Marketing Hub
Local Demo

Assistant

Collapse / Expand

Assistant is the operational workspace for assistant within assistant and agent operations, where teams manage concrete records, dependencies, and execution decisions rather than static summaries. The business and compliance goal is to keep governed rag assistant aligned with approved claims, consent controls, market rules, and audit evidence at every step. This page presents role-scoped actions, linked downstream impacts, and policy-aware guardrails so users can work quickly without violating promotional or safety obligations. Next, use this assistant workspace to execute governed tasks and route outcomes to connected teams.

RestrictedSection: Assistant + AgentsSeed: 20260224

Access Boundaries

SSO + RBAC/ABAC controls are evaluated by brand, market, function, agency zone, and risk tier before actions are enabled.

Field Surfaces: Approved-only enforcedExternal Partner: Submit-onlyRisk Tier Scope: Low/Moderate/High/Critical
Assistant: Grounded Responses
215-8% governed
Assistant: Citation Completeness
222+5% governed
Assistant: Confidence Median
229+1% governed
Assistant: Classification Accuracy
236-3% governed
Assistant: MLR Requests
243-7% governed
Assistant: Audit Logs
250+6% governed

Operational Focus

  • Ground output in approved sources only.
  • Expose confidence and classification.
  • Route outputs to MLR when needed.

What To Do Next

  • Run governed prompt.
  • Review cited sources.
  • Request MLR review on output.

Assistant Prompt Studio

Run grounded assistant prompts using approved content with citations, confidence, and output classification.

IDTitleMarketOwnerStatusRiskRelated
ASST-0001Assistant · Prompt run 1SGAI OpsEscalatedCriticalAgent HubMLR Queue
ASST-0002Assistant · Grounding source 2USMedicalDraftLowAgent HubMLR Queue
ASST-0003Assistant · Classification outcome 3CAComplianceIn ReviewModerateAgent HubMLR Queue
ASST-0004Assistant · Review request 4UKAI OpsApprovedHighAgent HubMLR Queue
ASST-0005Assistant · Prompt run 5DEMedicalActiveCriticalAgent HubMLR Queue
ASST-0006Assistant · Grounding source 6FRComplianceBlockedLowAgent HubMLR Queue
ASST-0007Assistant · Classification outcome 7ITAI OpsEscalatedModerateAgent HubMLR Queue
ASST-0008Assistant · Review request 8ESMedicalDraftHighAgent HubMLR Queue
ASST-0009Assistant · Prompt run 9JPComplianceIn ReviewCriticalAgent HubMLR Queue
ASST-0010Assistant · Grounding source 10AUAI OpsApprovedLowAgent HubMLR Queue
ASST-0011Assistant · Classification outcome 11BRMedicalActiveModerateAgent HubMLR Queue
ASST-0012Assistant · Review request 12MXComplianceBlockedHighAgent HubMLR Queue

Relationship Graph

Linked entities are enabled so users can traverse parent-child dependencies and downstream impacts without manual lookup.

Agent HubPolicy Engine

Assistant Tools

Grounded Query

Run RAG query against approved sources.

AssistantClaims
Citation Inspector

Inspect citation/source usage and confidence.

ReferencesAudit
Execution StepOwnerETA
Validate inputOps5m
Policy checksCompliance8m
Route next actionWorkflow3m

Governed Assistant Output

Grounding sources include approved claims, PI/SmPC requirements, references, strategy docs, and previously approved assets. Responses include citations, confidence, output classification, request-MLR CTA, and prompt-output audit logs.

Classification: PromotionalConfidence: 0.91

Agent Hub

Model-agnostic registry, tool permissions, policy checks, and human-in-the-loop approvals are deterministic for local demo runs.

Model Registry

  • mdl-gpt-enterprise · OpenAI · high
  • mdl-claude-enterprise · Anthropic · high
  • mdl-local-llama · Self-hosted · sandbox

Sample Run

1. Input Loaded
2. Policy Guardrails Checked
3. Source Retrieval + Traceability
4. Human Approval Gate
5. Final Routed Artifact

Risk Flags: Missing fair balance in social template (resolved), off-label drift score 0.03 (pass), unapproved claim attempt blocked.

Human Approval Phase: Medical reviewer sign-off required before output release.

Final Artifact: Routed to `/mlr/queue` with note `Awaiting final promotional approval`.

Audit Entry: EVT-AI-2026-02-24-04122 recorded with tool permissions snapshot.

Output Classification: Promotional · Confidence: 0.91
  • AST-00431: HCP Efficacy Leave-behind v3
  • CLM-0214: Primary endpoint response claim
  • REF-0550: SmPC Section 4.8 safety reference

Agent Catalog

Market Landscape AgentBrand Positioning AgentCampaign Brief AgentSegmentation AgentClaims Librarian AgentClaim-to-Asset Traceability AgentFair Balance / Safety Presence AgentOPDP/Social Space Constraint Checklist AgentMLR Pre-Check AgentModular Asset Composer AgentVariant Generator AgentLocalization Brief AgentSEO/Discoverability AgentNext Best Action AgentJourney Builder AgentField-Digital Coordination AgentPerformance Analyst AgentContent ROI AgentExperiment Design AgentAttribution Insights AgentWorkflow Triage AgentAsset Reuse Finder AgentExpiry & Withdrawal AgentAgency QA AgentAdverse Event Detector AgentProduct Complaint Classifier Agent

Governance Notes

  • Every approval, publication, policy block, and AI output writes an audit event.
  • Consent and frequency caps are evaluated before deployment packaging.
  • Expiry and withdrawal controls maintain cross-channel evidence status.
  • Safety routing is always flag-and-route with human adjudication.

Geo Coverage

US1,820 active
UK640 active
DE550 active
JP730 active
BR500 active
AU320 active

Seed Coverage Summary

brands: 8campaigns: 30briefs: 200assets: 6,000components: 20,000claims: 2,500references: 8,000markets: 12localizationJobs: 1,500mlrPackages: 15,000reviewSteps: 60,000annotations: 200,000interactionsRollup: 2,000,000socialSignals: 50,000aeSignals: 1,500experiments: 5,000reports: 3,000connectors: 250auditLog: 300,000