A conversation channel for sensemaking. It allows users to recover context (“Where am I?”), acquire domain knowledge (“What does this mean?”), and bridge the gulf of execution (“How do I do this?”).

💭 Help is primarily pull-based (user seeks understanding). When the system proactively offers to do things, that is embedded intelligence.

Why users seek help

Do: How do I execute this task?

Users need to act but face interface friction. Help here is operational—tooltips, input hints, empty states, wizards. The goal is immediate unblocking, not deep understanding.

Study: Why does this work this way?”

Users want to build mental models of the underlying system or domain logic. Help here is conceptual—explanations, documentation, “learn more” sidebars. Example: “Why is this option disabled? Because the fiscal year is closed.”

Locate: Where is X?

Users know something exists but can’t find it. This cuts across depth levels—locating a button is operational, locating a policy definition is conceptual. Mechanisms include search, index, command menu, and glossary.

Recover: Something went wrong—now what?

Users encounter errors or get stuck. Recovery may require quick interface correction (operational) or deeper diagnostic understanding (conceptual). Often begins with status feedback that signals the problem.

When recovery needs exceed what static content can provide, help shifts to conversation—agents or chat interfaces that can reason about the specific context.

Subject matter

Each help request concerns either:

• Interface: How to operate the UI—where to click, what fields to fill, how to navigate. Help here addresses execution mechanics.
• Domain: Why the system works this way—business rules, data relationships, regulatory constraints. Help here builds mental models of the underlying logic.

The same purpose can require either. A user seeking to “Do” might need interface help (“Where’s the export button?”) or domain help (“What format meets compliance requirements?”). “Locate” cuts across both—finding a button is interface; finding a policy definition is domain.

Contextual availability

Help effectiveness depends on proximity to point of need. The closer help is to where users encounter friction, the less cognitive overhead required to find and apply it.

• Inline: Guidance embedded directly in the flow—input hints, empty states, field-level validation messages.
• Just-in-time: Revealed on specific interaction—tooltips on hover, contextual panels on focus.

What belongs where

To-do

Guided tours and overlays