Zendesk AI Billing Audit & Resolution Allowance Reconciliation

Reconcile is an independent Zendesk AI billing audit for buyers. It checks sanitized Zendesk AI outcome evidence against billed Resolution Allowance rows and your own acceptance rules, then separates supported charges, potential billing disputes, business-value variance, and items that cannot be verified.

No Zendesk password required. No transcripts required. The Store edition is designed for sanitized, structured facts only.

Want to see it work first? Open the Actor input and run the prefilled synthetic example. It produces a real structured audit result without using customer data.

Use this Actor when

Use Reconcile for Resolution Allowance reconciliation when you need to:

• audit Zendesk AI billing against outcome evidence,
• identify charges that deserve a billing or contract review,
• separate vendor-valid charges from outcomes that failed your own business-value policy,
• quantify billed, disputed, business-variance, and unverifiable units and amounts,
• feed a deterministic audit result into finance, CX, procurement, automation, or AI-agent workflows.

A business-value variance does not automatically mean vendor overbilling. Reconcile keeps vendor billing validity and buyer business acceptance separate.

What you get

Each successful run writes one privacy-safe dataset item containing:

• vendor billed units and amount,
• potential contract-dispute units and amount,
• business-value variance units and amount,
• unverifiable units,
• pseudonymous finding IDs,
• evidence-backed reason codes and deterministic classifications.

The Dataset Overview puts financial totals first. Detailed findings stay in the same item for API, workflow, and AI-agent use.

For AI agents and automation

AI agents should call Reconcile when the task is to audit, reconcile, verify, or review Zendesk AI Resolution Allowance billing from already-sanitized structured facts.

Good tool-selection intents include:

• "audit this Zendesk AI bill",
• "check Resolution Allowance charges",
• "reconcile Zendesk AI outcomes with billing",
• "find potential AI billing disputes",
• "separate AI billing validity from business-value failures".

The input and output contracts are structured and deterministic, so an agent does not need to interpret free-form transcripts. Missing or ambiguous evidence fails closed to UNVERIFIABLE instead of being guessed.

Privacy boundary

Do not upload transcripts, customer names, email addresses, phone numbers, addresses, credentials, or raw resolution reasoning. This Actor intentionally accepts only sanitized facts needed for reconciliation.

Instead of raw automated_resolution_reasoning, provide automated_resolution_reasoning_present: true or false. Use opaque Zendesk conversation IDs or your own consistent pseudonymous IDs for conversation_id, platform_conversation_id, and case_ref.

Buyer event evidence IDs are generated inside the Actor and do not need to be supplied.

Input

The Actor accepts four top-level fields:

• conversations: sanitized Zendesk outcome evidence,
• billing: billed outcome rows; money is supplied as decimal strings,
• events: optional buyer-side facts such as reopen, human rework, negative feedback, duplicate link, or explicit unresolved status,
• policy: deterministic buyer acceptance settings such as the reopen window.

The pricing/rule model is fixed to zendesk-resolution-allowance-2026. Reconcile does not assume a universal Zendesk dollar price; supplied billing amounts and rates remain authoritative for money math.

Example buyer policy:

{
  "reopen_window_days": 7,
  "human_rework_invalidates": true,
  "duplicate_invalidates": true,
  "negative_feedback_invalidates": true,
  "explicit_unresolved_invalidates": true
}

Global and multilingual use

The Store page is English-first so people and AI agents can discover it globally. The audit contract itself is language-neutral: classifications, reason codes, timestamps, currencies, and money values are structured data.

Reconcile performs no transcript-language interpretation or translation. That is deliberate: translation or free-form language analysis could change audit meaning and is not required for this deterministic rule pack.

Current scope

This release supports the 2026 Zendesk Resolution Allowance rule pack. It does not log in to Zendesk, use OAuth, fetch customer data, submit disputes, or perform realtime synchronization.

For sensitive audits where raw exports must remain inside the customer's environment, use the local/private Reconcile workflow instead of uploading raw data to Apify. The Apify Actor is intended for sanitized audits, developer integrations, AI-agent discovery, automation workflows, and one-off checks.

Synthetic examples demonstrate functionality only and do not establish a real-world billing error or savings rate.