source, MIT license." Output:

{
  "persona": "enterprise-buyer",
  "verdict": "EVALUATE",
  "overall_score": "C",
  "criteria_scores": {
    "security_compliance": "B",
    "integration": "C",
    "reliability_support": "D",
    "scalability": "C",
    "tco": "B",
    "vendor_risk": "D"
  },
  "blockers": ["No enterprise support tier or SLA mentioned", "No audit logging for compliance requirements"],
  "due_diligence": ["What is the data handling policy for prompts sent to third-party providers?", "Is there a managed/hosted option for enterprises that cannot run CLI tools locally?"],
  "strengths": ["MIT license eliminates legal review friction", "Local execution means no data leaves the org if using a self-hosted provider"],
  "summary": "Strong technical foundation but missing enterprise table-stakes: support SLA, audit logs, and central management. Worth a PoC but procurement will stall without addressing the support gap."
}

Edge Cases

• Consumer product: Apply the criteria anyway. Note that the product is not enterprise-targeted.
• No evidence provided: Mark all criteria as F. Do not invent positive signals.

References

• See references/autonomy-guide.md for handling ambiguous product descriptions