IP operations
Agentic docketing: hard schema for legal truth, soft schema for operational context
Why a docket that runs the work has to be built in two layers, and what goes wrong when they are collapsed into one.
Draft — pending professional review. This piece has not yet been cleared by a named professional. Passages marked for verification contain legal specifics that must be confirmed against current law before publication.
Most docketing systems are filing cabinets with dates in them. They record what happened and, asked precisely enough, what is due. What they do not do is make the next thing happen — that remains a person's job, and it is the part that fails.
The two layers
The legal facts of a matter — the events, the deadlines, the communications, the approvals, the invoices — belong in a deterministic system of record. That record has to be rigid, versioned and boring. It is written once, appended rather than overwritten, and every entry carries its provenance. Rigidity is the point: it is what makes the record trustworthy a year later.
The operational context around those facts — what this matter needs next, what is missing, who has not replied, what should be escalated — cannot be rigid, because it changes continuously and it is not always reducible to fields. That belongs in a separate layer that reads the record, annotates it, and acts on it.
Hard schema for legal truth. Soft schema for operational context.
What goes wrong when they are collapsed
Collapse them one way — everything rigid — and you get a system that records accurately and does nothing. Every operational judgement has to be made by a person outside the system, which is exactly the work that was supposed to be absorbed.
Collapse them the other way — everything soft — and you get a system that is helpful and untrustworthy. If a statutory date can be produced by inference, then it can be produced wrongly, and nobody can tell by looking at it which dates were derived from a rule and which were guessed.
Provenance is what joins them
The operating layer is allowed to say anything it can attribute. A summary points at the entries it was built from. A deadline points at the rule that produced it. A report points at the playbook clause that named its recipients. An output that cannot be traced back to a record entry is not treated as a fact about the matter.
That constraint is what makes the layer safe to rely on. It is not that the operating layer is never wrong; it is that a wrong answer can always be traced to the thing that produced it.
This material is general information about Indian IP practice and procedure. It is not legal advice, it does not create an attorney–client relationship, and it should not be relied on for a specific matter without professional advice on the facts.