| UI or UX | UI + UX - Rule authoring surface for defining event triggers, conditions, actions, tests, activation, and runtime safeguards | UI + UX - Runtime checkpoint that pauses AI or automation until an eligible human authorizes the next step | UI + UX - Routed decision workflow for requests that require authorized approval | UI + UX - Dedicated user or app configuration management surface | UI + UX - Sectioned enterprise configuration form | UI + UX - Context-sensitive workflow action suggested for the user's current record, case, conversation, or task |
| UI guidance | Render automation rule builder as a structured authoring surface with rule name, trigger, scope, conditions, branch logic, actions, recipients or targets, test data, previewed matching records, activation state, run history, and owner. | Render a human approval gate as a paused automation checkpoint with the proposed action, tool or workflow step, triggering rule, risk level, payload snapshot, requester or agent, approver eligibility, timeout, and explicit approve, reject, edit, cancel, or bypass controls. | Render approval workflow as a routed request record with requester, object, requested action, approver eligibility, required rule, current gate, due date, decision controls, comment history, and outcome consequences. | Render settings management as a durable configuration surface with a clear Settings or Preferences entry point, grouped categories, current values, setting descriptions, ownership or scope labels, dependencies, save or immediate-apply behavior, status feedback, search or section navigation for larger sets, and reset or restore defaults where appropriate. | Render one coherent form surface with a clear title, section navigation or section headings, current section state, grouped controls, persistent labels, helper text, required indicators, error summary, field-level errors, review or summary state, and one save contract. | Render the recommended next action as a bounded suggestion card or action slot that names the action, trigger context, expected outcome, owner, due time or urgency, eligibility status, and why the system is suggesting it now. |
| UX guidance | Make automation safe to activate: validate missing pieces, preview matches, test with sample events, disclose permissions and side effects, handle conflicting rules, and provide pause, draft, rollback, and audit paths. | Use human approval gate when automation is ready to act but policy, risk, confidence, cost, access, publication, deployment, customer impact, or legal consequence requires a human decision before execution continues. | Make the approval route understandable over time: who has the request, what decision is needed, what happens on approval or rejection, what blocks self-review, and where the requester can revise or withdraw. | Use settings management when users need to review and change persistent app, account, workspace, notification, privacy, display, integration, or system behavior outside the immediate task flow. | Use a complete complex form when users must edit many related product settings together and need to inspect dependencies across sections before saving. | Keep recommended next action separate from broad recommendations, task lists, primary buttons, assignments, approvals, and review queues because it is optional, contextual, explainable, and directly tied to the current work state. |
| Good UI | A support admin creates a rule: When priority is high and sentiment is negative, assign to Escalations, notify the manager, and pause if the customer is VIP; the builder previews matching tickets and lets the admin run a dry test. | An AI support agent pauses before issuing a refund, shows the proposed amount, customer, policy match, confidence, source grounding, approver role, timeout, Approve refund, Edit amount, Reject, and Stop run controls. | A production deployment request shows requester, service, environment, change summary, required reviewer group, self-review restriction, wait timer, Approve and Reject with reason controls, and a timeline of route changes. | A notification settings page groups channels, quiet hours, digest frequency, and workspace scope; each row shows current value, effect, dependency, and whether changes save immediately. | A policy configuration form has Basics, Rules, Approvals, Notifications, and Review sections; Approvals shows an error because the threshold in Rules requires at least two approvers. | A support case sidebar recommends Send refund-policy article because the customer asked about a refund twice, shows confidence, source snippets, and opens a draft for review. |
| Bad UI | A rule says Send notification on update but hides the fields, recipients, repeat limits, and whether it will run on existing records. | A reviewer can approve an automation run after the action already executed, making the gate only decorative audit text. | A stale email approval applies to a request version that has changed. | Users change the wrong scope because local, account, workspace, and organization settings look identical. | A long settings page saves invalid combinations because validation checks each field in isolation and ignores dependencies between threshold, approvers, and escalation owner. | Required onboarding tasks and optional suggestions are mixed in one list labelled Next best actions. |
| Good UX | A team lead edits an assignment rule, sees that 18 current tickets would match, tests one sample ticket, notices a conflict with an older VIP rule, resolves priority order, and activates the rule with run logging enabled. | A billing lead opens the paused refund gate, sees that the amount is under policy but source grounding is partial, edits the refund to the verified amount, approves, and the agent resumes only that step. | A finance approval requires both manager and controller approval; the first approval advances the sequence, a rejection stops downstream processing, and the requester sees the exact rejection reason and revise action. | A user turns off weekly digest emails, sees the setting save immediately, keeps urgent security emails enabled, and understands the workspace-level override. | A user jumps from Review back to Rules, changes a dependency, and sees Review marked stale until the affected sections pass validation again. | A representative reviews the suggested reply, sees that it was triggered by customer intent and a matching knowledge article, edits the draft, and sends it. |
| Bad UX | A rule activates immediately while the user is still choosing conditions and sends messages to hundreds of customers. | A human approves a stale agent action from email and the agent applies it to a different customer state. | The first approver clicks Approve and the system marks the whole request approved even though policy required everyone to approve. | A user changes a privacy setting thinking it affects only one project, but the value applies to the whole account. | A user clicks Save, sees three generic errors somewhere in the page, and must scan many controls to discover which section owns each problem. | A user accepts a suggested discount and only afterward learns it changed contract terms. |
| Best fit | Users need to create or edit automations that run later based on events, conditions, schedules, or record changes. | An AI agent, workflow, deployment, or automation is ready to perform a high-impact step and must pause for human authorization. | A submitted request needs authorized approval before a deployment, purchase, access grant, publication, content change, policy exception, financial action, or workflow step can proceed. | Users need to inspect and change persistent app, account, workspace, privacy, notification, display, integration, device, or system behavior. | A product configuration has many related fields across named sections. | Users are working in a record, case, conversation, or workflow where choosing the next action is costly or error-prone. |
| Avoid when | The user only needs to approve one currently paused automation step. | The action has already happened and users only need an audit log. | The user only needs to check their own answers before submission. | The task is a one-time transaction, submission, setup wizard, or onboarding flow. | The form is a short related set of fields that fits a simple single-page form. | The action is always required and should be a task, validation, or workflow gate. |
| Required state | Empty draft state with rule name, trigger picker, condition builder, action picker, owner, and disabled activation. | Paused gate state with proposed action, payload snapshot, reason for gate, and run context. | Draft or submit-ready request handoff state | Settings overview with categories and current values | Initial draft state with named sections, required-field convention, grouped controls, and save unavailable or clearly pending validation. | Recommended action state with action title, reason, evidence, trigger, consequence, and accept or review path. |
| Accessibility burden | Use labelled groups for trigger, condition, branch, action, test, and activation sections. | Expose gate status, proposed action, target, payload summary, risk, approver rule, timeout, and current run state as text. | Use labelled request summary, approver, status, due date, decision, comment, and history regions. | Use headings, section labels, fieldsets, and persistent labels so settings groups and controls have clear programmatic names. | Use semantic headings, fieldsets, legends, and persistent labels so each section and field group has a programmatic name. | Use a labelled region or card heading that identifies the suggestion as recommended, optional, and scoped to the current work object. |
| Common misuse | Hiding which existing records or future events will match the rule. | Using a post-action audit record as if it were a pre-action human gate. | Treating the first approval as final when the rule requires more approvers. | Using settings as a dumping ground for unrelated navigation, billing, help, profile setup, onboarding, or destructive account actions. | Splitting tightly related settings across pages so dependencies are invisible during editing. | Hiding why the action was suggested, what evidence was used, or what will change. |