Key takeaways

  • Demand an evidence-based pilot, not a sales-only demonstration.
  • The vendor should explain reconciliation exceptions and preserve an audit trail.
  • AI features need source traceability, privacy terms and human-review controls.
  • Test full exports and account closure before committing critical records.
  • Compare total annual cost, including entities, users, credits, add-ons and implementation.

Demand a workflow demonstration

Provide a redacted sample with ordinary transactions, amendments, credit notes, reverse charge, imports, missing supplier data and one manual exception. Ask the vendor to take it from import to final return and then trace one figure back to the source record.

The demonstration should show errors and uncertainty, not only a clean happy path. Record what requires manual work and who can approve it.

Reconciliation and audit trail

Ask how matching tolerances work, how duplicates are handled, how exceptions are classified and whether a user can override a result. Every material override should record user, time and reason. Historical return versions, amendments, acknowledgements and payment evidence should remain linked.

If the algorithm is proprietary, the vendor may not disclose code, but it should still explain the business rule applied to each exception.

Notice readiness

Require the product to trace a notice figure back through returns, ledgers, invoices and prior reconciliation decisions. A notice feature should preserve the complete document, annexures, service date, deadline, allegation matrix, evidence and filed reply.

If legal research is included, test the applicable financial-year version and exact source support. A generated paragraph without verifiable authority creates more review work than it removes.

AI governance

Ask which model provider receives documents, whether data is retained or used for training, where it is processed, which subprocessors participate, how administrators control access and how deletion is proved. Contract wording should match the actual architecture.

Require visible uncertainty and source links. The system should not silently invent a missing invoice, circular or judgment. Confirm there is a human approval point before any external filing or communication.

Security and continuity

Review authentication, multi-factor support, role design, logs, encryption, backups, vulnerability handling, incident notification and business continuity. Ask how quickly access is revoked for a departing employee and how activity is investigated.

Test a full export. It should include source data, reconciliations, versions, documents, approvals and final outputs in usable formats. Confirm retention and deletion after cancellation.

Pricing and contract

Build a scenario covering every GSTIN, user, return, invoice, API, notice credit, storage tier and support requirement. Ask how renewal prices change and whether unused credits expire. Review service levels, liability limits, confidentiality, data ownership, termination assistance and subcontracting.

Score the pilot

Use a weighted scorecard: statutory accuracy, source traceability, reconciliation accuracy, operational time, security, exportability, support and total cost. Reject any product that cannot reproduce its result or preserve the evidence needed to defend it.

Primary sources

Frequently asked questions

What should a GST software demo include?

Use redacted representative data and require the vendor to show imports, exceptions, corrections, approvals, filing evidence, exports and a notice trace-back.

What should I ask about AI data?

Ask which provider processes prompts, location, retention, training use, subprocessors, access controls, deletion, incident response and contractual responsibility.

Why is export testing important?

A firm must retain working papers and continue operations if it changes vendors. Test human-readable and machine-readable exports before purchase.

How should pricing be compared?

Calculate the annual cost for actual GSTINs, users, filing volume, notice credits, storage, APIs, onboarding, support and taxes.

Explore the GST authority library

Research the underlying documents inside your AI assistant

Connect TaxByKK to Claude or ChatGPT to trace provisions, read source pages and build a cited working draft.

View TaxByKK plans