AI Studio - Complete Guide

Use AI Studio to configure the agents, data access, knowledge, and workflows behind Yarken's AI capabilities. This guide shows which component to use and links to the setup instructions for each task.

Before you begin

Who can configure AI Studio

AI Studio configuration requires Admin privileges, a Yarken AI licence, access to the enabled AI framework in your environment, and Ask Yarken Beta enabled under Admin > Settings > AI > Ask Yarken. Individual Beta email allowlisting is not required. Users without AI Studio administration access may still use enabled AI features according to their role, licence, and permissions.

Open AI Studio

1. Sign in to Yarken with an eligible administrator account.

2. Open Admin > AI Studio.

3. Choose the relevant section from the left navigation. The table below explains each section.

AI settings are separate from AI Studio. Use Admin > Settings > AI to configure Ask Yarken settings and Action Types before connecting them to AI Studio workflows.

Choose the component you need

Action Types are configured separately under Admin > Settings > AI > Action Types. Define what kind of finding an Automation can raise, its payload fields, default owner, card content, and matching rules for repeated findings. See Configure AI Action Types.

How components work together

Each component has a specific job. A Specialist handles a domain question. A Data Source provides governed reporting data. A Playbook supplies a repeatable procedure. A Skill guides the expected output. An Automation runs the procedure on a trigger and can raise an assigned action.

Example: Explain a budget variance

1. A user asks Ask Yarken why a cost centre is above its budget.

2. The system can select the relevant IT Cost Analyst Specialist.

3. The Specialist queries an authorized spend Data Source and checks budget, actuals, and variance drivers.

4. When available and applicable, a Playbook provides a standard investigation sequence and a Skill shapes the explanation.

5. The user receives an analysis grounded in the data the system can access.

Example: Monitor an issue without a user prompt

1. An administrator creates an Automation with a schedule or supported event trigger.

2. Its steps query data, evaluate a condition, and extract finding values.

3. Raise an action populates a configured Action Type. The Action Type controls the expected fields, card display, owner, and repeat-finding identity.

4. Administrators inspect Run History to verify the run and field mappings. For repeated findings, selected identity fields determine whether the existing action is updated.

Raise and handle an action: start here

Use this sequence when an Automation should identify a finding, assign it to a person, and continue work after the person selects Act on this.

1. Define the Action Type. Open Admin > Settings > AI > Action Types. Choose one Shape, define required fields, card impact, default recipient, and Lifecycle matching. See Configure AI Action Types.

2. Build the producing Automation. In Admin > AI Studio > Automations, choose a trigger, add analysis steps, then add Raise an action. Select the Action Type and map its required fields. See Raise an action from an Automation.

3. Check the action card. Use Test run and Run History, then inspect impact, evidence, and owner in Inbox.

4. Build the follow-up Automation. Create another Automation using On app event > An action is accepted. Select intended Action Types, add follow-up steps, and activate it. See Review findings with Take Action.

5. Test the full path. Select Act on this on the card and verify the follow-up run. Add Wait for approval when further work needs human authorization.

Set up a new AI use case

1. Define the outcome. State the question or recurring task and what a useful result must contain.

2. Check data access. Confirm necessary saved reports or approved documents exist, then set up Data Sources and Knowledge Base where required.

3. Configure analysis. Add or update the appropriate Specialist, Skill, or Playbook.

4. Configure execution. If the task needs to run automatically, create an Automation and set triggers, variables, conditions, and approvals.

5. Test. Run an Automation test or ask a representative question. Check the output against source data and user permissions.

6. Review ongoing activity. Use Automation Run History and AI Studio Insights to investigate results and usage.

Find instructions by task

Access and data checks

• Data Sources with Apply user authorisation enabled restrict query results to rows the current user is authorized to access.

• Review imported source permissions and available report fields before using them in an agent workflow.

• Use Wait for approval within an Automation before steps that require human authorization.

• For supported agent conversations, configure PII Guardrails to handle email addresses, phone numbers, and credit/debit card numbers. Select None, Mask, or Redact independently for each type, and verify the results with approved test data.

• Simulations appear in AI Studio navigation but do not yet provide execution controls.

Related content

• AI settings

• Ask Yarken (Admin Settings)

• Ask Yarken