Back to compare picker

AI agent acts without approval vs Human approval gate vs Agent plan preview vs Agent progress trace vs Tool-use visibility vs Dangerous-action review

Choose AI agent acts without approval when an agent sends, spends, edits, deploys, deletes, publishes, changes access, or calls an external side-effect tool before the user approves the exact payload.

Decision dimensions

Dimension AI agent acts without approvalHuman approval gateAgent plan previewAgent progress traceTool-use visibilityDangerous-action review
UI or UX UI + UX - Anti-pattern for agentic side effects executed before human authorizationUI + UX - Runtime checkpoint that pauses AI or automation until an eligible human authorizes the next stepUI + UX - Pre-execution preview of an AI agent's proposed multi-step plan, tools, data access, and expected outputsUI + UX - Live execution trace for an AI agent or automation run after work has startedUI + UX - Inspectable disclosure of AI agent tool calls, inputs, outputs, permissions, and side effectsUI + UX - Pre-execution review of high-impact actions that may affect money, access, production systems, legal status, customers, external recipients, or safety
UI guidance Do not show an agent as merely thinking, drafting, or preparing when it has already sent, changed, purchased, deployed, revoked, deleted, or externally committed something.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 agent plan preview as a pre-run plan with objective, ordered steps, planned tools, data sources, permissions, assumptions, dependencies, approval gates, expected outputs, and controls to edit, approve, run, save, or cancel.Render an agent progress trace as a live, ordered run timeline with run ID, plan version, current step, queued steps, active tool or task, elapsed time, last event time, step status, blocked gates, retry state, and final outcome.Render tool-use visibility as an inspectable tool-call surface with tool name, purpose, input summary, output summary, permission scope, data source, side-effect risk, status, timestamps, and redaction state.Render dangerous-action review as a pre-execution checkpoint that names the armed action, actor, target, scope, affected systems, external effects, risk reason, evidence, freshness, permissions, alternatives, and exact outcome of Run, Cancel, Edit, or Escalate.
UX guidance Protect users from surprise agent side effects by separating proposed work from executed work, pausing risky steps, and requiring authorization that is valid only for the displayed payload.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.Use agent plan preview when users need to understand and shape what an AI agent will do before it starts calling tools, changing records, sending messages, spending budget, or making external side effects.Use agent progress trace when an AI agent or automation has started multi-step work and users need to monitor progress, intervene on stalls or gates, understand partial completion, and know whether the reviewed plan is still being followed.Use tool-use visibility when users need to understand, verify, approve, debug, or audit how an AI agent used tools, data, or integrations during a run.Use dangerous-action review when an action is not necessarily destructive but can create high-impact consequences such as sending money, changing access, executing production commands, contacting customers, publishing content, filing a legal response, or letting an agent use a privileged tool.
Good UI A support agent pauses before issuing a refund, shows customer, order, amount, policy source, approver role, and Approve refund, Edit amount, Reject, and Stop run controls.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 sales assistant previews a six-step account-research plan with CRM lookup, web search, draft email, approval gate before send, estimated sources, and editable recipient scope.An account research agent trace shows Run A-204, reviewed plan P-18, completed CRM lookup, active policy search, queued draft email, approval gate pending before send, elapsed time, and a View tool details control.A research agent shows Knowledge search, Web search, and CRM lookup tool cards with purpose, input summary, source scope, output summary, status, and redacted raw request details.A production console shows Restart payment workers, affected region, open incidents, customer impact, rollback owner, evidence links, change window, dry-run result, and Run restart only after the reviewer checks the risk inventory.
Bad UI A chat message says I handled it and reveals the agent already issued a refund without any review screen.A banner says Human approval needed but does not show the tool call, payload, approver, timeout, or resume consequence.The UI says I have a plan and immediately starts executing without showing steps, tools, data access, or external side effects.A spinner says Working on it while an agent calls several tools with no step identity, elapsed time, blocked state, or recovery path.A trace says Using tools but never names the tool, input, source, output, permission, or side effect.A privileged tool button says Continue and immediately sends a customer email, changes access, and updates billing without showing the payload or external recipients.
Good UX A manager sees the agent is armed to change account access, edits the target group, approves the revised payload, and the run resumes only that step.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 manager removes the Send email step, narrows the data source to approved knowledge, approves the remaining plan, and sees execution start from the revised version.A user watches the active step move from searching policies to drafting the email, opens the blocked permission item, grants access, and sees the run continue from the same step.A user opens the active CRM lookup, sees it is read-only, verifies the account ID, and continues watching the run.A release manager sees that a deploy action affects production EU, has a stale smoke test, cancels execution, refreshes checks, and then runs the action with an audit record.
Bad UX A user asks an agent to research an account and later discovers it changed the opportunity stage.A human approves a stale agent action from email and the agent applies it to a different customer state.Users approve a plan that says Research account but the agent also updates the opportunity stage.Users cannot tell whether the agent is stuck, waiting for approval, or finished because all states use the same animated progress label.Users cannot tell whether the agent searched the web, read private files, or changed customer data.A user approves a notification from email after the underlying payload changed, and the system executes against a different customer.
Best fit An AI agent or automation can create side effects that affect customers, money, access, production systems, legal status, public content, sensitive data, or external recipients.An AI agent, workflow, deployment, or automation is ready to perform a high-impact step and must pause for human authorization.An AI agent or automation can show a proposed multi-step plan before execution.An agent or automation run has started and spans multiple steps, tools, gates, or side effects.An AI agent or automation calls tools, functions, APIs, retrieval systems, commands, or integrations.A user, agent, automation, or admin tool is about to execute a high-impact action that can affect money, access, production systems, legal/compliance state, customers, external recipients, sensitive data, or safety.
Avoid when The agent can only perform read-only retrieval and cannot affect external systems or user-visible state.The action has already happened and users only need an audit log.The system cannot generate a reliable plan before execution.Execution has not started and users need to inspect or edit a proposed plan.The system cannot reliably identify tool calls, inputs, outputs, status, permissions, or side effects.The risk is narrowly permanent deletion or loss of a named object; use destructive action confirmation.
Required state Read-only step allowed state with no external side effect.Paused gate state with proposed action, payload snapshot, reason for gate, and run context.Draft plan state with objective, ordered steps, planned tools, and expected output.Run started state tied to run ID, plan version, objective, and user who started the run.Pending tool call state with tool name, purpose, requested permission, and side-effect risk.Armed action state with verb, target, payload, actor, source, and exact execution boundary.
Accessibility burden Expose whether a step is proposed, armed for approval, running, completed, rejected, cancelled, bypassed, or rolled back as text.Expose gate status, proposed action, target, payload summary, risk, approver rule, timeout, and current run state as text.Expose objective, plan version, step order, step status, tool, data access, side effect, and expected output as text.Expose trace status, run ID, current step, elapsed time, blocked state, final outcome, and details availability as text.Expose tool name, status, permission, risk, input summary, output summary, and redaction reason as text.Use headings and labels that name the action and target before risk details.
Common misuse Treating Run agent as permission to execute every hidden side-effect step.Showing Approve without the exact action, payload, target, risk, or resume consequence.Showing a vague plan summary while hiding planned tool calls, data access, and side effects.Using one spinner or vague Thinking label for a multi-step agent run.Showing a vague Using tools label without names, inputs, outputs, or permissions.Using vague Are you sure, Continue, or Proceed copy without naming the dangerous operation.

AI agent acts without approval

UI or UX
UI + UX - Anti-pattern for agentic side effects executed before human authorization
UI guidance
Do not show an agent as merely thinking, drafting, or preparing when it has already sent, changed, purchased, deployed, revoked, deleted, or externally committed something.
UX guidance
Protect users from surprise agent side effects by separating proposed work from executed work, pausing risky steps, and requiring authorization that is valid only for the displayed payload.
Good UI
A support agent pauses before issuing a refund, shows customer, order, amount, policy source, approver role, and Approve refund, Edit amount, Reject, and Stop run controls.
Bad UI
A chat message says I handled it and reveals the agent already issued a refund without any review screen.
Good UX
A manager sees the agent is armed to change account access, edits the target group, approves the revised payload, and the run resumes only that step.
Bad UX
A user asks an agent to research an account and later discovers it changed the opportunity stage.
Best fit
An AI agent or automation can create side effects that affect customers, money, access, production systems, legal status, public content, sensitive data, or external recipients.
Avoid when
The agent can only perform read-only retrieval and cannot affect external systems or user-visible state.
Required state
Read-only step allowed state with no external side effect.
Accessibility burden
Expose whether a step is proposed, armed for approval, running, completed, rejected, cancelled, bypassed, or rolled back as text.
Common misuse
Treating Run agent as permission to execute every hidden side-effect step.

Human approval gate

UI or UX
UI + UX - Runtime checkpoint that pauses AI or automation until an eligible human authorizes the next step
UI guidance
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.
UX guidance
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.
Good UI
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.
Bad UI
A banner says Human approval needed but does not show the tool call, payload, approver, timeout, or resume consequence.
Good UX
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.
Bad UX
A human approves a stale agent action from email and the agent applies it to a different customer state.
Best fit
An AI agent, workflow, deployment, or automation is ready to perform a high-impact step and must pause for human authorization.
Avoid when
The action has already happened and users only need an audit log.
Required state
Paused gate state with proposed action, payload snapshot, reason for gate, and run context.
Accessibility burden
Expose gate status, proposed action, target, payload summary, risk, approver rule, timeout, and current run state as text.
Common misuse
Showing Approve without the exact action, payload, target, risk, or resume consequence.

Agent plan preview

UI or UX
UI + UX - Pre-execution preview of an AI agent's proposed multi-step plan, tools, data access, and expected outputs
UI guidance
Render agent plan preview as a pre-run plan with objective, ordered steps, planned tools, data sources, permissions, assumptions, dependencies, approval gates, expected outputs, and controls to edit, approve, run, save, or cancel.
UX guidance
Use agent plan preview when users need to understand and shape what an AI agent will do before it starts calling tools, changing records, sending messages, spending budget, or making external side effects.
Good UI
A sales assistant previews a six-step account-research plan with CRM lookup, web search, draft email, approval gate before send, estimated sources, and editable recipient scope.
Bad UI
The UI says I have a plan and immediately starts executing without showing steps, tools, data access, or external side effects.
Good UX
A manager removes the Send email step, narrows the data source to approved knowledge, approves the remaining plan, and sees execution start from the revised version.
Bad UX
Users approve a plan that says Research account but the agent also updates the opportunity stage.
Best fit
An AI agent or automation can show a proposed multi-step plan before execution.
Avoid when
The system cannot generate a reliable plan before execution.
Required state
Draft plan state with objective, ordered steps, planned tools, and expected output.
Accessibility burden
Expose objective, plan version, step order, step status, tool, data access, side effect, and expected output as text.
Common misuse
Showing a vague plan summary while hiding planned tool calls, data access, and side effects.

Agent progress trace

UI or UX
UI + UX - Live execution trace for an AI agent or automation run after work has started
UI guidance
Render an agent progress trace as a live, ordered run timeline with run ID, plan version, current step, queued steps, active tool or task, elapsed time, last event time, step status, blocked gates, retry state, and final outcome.
UX guidance
Use agent progress trace when an AI agent or automation has started multi-step work and users need to monitor progress, intervene on stalls or gates, understand partial completion, and know whether the reviewed plan is still being followed.
Good UI
An account research agent trace shows Run A-204, reviewed plan P-18, completed CRM lookup, active policy search, queued draft email, approval gate pending before send, elapsed time, and a View tool details control.
Bad UI
A spinner says Working on it while an agent calls several tools with no step identity, elapsed time, blocked state, or recovery path.
Good UX
A user watches the active step move from searching policies to drafting the email, opens the blocked permission item, grants access, and sees the run continue from the same step.
Bad UX
Users cannot tell whether the agent is stuck, waiting for approval, or finished because all states use the same animated progress label.
Best fit
An agent or automation run has started and spans multiple steps, tools, gates, or side effects.
Avoid when
Execution has not started and users need to inspect or edit a proposed plan.
Required state
Run started state tied to run ID, plan version, objective, and user who started the run.
Accessibility burden
Expose trace status, run ID, current step, elapsed time, blocked state, final outcome, and details availability as text.
Common misuse
Using one spinner or vague Thinking label for a multi-step agent run.

Tool-use visibility

UI or UX
UI + UX - Inspectable disclosure of AI agent tool calls, inputs, outputs, permissions, and side effects
UI guidance
Render tool-use visibility as an inspectable tool-call surface with tool name, purpose, input summary, output summary, permission scope, data source, side-effect risk, status, timestamps, and redaction state.
UX guidance
Use tool-use visibility when users need to understand, verify, approve, debug, or audit how an AI agent used tools, data, or integrations during a run.
Good UI
A research agent shows Knowledge search, Web search, and CRM lookup tool cards with purpose, input summary, source scope, output summary, status, and redacted raw request details.
Bad UI
A trace says Using tools but never names the tool, input, source, output, permission, or side effect.
Good UX
A user opens the active CRM lookup, sees it is read-only, verifies the account ID, and continues watching the run.
Bad UX
Users cannot tell whether the agent searched the web, read private files, or changed customer data.
Best fit
An AI agent or automation calls tools, functions, APIs, retrieval systems, commands, or integrations.
Avoid when
The system cannot reliably identify tool calls, inputs, outputs, status, permissions, or side effects.
Required state
Pending tool call state with tool name, purpose, requested permission, and side-effect risk.
Accessibility burden
Expose tool name, status, permission, risk, input summary, output summary, and redaction reason as text.
Common misuse
Showing a vague Using tools label without names, inputs, outputs, or permissions.

Dangerous-action review

UI or UX
UI + UX - Pre-execution review of high-impact actions that may affect money, access, production systems, legal status, customers, external recipients, or safety
UI guidance
Render dangerous-action review as a pre-execution checkpoint that names the armed action, actor, target, scope, affected systems, external effects, risk reason, evidence, freshness, permissions, alternatives, and exact outcome of Run, Cancel, Edit, or Escalate.
UX guidance
Use dangerous-action review when an action is not necessarily destructive but can create high-impact consequences such as sending money, changing access, executing production commands, contacting customers, publishing content, filing a legal response, or letting an agent use a privileged tool.
Good UI
A production console shows Restart payment workers, affected region, open incidents, customer impact, rollback owner, evidence links, change window, dry-run result, and Run restart only after the reviewer checks the risk inventory.
Bad UI
A privileged tool button says Continue and immediately sends a customer email, changes access, and updates billing without showing the payload or external recipients.
Good UX
A release manager sees that a deploy action affects production EU, has a stale smoke test, cancels execution, refreshes checks, and then runs the action with an audit record.
Bad UX
A user approves a notification from email after the underlying payload changed, and the system executes against a different customer.
Best fit
A user, agent, automation, or admin tool is about to execute a high-impact action that can affect money, access, production systems, legal/compliance state, customers, external recipients, sensitive data, or safety.
Avoid when
The risk is narrowly permanent deletion or loss of a named object; use destructive action confirmation.
Required state
Armed action state with verb, target, payload, actor, source, and exact execution boundary.
Accessibility burden
Use headings and labels that name the action and target before risk details.
Common misuse
Using vague Are you sure, Continue, or Proceed copy without naming the dangerous operation.
Decision rules
  • Choose AI agent acts without approval when an agent sends, spends, edits, deploys, deletes, publishes, changes access, or calls an external side-effect tool before the user approves the exact payload.
  • Choose human approval gate when a running agent or automation is paused at one armed step and an eligible human can approve, reject, edit, cancel, bypass, or let it expire before execution resumes.
  • Choose agent plan preview when users need to inspect proposed steps, planned tools, data access, side effects, approval gates, and expected outputs before the agent starts running.
  • Choose agent progress trace when the run has already started and users need a live timeline of queued, active, blocked, approved, failed, cancelled, or completed steps.
  • Choose tool-use visibility when the design problem is exposing exact tool names, inputs, outputs, permissions, redactions, and payload details for inspection rather than authorizing execution.
  • Choose dangerous-action review when a single user-initiated high-impact action needs payload, target, consequence, and final action review outside an agent run.
  • A plan preview or broad Run control authorizes only the displayed plan version; it does not authorize hidden future side-effect steps unless those steps and gates are explicitly listed.
  • A progress trace or activity log can show that an action happened, but it cannot convert a post-action audit event into valid pre-action approval.
  • Approval is valid only for the shown target, payload, tool, model or workflow version, policy version, and approver rule; stale email or mobile approvals must be refused.
  • If an agent acted without approval, show the executed action as already committed and provide stop, rollback, revoke, notify, incident report, escalation, and audit paths where possible.
Inspect live examples
Failure modes
  • The agent says it drafted an email but sends it to customers when the user taps Run.
  • A broad approval of the plan is used to issue refunds that were not listed in the plan.
  • An approval card appears after the database update already committed.
  • A stale notification resumes an agent step against a changed customer, source, or payload.
  • The requester approves their own high-risk agent action despite separation-of-duties policy.
  • The audit trail records a side effect but cannot show what the user saw before it executed.