AI Studio (Beta) - Complete Guide

Use this guide to understand how AI Studio works, how its features are invoked in Yarken, and where to configure each AI capability.

Overview

AI Studio is the admin workspace for configuring, governing, testing, and monitoring Yarken's AI capabilities.

It brings together the components that support Ask Yarken, specialist agents, governed data access, reusable skills, playbooks, automations, document-based knowledge, guardrails, copilots, simulations, and insights.

Access to AI Studio requires Admin privileges, a Yarken AI licence, the required environment access, and Ask Yarken Beta enabled for the environment. AI Studio is available to all administrators who meet these requirements and is not limited to Yarken administrators. Individual email addresses are no longer used to grant Beta access. See Ask Yarken (Admin Settings).

AI Studio has two main flows:

User flow: how AI capabilities are invoked from Ask Yarken, Copilots, Automations, or other AI-enabled areas of the platform.
Admin flow: how administrators configure, test, govern, and monitor those capabilities in AI Studio.

Section	Menu item	Primary purpose
Configure	Specialists	Configure domain-focused agents that the orchestrator can delegate work to.
Configure	Data Sources	Define governed datasets and query tools available to AI workflows.
Configure	Skills	Define reusable response, reasoning, and output behaviors.
Orchestrate	Playbooks	Define standard operating procedures the agent can follow.
Orchestrate	Automations	Configure scheduled, manual, event-triggered, and approval-based agent flows.
Knowledge	Knowledge Base	Upload and manage documents the agent can search and cite.
Safety	Guardrails	Configure content filters, denied topics, and data privacy rules.
Channels	Copilots	Configure chat, wizard, and investigative Copilot modes.
Evaluate	Simulations	Run simulated conversations to evaluate agent quality.
Observe	Insights	Review execution metrics, latency, and skill usage logs.

Configure AI Studio Copilots under Admin > AI Studio > Channels > Copilots. Copilot configuration for the agentic framework has moved from general AI settings into AI Studio. The Dashboard AI Narrative is one of the configurable Copilots available under Admin > AI Studio > Channels > Copilots.

Within Orchestrate, Playbooks appears before Automations in the current navigation order.

User flow: how AI Studio features are invoked

Users do not normally open AI Studio during day-to-day analysis. They use AI capabilities from places such as Ask Yarken, supported Copilots, scheduled workflows, or other AI-enabled experiences.

AI Studio controls what happens behind the scenes.

Feature	How users experience it
Specialists	Invoked by Ask Yarken through the orchestrator when the prompt matches a specialist domain.
Data Sources	Used behind the scenes when a Specialist, Playbook, or Automation needs governed data.
Skills	Applied when the response needs a specific behavior or output format, such as outlier detection, root-cause analysis, or an HTML/PDF deliverable.
Playbooks	Applied when a user request matches a repeatable operating procedure, such as variance analysis, vendor review, QBR creation, or cloud optimization.
Automations	Run when a manual trigger, schedule, insight, or workflow condition starts an agent flow.
Knowledge Base	Used when the agent needs to search and cite uploaded reference documents.
Guardrails	Applied across AI interactions to enforce content, topic, and data privacy boundaries.
Copilots	Invoked from specific AI modes, such as chat, wizard, or investigative experiences.
Simulations	Used by admins to test AI behavior before users rely on the configuration.
Insights	Used by admins to review execution metrics, latency, skill usage, and operational signals over time.

The Dashboard AI Narrative Copilot generates a written summary from selected custom-dashboard widgets or instructions and can follow the dashboard's active filters. An administrator enables the Copilot under Admin > AI Studio > Channels > Copilots, while dashboard editors add and configure its widget from Analytics > Custom Dashboards. See Add and configure a Dashboard AI Narrative widget.

End-to-end AI flow

How AI Studio decides what to use

AI Studio components work together. The orchestrator decides what to use based on the user’s intent, available context, required data, and expected output.

Signal in the user request	Component likely used
The task follows a known repeatable process	Playbook
The question is about a specific domain	Specialist
The answer needs governed Yarken data	Data Source
The agent needs to ask for missing context, generate a chart, or perform a system action	Tool
The answer needs approved document context	Knowledge Base
The output needs a specific format or behavior	Skill
The workflow should run on a trigger or schedule	Automation
The user is working inside a specific AI-enabled mode	Copilot
The request needs safety, topic, or privacy checks	Guardrails

A single response can use more than one component.

Example: if a user asks for a QBR report on cloud spend, the orchestrator may use:

A Playbook for the QBR workflow.
A Cloud FinOps Analyst for domain analysis.
Data Sources for cloud and spend data.
A Skill for report formatting.
A Tool to generate charts, if the output needs visual analysis.

Admin flow: how to configure AI Studio

Use this sequence when setting up AI Studio for a new environment or new AI use case.

Feature relationship map

Feature	Depends on	Feeds into
Specialists	Instructions, Data Sources, conversation starters	Ask Yarken and orchestrated AI workflows
Data Sources	Governed Yarken data and authorization settings	Specialists, Playbooks, Automations
Skills	Response instructions and exposure settings	Ask Yarken, Specialists, Playbooks, Automations, report-style outputs
Playbooks	Operating procedure content, Specialists, Data Sources, Skills	Repeatable analysis and structured AI workflows
Automations	Triggers, steps, data access, variables, approvals, and notification settings	Scheduled, manual, or event-based agent flows
Knowledge Base	Approved uploaded documents	Document-grounded answers and citations
Guardrails	Content filters, denied topics, privacy rules	All AI interactions
Copilots	UI mode, workflow context, and configured AI behavior	Chat, wizard, and investigative in-product assistance
Simulations	Test prompts and expected behavior	Configuration validation before rollout
Insights	AI execution data and usage logs	Monitoring and improvement after rollout

Finalized AI Studio scope

The finalized AI Studio catalog defines seven default specialists and maps them to governed data tools, reusable skills, playbooks, automations, and eleven technology-finance outcomes. A component appearing in the catalog does not by itself mean that it is enabled in every client instance.

Component	Finalized catalog scope	How to interpret it
Specialists	7	Domain-focused agents in the default client roster.
Data tools	20	Live data sources plus cataloged extensions and dependencies.
Skills	35	Reusable analysis, reasoning, and output behaviors.
Playbooks	15	Interactive standard operating procedures.
Automations	14	Scheduled, event-driven, or manually triggered workflows.
Outcome workflows	11	Business outcomes supported by one or more components.

Default specialists and availability

Specialist	Primary focus	Availability basis
IT Cost Analyst	Actuals, budget, forecast, variance, and overruns	Uses live spend, line-item, and dimension-resolution tools.
Cloud FinOps Analyst	Cloud and AI platform consumption, anomalies, attribution, and commitments	Uses live cloud-spend and dimension-resolution tools; some commitment and action workflows require additional capabilities.
Vendor & Contract Analyst	Contracts, invoices, purchase orders, renewals, and vendor concentration	Uses live contract, spend, line-item, and dimension-resolution tools; some pipeline and invoice-tier workflows require additional capabilities.
License & SaaS Optimization Analyst	Entitlement, usage, shelfware, reclaim, downgrade, and true-up	Complete results require report_license_usage.
Application & Service TCO Analyst	Layered application, product, service, and Solution TCO	Initial aggregates use report_spend; complete TCO requires asset and supporting metric data.
Chargeback & Showback Analyst	Consumer allocation, Bill of IT, bill movement, and disputes	Complete results require report_showback and allocation-method detail.
Cost Model Steward	Mapping rules, coverage, model quality, restatement, and audit	Complete model-health results require report_spend_mapping_quality; rule and audit reviews require their corresponding sources.

Finalized data-tool coverage

Specialist	Core tools	Additional cataloged capabilities
IT Cost Analyst	report_spend; report_spend_line_items; search_dimension_values	report_spend_mapping_quality; extend_spend_variance
Cloud FinOps Analyst	report_cloud_spend; search_dimension_values	report_cloud_commitments; report_cloud_forecast; report_recommendations; tag_upload_rule_actions
Vendor & Contract Analyst	report_contracts; report_spend; report_spend_line_items; search_dimension_values	list_renewal_pipeline; invoice_po_asset_tier_reads
License & SaaS Optimization Analyst	report_license_usage; report_spend; search_dimension_values	report_recommendations
Application & Service TCO Analyst	report_spend; search_dimension_values	report_asset_tco; report_companion_metrics; report_business_case
Chargeback & Showback Analyst	report_showback; report_spend; search_dimension_values	Allocation-method and consumption detail where configured
Cost Model Steward	report_spend_mapping_quality; report_spend; search_dimension_values	rule_inventory_read; audit_log_read

Finalized skill coverage

Specialist	Cataloged skills
IT Cost Analyst	Overrun monthly summary; Variance reconciliation pack; Spend root-cause drivers; RGT / mix analysis; Forecast & budget pack; Scenario / what-if pack; KPI drift & breach narrative; Forwardable narrative / exec summary
Cloud FinOps Analyst	FinOps anomaly review; AI/LLM spend visibility; Commitment health review; Tagging compliance findings; Cost-to-performance advisor; Cloud optimization suggestions; FinOps quick-start tags pack; Forwardable narrative / exec summary
Vendor & Contract Analyst	Contract intelligence brief; Renewal / renegotiation prep; MSP invoice audit; Invoice–PO–contract recon; Vendor consolidation candidates; SLA performance monitor; Forwardable narrative / exec summary
License & SaaS Optimization Analyst	License optimization pack; Software shelfware suggestions
Application & Service TCO Analyst	App portfolio / TCO review; Value realization analytics; Business case builder; Benchmark & cost-to-serve
Chargeback & Showback Analyst	Bill of IT statement; Chargeback zero-usage; Portfolio showback rollup
Cost Model Steward	Model health review; Data quality assistant; Human change & override audit; Agent action audit

Finalized orchestration coverage

Specialist	Playbooks	Automations
IT Cost Analyst	Budget overrun investigation; Month-end variance explanation; Planning cycle variance review; KPI breach investigation	Monthly spend & overrun report; Budget / forecast watchlist; Reporting automation
Cloud FinOps Analyst	Cloud anomaly triage; Tagging gap remediation; AI / LLM spend governance review	Cloud spike scan; Spend commitment monitor; Tagging compliance monitor; FinOps quick-start tags run
Vendor & Contract Analyst	Renewal decision tree; Invoice dispute prep; Consolidation business case	90-day renewal watch
License & SaaS Optimization Analyst	License reclaim before renewal	Pre-renewal license reclaim
Application & Service TCO Analyst	App retire / consolidate / keep	No specialist-specific automation in the finalized catalog
Chargeback & Showback Analyst	Showback dispute handling; Bill change FAQ flow	Zero-usage chargeback detector; Automated Bill of IT
Cost Model Steward	Mapping exception handling	Mapping / DQ monthly check; Human override audit; Agent governance audit

Finalized outcome coverage

Monthly IT budget overrun
Month-end IT cost close
Cloud budget guardrails
Contract renewal readiness
Business-unit showback
Taxonomy and mapping trust
Commitment and waste optimization
Forecast and budget planning cycle
Application portfolio cost rationalization
Vendor consolidation readiness
Shared-service allocation policy trust

How component availability works

The catalog is broader than the set of components that may be enabled in a particular client instance. Administrators should use the AI Studio component cards and the relevant specialist guide to confirm availability before relying on a workflow.

Live components can run when the user has access to the governed source data.
Dependency-based components require the named data source, action, or audit capability before complete results are possible.
A cataloged skill, playbook, or automation can still require configuration, recipients, schedules, approvals, or supporting tools.
When a required capability is unavailable, the specialist must report the limitation and must not fabricate a result.
Platform-wide capabilities and non-client concepts are not part of the seven-specialist default roster.

Governance notes

AI Studio configuration does not replace Yarken access controls.

Administrators should confirm that:

Data Sources use the right authorization settings.
Specialists only use the Data Sources they need.
Skills define response behavior, not source data.
Playbooks use approved operating procedures and do not invent missing data.
Automations use the right trigger, data access, approval, and notification settings.
Knowledge Base contains approved, current documents only.
Guardrails reflect the organization's content, topic, and privacy requirements.
Simulations are run before broad rollout.
Insights and Automation Run History are reviewed after rollout.