ZynoviqThe law changed last night. Is your AI still right?
Indian indirect tax changes several times a year through council decisions, notifications and circulars. Each change can turn a correct answer into a wrong one without anyone noticing. We watch the changes, re-run the affected cases on your system, and send you an impact note before your customers find the problem.
How it works
- Map your system to the benchmark
Starting from your baseline evaluation or diagnostic, we record which benchmark cases and topics your AI is tested on, and in which version.
- Track rule changes
Qualified professionals read each council decision, notification, circular and significant ruling as it is issued and decide which certified cases it touches.
- Revise the affected cases
Each affected case goes back through the full quality chain: re-solved independently, reviewed, adjudicated and tied to the new authority, with a new valid-as-of date and version.
- Re-run on your system
The revised cases run against your AI in your test environment. Synthetic cases only; no production access.
- Impact note
You receive a one-page note on what the change did to your AI, with the case-level results and the updated regression set.
What you receive each time
| Deliverable | What it is |
|---|---|
| Impact note | One page: what changed, which cases it affected, what your AI now gets wrong, and how serious each failure is |
| Case-level results file | Every affected case with the revised certified answer, the new rule and your AI's response, as a spreadsheet or JSON file |
| Updated regression set | The revised cases, versioned with a new valid-as-of date, ready to re-run in your own pipeline |
| Running record | Every re-run logged by benchmark version and date, so you can show when your AI was last checked against the current law. Evidence |
We don't tell you the law changed. We tell you what it did to your AI.
Plenty of services summarise GST notifications. None of them know what your AI will say about them. The impact note is specific to your system: which answers flipped from right to wrong, which stayed right, and which rule each one now turns on. Your team gets a short list of things to fix rather than a reading assignment.
Who it is for
- Tax-tech and ERP vendors. One contract covers a product used by many customers, so a rule change is caught once, before the next release ships. For software vendors and partners.
- Enterprises. AI that posts entries or answers compliance questions on your own books, checked against the current law each time it changes. For enterprise AI teams.
- Government AI vendors. A running record that the system was re-checked after each change, ready for acceptance and audit. For government AI vendors.
Commercial form
- Annual subscription per system, quoted.
- Covers scheduled re-runs and re-runs after material changes in the law.
- Extra releases on your side are priced per run.
- Starts after a baseline evaluation or an AI Quality Diagnostic, so there is a mapped system to re-run.
Coverage
| Domain | Status |
|---|---|
| GST and Indian indirect tax | Live |
| Payroll under the labour codes | Active — on request |
| Company law | Active — on request |
| TDS and income tax | Planned |
| Customs | Planned |
Only GST is live today. Everything else is a stated plan, not a claim that the service exists in that domain. Every domain and its status.
Regulatory Regression Service — FAQ
How quickly after a rule change do you re-run?+
A target turnaround is written into your contract and depends on the scale of the change: a rate notification affects fewer cases than a new section. It is a target, not a guarantee. Revising a case properly means a professional reads the change, the case is re-solved and reviewed, and only then re-run.
Do you need access to our production system?+
No. Re-runs happen in your test environment, with synthetic cases, exactly as the baseline evaluation did. How tests run.
What if the change itself is ambiguous?+
Some notifications and circulars admit more than one reading until clarified. Affected cases are marked ambiguous, the professional consensus and any dissent are recorded on the case, and your AI's answer is reported against that position rather than scored as a simple failure.
Is the impact note tax advice?+
No. It tells you what a change did to your AI's answers on certified cases. It is not a legal or tax opinion on your affairs or your customers' affairs, and it is not a certification of your system.
Can we combine this with release testing?+
Yes. Most customers run it alongside continuous evaluation: scheduled re-tests and re-tests after your releases cover changes on your side; this service covers changes in the law.