Everything you need to know about Tobie

Tobie is a benefits communication operating system. Send us one benefits guide and we turn it into a branded employee microsite, a source-grounded AI assistant, branded video content, multi-channel campaign messaging, and a full admin oversight dashboard, all from your existing documentation.

Every answer the AI gives traces back to your approved materials. This isn't a generic chatbot, it's a system that stays within the boundaries of what you've approved. That means credible answers, auditable sources, and no hallucinated benefits information.

HR teams, benefits brokers, and consultants who manage benefits communication for mid-to-large employers. If you're sending employees dense PDFs and fielding the same questions every enrollment season, Tobie replaces that with a structured, governed experience.

The assistant is intentionally narrow. It only answers from approved source documents, refuses when evidence is insufficient, and escalates to HR when the question falls outside its boundaries. Every answer includes source citations that admins can verify. No internet access, no improvisation.

A branded benefits microsite with organized plan information, provider directory, FAQs, and an embedded AI assistant. It replaces dense PDFs with clear pathways, employees can browse by category or ask questions directly.

Plan summaries, definitions, deadlines, eligibility language, cost-sharing breakdowns, provider contacts, and policy explanations, anything that's clearly documented in the approved benefits materials.

No. The system does not make recommendations, provide investment advice, or suggest personal benefit choices. It can compare documented differences between options, but it will not cross into personalized guidance.

The assistant says so. It will refuse to speculate, ask for clarifying details, or direct the employee to HR, the carrier, or another official support channel. This is a core trust feature, not a limitation.

No. The system is retrieval-only from approved sources. It never searches the web or draws from general knowledge. This closed-corpus design is what makes every answer auditable.

The system retrieves approved source material first, builds a bounded evidence set, generates a draft answer, then verifies the answer is actually supported before showing it. If evidence is insufficient, it refuses rather than guessing.

Every claim the AI makes is tied to specific pages, sections, and document versions from your approved materials. The source trail is stored for every conversation, admins can inspect exactly where any answer came from.

The employee-facing chat stays clean and simple. Source tracking happens in the background and is available in the admin dashboard for review, audit, and export.

Medical advice, legal advice, tax advice, investment guidance, plan-selection recommendations, unsupported eligibility determinations, and any attempt to get the AI to ignore its source boundaries.

The system discourages employees from sharing SSNs, claim numbers, or medical records. Transcripts are stored with controlled access, and the design follows HIPAA minimum-necessary principles for all data handling.

Yes. Every conversation is logged with timestamps, source evidence, confidence scores, and outcomes. This is a core feature, every answer the AI gives can be reviewed after the fact.

Access is role-based. Employees use the assistant. HR admins use a separate dashboard to review conversations, inspect source evidence, search transcripts, and export records. All admin access is logged.

Conversation lists, individual transcript views with source citations, search across all conversations, export for audit, escalation workflows, and analytics on conversation volume and topics.

Separate employee and admin experiences, role-based access controls, encrypted storage, append-only audit logs, and monitoring on all admin actions. HIPAA-aware, SOC 2 ready architecture.

Retention is policy-driven and configurable per client. Transcript storage, export, deletion, and legal-hold behavior are explicit product settings, not left undefined.

Yes, that's the whole point. The system is designed around your company-provided benefits guide and supporting documents. The ingestion process preserves tables, footnotes, page references, and versioning.

Absolutely. Send us one benefits guide and we'll return a working demo within 48 hours showing the employee experience, AI guardrails, and admin review model. Currently offering limited pilot pricing.

Source document ingestion, a branded employee-facing microsite, the bounded AI assistant, transcript logging, and an admin dashboard for conversation review, search, and export.

Tobie is not a claims system, eligibility engine, legal interpretation tool, or substitute for formal plan documents. It makes approved benefits information easier to find and easier to audit.

Clear employee UX, disciplined answers, visible guardrails, reviewable transcripts, strong source traceability, and a security posture that's serious from day one. Not just that the AI answers questions, that the experience is credible to a buyer.

A production path adds identity and access integration, customer-specific policy review, finalized retention rules, operational monitoring, formal evaluation gates, and a tighter document-governance process around source updates.