Reports

When analysis completes, every call gets a structured report: a single JSON object with a fixed, strict schema. The same shape is used to constrain the model's output and to document what you can rely on. Fetch it from GET /v1/calls/{id}/report.

The report is a strict object — unknown top-level keys are not added. The fields below in bold are always present; the rest may be present depending on what the call contained.

Annotated example

{
  "overall_score": 78,                // integer 0–100, the headline grade
  "score_breakdown": {                // per-category integers, each 0–10
    "rapport": 8,
    "discovery": 7,
    "needs_analysis": 7,
    "value_articulation": 8,
    "objection_handling": 5,
    "active_listening": 8,
    "talk_ratio": 6,
    "next_steps": 7,
    "qualification": 7,
    "pricing_confidence": 5,
    "closing": 6,
    "compliance": 9
  },
  "summary": "Strong discovery and rapport; a pricing objection was raised late and left unresolved. Clear next step booked.",
  "call_outcome": "follow_up_scheduled",   // free-text outcome label
  "close_likelihood": 62,                  // integer 0–100, deal-close probability
  "manager_review_required": false,        // boolean (optional)
  "customer_needs": ["faster onboarding", "SOC 2 evidence"],
  "pain_points": ["manual call review takes hours"],
  "objections": [                          // array of objects
    { "objection": "Price is higher than incumbent",
      "raised_by": "customer",
      "handled": false,
      "suggested_response": "Anchor on time saved per rep per week." }
  ],
  "buying_signals": [                      // array of objects
    { "signal": "Asked about contract start dates", "strength": "high" }
  ],
  "missed_opportunities": [                // array of strings
    "Did not ask for the economic buyer to join the next call"
  ],
  "coaching_notes": {                      // object, keyed coaching guidance
    "priority": "Practice the pricing-objection reframe",
    "detail": "Reps lose momentum when price comes up late; pre-empt it during value articulation."
  },
  "rep_strengths": ["Excellent rapport", "Asked layered discovery questions"],
  "rep_weaknesses": ["Avoided the pricing conversation"],
  "recommended_training": ["Objection handling: price"],   // array of strings (optional)
  "suggested_follow_up": "Email a one-page ROI summary and propose a buyer-present call Thursday.",
  "crm_note": "Acme — warm. Pricing objection open. Next: ROI summary + buyer-present call.",
  "compliance_alerts": [                   // array of objects (empty when clean)
    { "type": "missing_disclosure", "severity": "low",
      "detail": "Call recording disclosure not clearly stated." }
  ],
  "transcript_highlights": [               // array of objects (optional)
    { "speaker": "customer", "quote": "If this saves my team Fridays, that's huge.",
      "label": "buying_signal", "t": 742 }
  ],
  "timeline": [                            // array of objects, call structure over time
    { "t": 0,   "phase": "rapport" },
    { "t": 180, "phase": "discovery" },
    { "t": 900, "phase": "pricing" }
  ],
  "confidence_level": "high"               // "low" | "medium" | "high"
}

Fields

Field	Type	Description
`overall_score` *	integer 0–100	The headline quality grade for the call.
`score_breakdown` *	object	Map of category → integer 0–10. Covers the ~12 evaluation categories (rapport, discovery, needs analysis, value articulation, objection handling, active listening, talk ratio, next steps, qualification, pricing confidence, closing, compliance).
`summary` *	string	One-paragraph human summary of the call.
`call_outcome` *	string	Outcome label, e.g. `follow_up_scheduled`, `closed_won`, `no_decision`.
`close_likelihood` *	integer 0–100	Estimated probability the deal closes.
`manager_review_required`	boolean	True if a human manager should review this call.
`customer_needs`	string[]	Needs the customer expressed.
`pain_points`	string[]	Problems the customer is trying to solve.
`objections` *	object[]	Each objection raised, who raised it, whether it was handled, and a suggested response.
`buying_signals` *	object[]	Positive intent signals with a strength.
`missed_opportunities` *	string[]	Things the rep should have done but didn't.
`coaching_notes` *	object	Actionable coaching guidance for the rep.
`rep_strengths` *	string[]	What the rep did well.
`rep_weaknesses` *	string[]	Where the rep needs to improve.
`recommended_training`	string[]	Suggested training topics.
`suggested_follow_up` *	string	Recommended next action / follow-up.
`crm_note` *	string	A short note ready to write back into your CRM.
`compliance_alerts` *	object[]	Compliance issues detected (empty array when clean). Each has a type, severity, and detail.
`transcript_highlights`	object[]	Notable quotes with speaker, label and timestamp.
`timeline` *	object[]	Call structure over time — phase markers with timestamps.
`confidence_level` *	enum	The model's confidence in the analysis: `low`, `medium`, or `high`.

* Always present.

Building an agent on top of these fields? See AI agents for which fields to act on and a ready-to-use tool schema.