Organise

Classification Application

Auto-tag every incoming item against your classification scheme.

Runs each arriving item — document, ticket, transaction, product — through a classification scheme and assigns category labels. Rules, ML classifiers, or hybrid. The output is a continuously labelled stream that makes downstream search, routing, and reporting possible. Nothing else in the stack changes what kind of thing an item is — this does.

Shape

classification schemebillingtechaccountinboundclassifierrules + modelconf ≥ 0.85billingtechaccount?low-conf → review

Operational dimensions

No human in loop

Runs without a person in the path.

Event-triggered

Fires when an upstream condition occurs.

Medium data gravity

Holds working state that compounds over runs.

Two-way integration

Reads from and writes to external systems.

Inputs

  • item stream (documents, tickets, transactions, products, content)
  • classification scheme (taxonomy or label set)
  • rules, model weights, or hybrid config
  • confidence threshold settings

Outputs

  • per-item label assignments with confidence scores
  • ambiguous / low-confidence cases flagged for review
  • classification distribution metrics (optional)

Mechanism

Applies a defined classification scheme to incoming items by assigning one or more category labels per item — the tagged items are the deliverable.

Why this is a primitive

Cannot be decomposed: the inspect-item → predict-or-rule-match-categories → assign-labels operation is a single act of categorisation. It assumes the scheme already exists (vocabulary-authoring), assumes the item is in canonical form, and does not link the item into a graph (graph-instantiation). It just answers 'what category/categories does this item belong to?'

Where it shows up

Insurance — classifies incoming claims by type, sub-type, and severity so they route to the correct handler queue immediately on receipt
E-commerce — auto-tags product listings with category, material, and occasion labels so search and recommendation systems have consistent facets
Legal ops — classifies contract clauses by obligation type (payment, liability, IP, termination) so counsel can scan risk exposure without reading full text
Customer success — tags inbound support tickets by issue category and urgency so SLA timers start on the right queue

Related primitives

Tags

AIautonomousstructured-databatchreal-time

See where it fits.

Primitives are configured into named solution shapes for each client’s domain. The fastest next step is a conversation about which shape fits your problem.

Start a conversation