Prompt classification

Understand how your users are using AI.

Prompt classification answers a simple question: How are your users using AI? Classify intent before the prompt, upload, or tool action crosses the boundary so teams can review, govern, and act.

SD-1 / Prompt classification Detect Classify Control Drafting Coding Research Data handling Tool use Allow Warn Block Evidence retained before the boundary
SD-1: Capability-specific hero showing intent lanes before an AI prompt crosses the boundary. Production metadata: 3ls-sd-1-solutions-prompt-classification-composite.svg; SVG master rendered fluidly with explicit 1600x1200 dimensions; crop-safe for 4:3 and stacked mobile hero placement.; task-1093-inline-svg-master

How are your users using AI?

3LS helps teams distinguish drafting, coding, research, summarization, data handling, and tool-driven work so AI usage becomes understandable instead of opaque.

Phase 1

AI activity appears

A person or tool starts using AI for a concrete task.

Phase 2

Intent is classified

3LS categorizes whether the interaction looks like drafting, coding, research, summarization, data handling, or tool use.

Phase 3

Risk is prioritized

Teams can separate routine activity from interactions that need closer review.

Phase 4

Operators get context

Usage patterns become visible enough to govern, coach, and control.

SD-2 / Detect, classify, control Detect Classify Control Action Analysis Decision Evidence Allow Warn Block Evidence retained before the boundary
SD-2: Three-step visual showing user action, analysis, then policy decision and evidence. Production metadata: 3ls-sd-2-solutions-policy-flow-diagram.svg; SVG master rendered fluidly with explicit 1600x900 dimensions; usable as 16:9 desktop panel and full-width mobile diagram.; task-1093-inline-svg-master

What teams gain

Visibility that supports governance, coaching, and review.

Drafting

See writing and summarization patterns

Understand when people are using AI to draft, summarize, and restructure information for everyday work.

Outcome: Baseline normal AI usage

Coding and research

Separate technical work from higher-risk handling

Distinguish coding, troubleshooting, and research assistance from interactions that start to involve sensitive business material.

Outcome: Faster triage for reviewers

Data handling and tool use

Highlight actions that deserve extra attention

Surface when prompts point toward data handling, sharing, or tool-driven behavior so operators know where controls may matter.

Outcome: Better governance decisions

From the blog

Prompt Risk and Classification Signals

Articles that show how prompt intent turns into risk: injection, tool steering, and the classification signals teams need before enforcement.