Prototype Exploration

Two Paths to AI-Powered Onboarding

Compare LLM-assisted workflows (user-driven) with Agentic AI (goal-driven, autonomous). Click into each to explore.

🧭
Current Scope

LLM-Assisted Onboarding

A guided wizard where you control progression. AI extracts data and suggests mappings, but you decide each step.

  • → Predefined flow — 7 fixed screens
  • → User-controlled — Click "Next" to advance
  • → AI assists — Extraction, suggestions, validation
  • → Explicit confirmation — Review every object
Open Wizard Prototype →
🤖
Future Vision

Agentic AI Onboarding

Give the agent a goal and documents. It plans, executes, validates, and adapts autonomously. You oversee and approve.

  • → Goal-driven — "Onboard Houston FCC" is the input
  • → Autonomous planning — Agent decides the steps
  • → Tool/API use — Calls OA APIs, validates, retries
  • → Human oversight — Approve checkpoints, not clicks
Open Agentic Prototype →
Side-by-Side Comparison
Dimension LLM-Assisted Agentic AI
Input Form fields + uploads per step Goal statement + documents upfront
Control Flow User advances through fixed screens Agent plans and executes; user oversees
AI Role Suggests, extracts, validates on request Reasons, acts, adapts, retries autonomously
Tool Use Backend processing, no visible API calls Visible tool calls: OA API, validation, parsing
Error Handling User sees errors, manually fixes Agent detects, reasons, retries or asks
UX Pattern Wizard / Stepper Chat + Activity Log + Approval Gates
Trust Model Confirm every step Approve at checkpoints, audit anytime