A call disposition is useful in an export only when everyone understands its meaning. For a Zanus AI CRM evaluation, define a small dictionary for the fields your reporting process needs. Ask which fields the proposed export actually supplies; this guide proposes a schema rather than documenting vendor API names.
Keep four concepts separate
Contact status describes whether a person answered or the attempt reached voicemail. Business outcome describes what was agreed or completed. Action state shows whether related work is pending or saved. Review state explains whether staff have checked the automated interpretation.
Combining these into one “successful” label makes reconciliation difficult. For example, a connected call can contain a requested callback whose CRM write failed. Each part is relevant and none should silently overwrite the others.
Write the dictionary with examples
| Proposed field | Example meaning |
|---|---|
| contact_status | answered, voicemail or no_answer |
| business_outcome | information_given, callback_requested or unresolved |
| action_state | not_required, pending, saved or failed |
| review_state | unreviewed, accepted or corrected |
These are illustrative internal labels, not confirmed Zanus field names. Map the actual export to your dictionary and preserve any unmapped values for review.
Make corrections and unknowns visible
Include the source interaction ID, customer or request reference, event time and dictionary version. Where the source permits it, preserve the original automated classification beside the reviewed result and correction reason.
Keep an explicit unknown or unmapped state. Replacing a new vendor value with the closest-looking old code can change reporting without anyone noticing. Date your mapping and ask who owns changes when the export format or business definitions evolve.
Reconcile a small export before reporting
Use test calls for voicemail, a tentative booking, a completed booking and a failed CRM update. Compare the source conversation, export values and destination records. Replay an event and check whether your reporting counts the same action twice.
Have operations staff read the dictionary and explain each example in their own words. Then calculate outcome rates using the appropriate fields and denominator. Keep raw imports available under the organisation’s retention policy so a later correction can be traced to its source rather than guessed from a chart.
Questions and answers
Are these example disposition codes documented Zanus API fields?
No. They are a proposed internal reporting dictionary. Confirm the actual vendor export and map it explicitly before using it in operational reporting.
Building a Mapping Governance Process Before the First Live Export
Once a dictionary exists on paper, the practical challenge shifts to governing who interprets it, who updates it and how disagreements get resolved before they reach a report. A dictionary without an owner behaves like an undated policy: staff apply it according to their own reading until a discrepancy surfaces in a quarterly review, at which point the root cause is hard to establish. The sections below propose an operational governance structure a business can put in place before committing to any specific export configuration.
Assign a named dictionary owner at the outset, not after the first export run. That person need not be a technical specialist. Their responsibility is to ensure that the written definition matches the operational intent, that proposed changes are reviewed before they enter production, and that any version of the dictionary can be retrieved by date. In practice this often sits with a reporting or operations lead rather than with an individual agent or an IT administrator, because both operational meaning and data destination need to be visible to one person simultaneously.
Establish a simple change log alongside the dictionary itself. Each row should carry the field affected, the previous definition, the new definition, the reason for the change and the date from which it applies. This does not need dedicated software. A shared document with edit history or a version-controlled file serves the same purpose. What matters is that anyone querying historical data can reconstruct which definition was in force at the time the record was created, rather than applying today's definition retrospectively.
Separating Customer Identity from Request Identity in Export Records
A call disposition export frequently conflates two distinct things: the person who was contacted and the request that was handled. These can diverge in ways that affect both reporting accuracy and downstream workflows. A single caller may place multiple requests within one conversation. One request may involve multiple callers, as in a household or a business account where different individuals are authorised to act. Treating the call record as the unit of analysis can obscure both patterns.
When designing the export schema, consider whether each row represents a call event, a request, or a contact. If the system exports one row per call, a single conversation containing two bookings and one cancellation may appear as a single undifferentiated outcome. If reporting later asks how many bookings occurred in a period, the figure drawn from call-level records may not match the figure drawn from booking records, and the discrepancy will be difficult to diagnose without understanding the original unit of measurement.
A practical approach during evaluation is to construct a fictional but realistic scenario and trace it through the proposed export. Consider a caller who first asks about availability, then requests a callback for a specific time, then separately asks about cancellation policy. The conversation ends without a definitive booking. Trace which of the four proposed dictionary fields would apply to each element: the contact was answered, the business outcome might be callback_requested alongside information_given, the action state for the callback is pending while the information request requires no further action, and the review state is unreviewed. The export should ideally support all of these simultaneously rather than collapsing them into a single cell.
Where the export format does not natively support multiple outcomes per call, document that constraint explicitly in your evaluation notes. It is not a reason to disqualify a system automatically, but it is a reason to decide in advance how your team will handle multi-outcome calls during reconciliation and whether a supplementary manual record will be maintained.
Designing Exception Handling for Automated Classification Disagreements
Automated interpretation of call content will occasionally produce a classification that does not match what a staff member who listens to the recording would assign. This is not a fault specific to any particular vendor; it is a structural feature of classification systems that a governance process should anticipate rather than react to after a reporting problem surfaces.
The review_state field proposed in the dictionary exists precisely for this reason. An unreviewed record carries the automated classification. An accepted record confirms that a staff member agreed with it. A corrected record shows both the original automated classification and the reviewed replacement, along with a reason code. That reason code is worth designing carefully. Generic reasons such as system error or agent override convey little analytical value. More useful reason categories might include: the caller's intent was ambiguous and the automated interpretation chose one reading, the conversation ended without a clear resolution and the classification defaulted to the nearest available code, or the request involved a category not yet defined in the dictionary.
Reviewing every record manually is impractical at volume, so the evaluation process should include a discussion with the vendor about which records are flagged for mandatory review and which are passed through automatically. Whatever the answer, your governance process should define a sample review routine: a defined number of randomly selected records reviewed each period by an operations staff member who was not involved in the original call. This provides a continuing check on whether the automated classification drift has occurred without waiting for a formal audit cycle.
When a correction is made, the corrected record should not overwrite the original automated value. Both should be stored, with the correction clearly identified as a human review rather than an updated automated output. This distinction matters if the organisation later analyses classification accuracy, adjusts the system configuration, or needs to defend a reporting figure to an internal or external stakeholder.
Acceptance Exercise: Observable Evidence Before Operational Sign-Off
An acceptance exercise for a call disposition export should generate observable evidence that a non-technical reviewer can examine. The following is a structured approach a business evaluation team could conduct.
Select a small set of test interactions covering the range of outcomes the system will be expected to handle. For each, write down in plain language what occurred: who called, what they asked, what was agreed or not agreed, and what follow-up action was created or deferred. This written record becomes your reference against which the export will be compared.
After each test interaction, retrieve the corresponding export record and map each export value to the dictionary. Produce a simple side-by-side table: the plain-language description in one column, the export value in the adjacent column, the dictionary definition in the next, and a notes column for any discrepancies or decisions made during mapping. Operations staff, not only technical staff, should participate in reading the table. If an operations staff member cannot explain what a given export value means after consulting the dictionary, the dictionary needs revision before live deployment.
Specifically test the following cases, because they reveal gaps that straightforward successful calls do not:
- A call that reached voicemail and left a message, where no action was taken by the business
- A call where the caller requested a callback and the callback request was recorded in the CRM
- A call where the caller requested a callback but the CRM write did not succeed
- A call where the caller's intent was unclear and the automated classification assigned an outcome
- A call involving a request category not yet included in the dictionary
For each case, confirm whether the export produces a distinct and unambiguous record. If two cases produce identical export values despite representing meaningfully different operational situations, that is a finding to document and discuss with the vendor before the export is used in reporting.
After completing the table, have the dictionary owner and at least one operations staff member sign off on the findings in writing. This creates a dated record of what was verified, what was outstanding and what decisions were made. It also establishes accountability: if a reporting discrepancy surfaces later, the sign-off record identifies what was known at deployment and what was not.
Handover Arrangements When Staff or Systems Change
A well-designed dictionary and governance process can become ineffective quickly if the person who built it leaves or if the export format changes without the dictionary being updated. Handover arrangements are worth designing deliberately rather than leaving to informal knowledge transfer.
The dictionary documentation should be written assuming the reader has no prior knowledge of the project. This means including not only the field names and definitions but also the reasoning behind each choice. Why was callback_requested separated from information_given rather than merged? What would happen to reporting if they were combined? What real operational questions does each distinction allow the business to answer? These explanations are as important as the definitions themselves, because they enable a future owner to make consistent decisions when a new situation arises that the original dictionary did not anticipate.
When the export format changes, whether because the vendor updates their output structure or because the organisation changes its CRM configuration, the dictionary update process should be triggered explicitly. A useful question to include in any vendor change notification process is: does this change affect the values, structure or completeness of the call disposition export? If the answer is yes or uncertain, the dictionary owner reviews the change against the current schema before the updated export enters production reporting.
If operational responsibility transfers to a different team or individual, a structured handover exercise is more reliable than documentation review alone. The incoming owner should repeat a simplified version of the acceptance exercise using live records, describe in their own words what each field means and how corrections are made, and confirm they know who to contact if the export produces an unmapped value. These are observable checks rather than assumptions about what was read and understood.
Questions That Should Change a Purchase or Configuration Decision
Certain questions, when answered by a vendor during evaluation, carry enough weight to change either the decision to proceed or the specific configuration chosen. The following are worth raising directly, with the understanding that answers should be confirmed in writing rather than accepted as verbal assurances.
Does the export include the original automated classification alongside any reviewed or corrected value, or does a correction overwrite the source? If the export only carries the most recent value, the organisation loses the ability to analyse classification accuracy over time or to reconstruct the original record if a correction is disputed.
What happens to export records when a call involves an interaction category that the system has not encountered before? Specifically, does the record carry an explicit unmapped or unknown value, or does the system assign the nearest existing code silently? Silent assignment is operationally riskier than an explicit unknown, because the record appears complete when it is not.
Is the export format versioned, and how are changes communicated? A format that changes without notice will break mapping logic that the organisation has built against it. Understanding the communication and notice process allows the organisation to design its own update procedures accordingly.
Can the export be filtered or segmented by dictionary field values, or is a full export the only option? This affects how practical it is to isolate records requiring review, to audit a specific outcome category or to supply a subset of records to a downstream process.
What does the export contain when a call did not complete, was disconnected early or was otherwise anomalous? Understanding the edge case behaviour of the export is as important as understanding the standard case, because edge cases often produce the records that are hardest to reconcile and most likely to distort aggregate reporting.
A practical next step before finalising any export configuration is to collect answers to each of these questions, map each answer against your dictionary and governance design, and identify the gaps that remain. Those gaps represent the decisions your organisation needs to make, not the vendor. They include how unmapped values will be handled, who reviews corrections, how the dictionary will be maintained as business definitions evolve and what evidence will confirm that the export is working as intended after each system change. Resolving those decisions in advance, with clear ownership and written documentation, is the operational foundation that makes a call disposition export usable rather than merely available.
Sources and next steps
This independent Sysevo guide proposes evaluation steps and reports no hands-on vendor test or compliance certification. Check the linked primary sources and Zanus AI’s current Front Office overview, then confirm the scope of your proposed deployment in writing.
Read Zanus AI call outcomes and next actions for related context. To compare providers using the same requirements, explore Sysevo CRM or book a Sysevo demo. Sysevo publishes this guide and offers its own business software.
Related workflow guides
For current Sysevo subscriptions and usage options, compare plans and pricing.
