Inspiration

Traditional Digital Adoption Platforms (like WalkMe, Whatfix, or Pendo) rely on rigid, hardcoded click sequences and screen coordinates. Whenever a SaaS team updates their UI, these tours break instantly and require constant engineering maintenance. We built AdoptionLayer to be the world's first AI-native Digital Adoption Platform (DAP): an autonomous, real-time interface guide that understands what is on the screen without brittle coordinate maps.

What it does

AdoptionLayer embeds into any web application or SaaS with a single lightweight <script> tag. Using advanced LLMs (Gemini, Claude, GPT) combined with a real-time live DOM parsing engine, it:

  • Dynamically analyzes the visible UI and interactive elements.
  • Highlights target buttons and triggers contextual tooltips.
  • Guides users through complex multi-step forms.
  • Answers user questions contextually right inside the host application.
  • Adapts automatically to UI changes, redesigns, and dynamic Single Page Apps (SPAs) with zero maintenance.

How we built it

  • Frontend Widget: Lightweight ES6/Vanilla JavaScript widget bundled with esbuild, built with strict host isolation, custom shadow styling, and a state machine.
  • Live DOM Parser: Compact semantic client-side parser extracting interactive elements, forms, and landmarks without screenshots or privacy leakage.
  • Backend & Guidance Engine: Ruby on Rails 8 API orchestrating LLM tool loops via Vertex AI (Gemini 3 Flash) and OpenRouter with strict server-side output policies.

Challenges we ran into

Building low-latency guidance that reacts in real-time to dynamic DOM modifications while guaranteeing 100% host page isolation and GDPR compliance.

Accomplishments that we're proud of

  • Delivering a sub-50ms lightweight embed widget.
  • Zero-maintenance guidance that doesn't break when CSS classes or DOM structures change.
  • A seamless developer experience with 1-line script installation.

Built With

Share this project:

Updates