AI feature settings and change history
Configure which AI behaviour runs for your visitors and keep a history of saved changes.
What it is
AI Features defines the AI jobs available on your website. You choose a type, an account, instructions and any visitor conditions. Each save of an existing feature creates a new version, so you can inspect earlier instructions.
Who uses it and why
Administrators configure the AI behaviour used by operators and visitors. Developers maintain more specialised instructions, functions and targeting.
Where to find it
Open Settings › AI › Features. You need the Settings role with AI allowed under Settings Permissions, or the Developer role. Powerfront manages Settings › AI › Accounts and Settings › AI › Models.
How it works
- Choose New and select the job in Select Feature Type.
- Enter a Name, choose an Account and configure that type's options.
- Set Visitor Filters (drag and drop) if the feature should apply only to particular visitors.
- Set Priority, check Enabled and save.
For ordinary visitor-based selection, INSIDE considers enabled features of the requested type with a suitable account and model. It checks their visitor conditions and selects the highest priority match. An empty filter matches without additional visitor conditions.
Saving an edit keeps the feature's identity and creates the next version. The previous version remains in history. Assigned skills carry forward. Autopilot saves also update its linked bot configuration.
Use the feature's history action to open History. It shows Version, Date, Operator and Rules, including instruction differences between versions.
Configure it
Ask Powerfront to enable AI and prepare the required accounts and models. Then create the features you need. Use distinct priorities where visitor filters overlap, and keep names descriptive enough to distinguish similar configurations.
Configure each type's instructions and options in its editor. A positive Max Tokens overrides the account value; 0 keeps the account default.
Settings
| Setting | Where | Applies to | Default | Who can change it | What it does |
|---|---|---|---|---|---|
| Enabled | Settings › AI › Features | website | On for a new feature | Settings role, Developer role | Makes the feature eligible for selection when its other requirements match. |
| Priority | Settings › AI › Features | website | 0 | Settings role, Developer role | A higher number takes precedence when several features of the same type match a visitor. |
| Max Tokens | Settings › AI › Features | website | 0, use the account default | Settings role, Developer role | A positive value overrides the account token limit for this feature. |
| Visitor Filters (drag and drop) | Settings › AI › Features | website | Not set: no visitor restriction | Settings role, Developer role | Restricts selection to visitors who meet the conditions; an empty filter adds no restriction. |
| Account | Settings › AI › Features | website | Not set: choose a configured AI account | Settings role, Developer role | Selects the configured AI account used to run the feature. |
| Type | Settings › AI › Features | website | Not set: choose the AI job when creating the feature | Settings role, Developer role | Selects the AI job, such as Autopilot, Copilot or Conversation Summary. |
Fields
| Field | Where | Required | What it is for |
|---|---|---|---|
| Name | Settings › AI › Features | Yes | Identifies the feature in lists and selection controls. |
| Description | Settings › AI › Features | No | Records what this feature is for. |
Permissions
| Permission | Lets you |
|---|---|
| Settings | Manage AI features when the AI permission is allowed. |
| AI | Allows access to AI configuration within Settings Permissions. |
| Developer | Manage AI features. |
Limits and good to know
- Name is required and cannot exceed 1,000 characters. New feature names must be unique on the website.
- Priority and Max Tokens cannot be negative.
- History is a record of saved versions. It does not provide a documented approval or restore process.
- Proactive agent selection has additional eligibility and ordering rules. See AI-005.
- Saving a feature does not by itself prove that a visitor matches it or that its account supports its type.
Depends on and works with
+49 more · used by
- Fix and Re-Phrase for operator draftsConfigured by
Dedicated feature selection uses account support, visitor filters and priority.
- Ask AI beside a visitor conversationConfigured by
Ask AI selection uses feature type, visitor filters and priority.
- AI conversation summariesConfigured by
Summary features are selected using account support, visitor filters and priority.
- Chat Translate and language detectionConfigured by
Language features use configured accounts, visitor filters and priority.
- AI spam checks for incoming conversationsConfigured by
Feature enablement, account, priority and visitor filters control selection.
- Email Autopilot for ticketsConfigured by
The feature selects the account, enablement and instructions.
- Voice Autopilot call handlingConfigured by
The AI feature stores voice behaviour and account selection.
- AI run limitsConfigured by
The feature selects the account and feature token allowance.
- Context Compiler for AI conversationsConfigured by
Each supported feature chooses its context mode.
- AI functions and required inputsConfigured by
AI features choose the functions they can use.
- AI knowledge and web searchConfigured by
The feature assigns functions and instructions to the agent.
- AI integrations and automation toolsConfigured by
Feature instructions and function assignments determine available external actions.
- AI updates to customer and conversation dataConfigured by
The AI feature assigns functions and supplies collection instructions.
- Reusable AI skills and instructionsConfigured by
Skills are assigned to a saved AI feature version.
- Customer goals and AI progress trackingConfigured by
Each AI feature has its own goal tracking scope and instructions.
- AI conversation plans and task progressConfigured by
Work planning is enabled independently on each supported AI feature.
- Specialist sub-agents for AI tasksConfigured by
The target is a configured AI feature on the same website.
- AI agent service connectionsConfigured by
Supported features select the AI account used for execution.
- Rich content in AI repliesConfigured by
Each supported AI feature controls its rich content settings.
- AI product suggestion settingsConfigured by
Feature enablement, account, filters and priority determine which configuration applies.
- AI answer citations and summariesConfigured by
Citation Summary is a reusable feature selected by the answering feature.
- AI conversation quality scoringConfigured by
A matching enabled Conversation Scoring feature supplies the scorecards and account.
- AI resolution and first-contact metricsConfigured by
Matching feature rules, filters and account select the Metrics configuration.
- AI Experiments Lab for response testingConfigured by
Feature-mode experiments use the selected feature configuration.
- AI assistance activityConfigured by
Feature configuration sets the feature token override.
- INSIDE AiFlow test dashboard · managed by PowerfrontConfigured by
Scenario tests use the selected AI feature configuration.
- Virtual Try-On with customer photosConfigured by
Feature configuration determines which agent can call the function.
- Ask AI agents in Team ChatConfigured by
The Ask AI feature must be active, enabled and configured with an account and model.
- AI reply streaming settingsConfigured by
Streaming and interruption policy are selected on each supported feature.
- Conversation history tools for AIConfigured by
The feature controls whether each history capability is available.
- Images and attachments for AI agentsConfigured by
Image analysis is configured on the supported feature.
- Knowledge Base connections for AIConfigured by
The AI feature determines which knowledge functions it can use.
- Algolia Search for AI knowledgeConfigured by
The AI feature selects which search functions it can use.
- Saved responses as AI knowledgeConfigured by
The intended AI feature must be allowed to use the knowledge function.
- Shared AI service connectionsConfigured by
Each supported AI feature selects the prepared connections it uses for prompts, tracing or memory.
- External AI prompts and versionsConfigured by
Each AI feature stores its own prompt selection and fallback instructions.
- AI execution tracing with LangfuseConfigured by
Each supported AI feature selects its own tracing connection.
- Persistent AI memory connectionsConfigured by
The feature selects the memory service, limit and identity scope.
- MCP tool connections for AIConfigured by
The AI feature determines which tool functions it may use.
- Autopilot connections through A2AConfigured by
Feature configuration controls availability, name, description and endpoint enablement.
- ACP compatibility for AI agentsConfigured by
Enabled, active features and their priorities determine which feature executes a run.
- AI platform execution limitsConfigured by
Max Tokens on the AI feature sets a feature-specific allowance or inherits the account default when zero.
- Voice Autopilot connection setupConfigured by
Feature availability and saved configuration determine the voice agent used by a number.
- AI customer and order searchConfigured by
AI feature configuration determines the assistant's functions and permitted use.
- AI commerce functions and access limitsConfigured by
The AI Feature determines the context in which the function is available.
- Autonomous AI performance reportsConfigured by
Autopilot feature selection includes the historical versions that handled conversations.
- AI conversation scoring reportsConfigured by
Scoring features define the scorecards available to filter.
- Operator QA and coaching dashboardConfigured by
Enabled scorecards and their minimum message counts affect scoring and coverage.
- Custom integrationsConfigured by
AI feature function assignments control which agents can use a script-backed integration tool.
View all connections as a list
This feature uses
- Admin settings and feature switches · managed by PowerfrontLicence
Powerfront must enable AI for the website.
- Roles and permissionsConfigured by
Roles and permissions control AI configuration access.
- AI instructions and templatesConfigured by
Instruction sources and templates define what the feature is told to do.
- Reusable AI skills and instructionsEnhances
Autopilot, Copilot and Ask AI can use assigned published skills.
- AI accounts and shared defaultsRequires
Each feature needs a configured AI account.
- AI model catalogueConfigured by
For accounts that use a model, model setup determines the available AI capabilities.
- Visitor event targeting conditionsConfigured by
Visitor targeting conditions restrict feature selection.
Used by
- Conversation translationConfigured by
AI feature setup controls enablement, account and translation options.
- Streaming AI replies for visitorsConfigured by
Streaming and interruption choices belong to the selected AI feature.
- Virtual Assistant tasks started by AutopilotConfigured by
The AI feature determines which functions the agent can use.
- AI instructions and templatesConfigured by
Instructions are saved within versioned AI features.
- Autopilot for live chatConfigured by
Autopilot is configured as a versioned AI feature.
- Autopilot handover and error handlingConfigured by
Recovery options are saved in the Autopilot feature.
- Autonomous agents for proactive chatConfigured by
Feature enablement, visitor filters and priority affect agent selection.
- Copilot reply suggestionsConfigured by
Feature selection uses enablement, account, visitor filters and priority.
- Fix and Re-Phrase for operator draftsConfigured by
Dedicated feature selection uses account support, visitor filters and priority.
- Ask AI beside a visitor conversationConfigured by
Ask AI selection uses feature type, visitor filters and priority.
- AI conversation summariesConfigured by
Summary features are selected using account support, visitor filters and priority.
- Chat Translate and language detectionConfigured by
Language features use configured accounts, visitor filters and priority.
- AI spam checks for incoming conversationsConfigured by
Feature enablement, account, priority and visitor filters control selection.
- Email Autopilot for ticketsConfigured by
The feature selects the account, enablement and instructions.
- Voice Autopilot call handlingConfigured by
The AI feature stores voice behaviour and account selection.
- AI run limitsConfigured by
The feature selects the account and feature token allowance.
- Context Compiler for AI conversationsConfigured by
Each supported feature chooses its context mode.
- AI functions and required inputsConfigured by
AI features choose the functions they can use.
- AI knowledge and web searchConfigured by
The feature assigns functions and instructions to the agent.
- AI integrations and automation toolsConfigured by
Feature instructions and function assignments determine available external actions.
- AI updates to customer and conversation dataConfigured by
The AI feature assigns functions and supplies collection instructions.
- Reusable AI skills and instructionsConfigured by
Skills are assigned to a saved AI feature version.
- Customer goals and AI progress trackingConfigured by
Each AI feature has its own goal tracking scope and instructions.
- AI conversation plans and task progressConfigured by
Work planning is enabled independently on each supported AI feature.
- Specialist sub-agents for AI tasksConfigured by
The target is a configured AI feature on the same website.
- AI agent service connectionsConfigured by
Supported features select the AI account used for execution.
- Rich content in AI repliesConfigured by
Each supported AI feature controls its rich content settings.
- AI product suggestion settingsConfigured by
Feature enablement, account, filters and priority determine which configuration applies.
- AI answer citations and summariesConfigured by
Citation Summary is a reusable feature selected by the answering feature.
- AI conversation quality scoringConfigured by
A matching enabled Conversation Scoring feature supplies the scorecards and account.
- AI resolution and first-contact metricsConfigured by
Matching feature rules, filters and account select the Metrics configuration.
- AI Experiments Lab for response testingConfigured by
Feature-mode experiments use the selected feature configuration.
- AI assistance activityConfigured by
Feature configuration sets the feature token override.
- INSIDE AiFlow test dashboard · managed by PowerfrontConfigured by
Scenario tests use the selected AI feature configuration.
- Virtual Try-On with customer photosConfigured by
Feature configuration determines which agent can call the function.
- Ask AI agents in Team ChatConfigured by
The Ask AI feature must be active, enabled and configured with an account and model.
- AI reply streaming settingsConfigured by
Streaming and interruption policy are selected on each supported feature.
- Conversation history tools for AIConfigured by
The feature controls whether each history capability is available.
- Images and attachments for AI agentsConfigured by
Image analysis is configured on the supported feature.
- Knowledge Base connections for AIConfigured by
The AI feature determines which knowledge functions it can use.
- Algolia Search for AI knowledgeConfigured by
The AI feature selects which search functions it can use.
- Saved responses as AI knowledgeConfigured by
The intended AI feature must be allowed to use the knowledge function.
- Shared AI service connectionsConfigured by
Each supported AI feature selects the prepared connections it uses for prompts, tracing or memory.
- External AI prompts and versionsConfigured by
Each AI feature stores its own prompt selection and fallback instructions.
- AI execution tracing with LangfuseConfigured by
Each supported AI feature selects its own tracing connection.
- Persistent AI memory connectionsConfigured by
The feature selects the memory service, limit and identity scope.
- MCP tool connections for AIConfigured by
The AI feature determines which tool functions it may use.
- Autopilot connections through A2AConfigured by
Feature configuration controls availability, name, description and endpoint enablement.
- ACP compatibility for AI agentsConfigured by
Enabled, active features and their priorities determine which feature executes a run.
- AI platform execution limitsConfigured by
Max Tokens on the AI feature sets a feature-specific allowance or inherits the account default when zero.
- Voice Autopilot connection setupConfigured by
Feature availability and saved configuration determine the voice agent used by a number.
- AI customer and order searchConfigured by
AI feature configuration determines the assistant's functions and permitted use.
- AI commerce functions and access limitsConfigured by
The AI Feature determines the context in which the function is available.
- Autonomous AI performance reportsConfigured by
Autopilot feature selection includes the historical versions that handled conversations.
- AI conversation scoring reportsConfigured by
Scoring features define the scorecards available to filter.
- Operator QA and coaching dashboardConfigured by
Enabled scorecards and their minimum message counts affect scoring and coverage.
- Custom integrationsConfigured by
AI feature function assignments control which agents can use a script-backed integration tool.
FAQ and troubleshooting
Why is my enabled feature not running?
Check website AI enablement with Powerfront, then check Account, model support and visitor filters. Another matching feature may have a higher Priority.
Which feature wins when priorities are equal?
Do not rely on an ordering for ordinary selection when priorities are equal. Give overlapping features distinct priorities.
Why can I see AI settings but cannot save?
Check the AI permission under Settings Permissions. The Settings role can be restricted from AI configuration.
Can I inspect previous instructions?
Yes. Open the feature's history action and review the saved versions and instruction changes in History.