## 1. The Build Architecture (Zero Runtime AI) > [!NOTE] > The deployed app contains no runtime AI — the AI was used exclusively at build time to synthesize dense CRM documentation into a single static SPA. This was a deliberate choice for performance, offline capability, and zero server cost. Rather than building a standard chatbot that queries a database at runtime, I used generative AI to synthesize our complex Dynamics 365 process documentation into a self-contained, standalone web application. Here is the workflow used to generate the simulator: ```mermaid graph TD A[(Dynamics 365 Knowledge Base)] -->|Exported PDFs & Policies| B(Context Preparation) B -->|Massive structured prompt| C{AI Code Generation} C -->|Synthesizes knowledge & logic| D[Standalone HTML/JS App] D -->|Deployed internally| E(Gamified Simulator) style C fill:#0f172a,stroke:#38bdf8,stroke-width:2px,color:#fff style A fill:#0f172a,stroke:#facc15,stroke-width:2px,color:#fff ``` ### Prompt Engineering for App Generation The core technical challenge was not runtime AI hallucinations, but rather structuring an incredibly dense prompt that forced the AI to generate a functional, gamified Single Page Application (SPA) in one shot. I designed a rigorous system prompt that dictated strict technical constraints, state management, UI design tokens, and exact information architecture derived from our CRM. Here is an excerpt of the system prompt used to orchestrate the build: ```markdown # Project: IE Student Services Onboarding Companion ## Role and goal You are a senior frontend engineer with strong design sensibility. Build a **single-page web app** that serves as an onboarding companion for new advisors joining the Immigration & Relocation team... ## Technical constraints (non-negotiable) - One self-contained `.html` file. No build step, no frameworks, no external dependencies. - Vanilla HTML + CSS + JavaScript only. - Persistence via `localStorage` under the key `ie-onboarding-v1`. Migrate any legacy state shapes. - Hash-based routing: `#home`, `#tasks`, `#progress`, `#roleplay`, `#guide/{id}`... ## Information architecture Five visible categories on the tasks dashboard, plus a resources bucket. Each guide has `{ id, title, icon, category, subtasks[] }`. Only subtask checkboxes count towards progress. 1. **Essentials for arrival** 🎯 — health insurance, public transport... 2. **Housing** 🏠 — long-term accommodation and temporary accommodation... 3. **Immigration & paperwork: Non-European students** 🛂 — padrón, fingerprinting, NIE card... ... ``` By heavily constraining the AI—forcing vanilla JS, local storage for persistence, and hardcoding the institutional design system—I ensured the output was a predictable, highly performant tool that required zero server maintenance. --- ## 2. Executive Summary The Problem Standard onboarding involved dense PDF guides for complex Dynamics 365 processes, overwhelming new hires and consuming hours of senior staff time for 1-on-1 teaching. The Solution A gamified, interactive web app powered by Microsoft 365 Copilot Cowork, leveraging our internal Dynamics 365 knowledge bases to create self-paced roleplay simulators. ### Business Impact - **Reclaimed Senior Staff Time:** ~1 hour/day reduction in 1-on-1 training per new hire, based on advisor self-reports during the first three months post-rollout (n=4). - **Accelerated Time-to-Value:** Shortened the standard onboarding cycle by weeks, allowing new hires to operate independently much faster. - **Live Reference Tool:** The app acts as an interactive reference guide that advisors actively use while on calls with students, reducing interruptions and escalating issues. --- ## 3. Product Demo: The Simulator *Experience the interactive gamified onboarding process below:*

--- ## 4. The Outcome By intelligently applying AI at build-time rather than runtime, we modernized our operations and proved the tangible ROI of AI adoption in the workplace without incurring recurring API costs. Instead of feeling overwhelmed, new hires found the onboarding process refreshing and self-driven, allowing our job specialists to pivot their focus back to high-value, strategic tasks.