Comparison Tables

This page consolidates the key comparison tables referenced throughout the documentation into a single reference. Each table links out to the relevant deep-dive page.

AI Desktop UI Automation Techniques & Tools

A comparison of the automation techniques the PHX Terminal platform draws on, adapted from the Technical Specification. See AI Computer Vision, RPA & Intelligent Automation, and LLMs & NLP for full detail.

Technique / Tool	Description	Strengths	Limitations	Relevance to Platform
RPA (Traditional)	Automates repetitive tasks by mimicking user interactions at the UI level.	Improves efficiency and accuracy for rule-based tasks; quick to deploy without altering underlying systems.	Most effective only with structured, predictable data; cannot read unstructured data; does not learn or improve on its own.	Foundation for basic, repetitive data-entry automation, especially structured forms.
AI-Powered RPA (e.g., UiPath)	Combines traditional RPA with AI (ML, LLMs) for complex, intelligent tasks.	Processes unstructured data; enables decision-making; extends automation to new areas; improves accuracy.	Requires training data; can be complex to implement for highly nuanced tasks.	Essential for processing diverse legal documents (contracts, emails) and extracting data for the hover opaque application.
AI Computer Vision (e.g., UiPath CV, OmniParser)	Enables robots to “see” and interact with UI elements like humans, using neural networks, OCR, and text matching.	Technology-agnostic across desktop environments (.NET, Java, VDIs); adapts to UI changes in real time; reliable where selectors fail.	Requires robust training datasets; potential ambiguity with repeated UI elements or coarse bounding boxes.	Core technology for the hover opaque application to identify fields and interact with any desktop application regardless of technology or age.
NLP / LLM Integration for UI	Uses NLP and LLMs to understand user intent and interface context.	Interprets human language contextually; understands nuance; can predict user needs and generate content proactively.	Sensitive to prompt quality; can produce incorrect or biased information if not constrained; resource-intensive.	Enhances user-intent recognition, letting the hover app anticipate lawyer needs and suggest data entry.
Accessibility APIs (e.g., Microsoft UI Automation, Apple Accessibility API)	Provides programmatic access to UI elements and properties, designed for assistive technologies.	Enables robust “opaque box” interaction regardless of application internals; supports cross-app UI testing; consistent across platforms.	Platform-specific (separate Windows/macOS implementations); may not expose all UI elements/properties.	Reliable programmatic layer for interacting with standard UI controls, complementing vision-based approaches for precise data injection.

Marketplace Revenue Models

Common revenue models evaluated for the PHX Terminal marketplace. The platform is expected to lead with a commission-based model, optionally combined with freemium or featured listings. See Marketplace & App Review and Business Model.

Model	How It Works	Pros	Cons
Commission	Charges a percentage or fixed fee per transaction (e.g., Salesforce AppExchange charges ~15% revenue share).	Most popular; attracts suppliers who pay only on a sale; aligns incentives.	Revenue depends on transaction volume.
Subscription	Users pay a recurring fee for platform access.	Predictable recurring revenue.	Requires consistent value to justify the recurring fee.
Listing Fee	Sellers pay to post listings/advertisements.	Revenue from all listings.	Worsens the “chicken-and-egg” problem; sellers pay before profit.
Freemium	Basic services free; users pay for premium features.	Fast lead generation.	Challenge of converting free users to paying ones.
Featured Listings / Ads	Sellers pay for enhanced visibility.	Additional revenue stream.	Can be difficult to monetize effectively.
Lead Fee	Users post requests; suppliers pay to bid.	Better value than listing fees.	Requires high-value leads.
Mixed Models	Combine several of the above.	Multiple revenue sources.	Added complexity to manage.

Cloud Deployment Options

PHX Terminal uses a hybrid model, placing each workload where it best balances security, compliance, and cost. See Hybrid Cloud & Serverless.

Option	Best For	Trade-offs
Public Cloud	Scalable compute, AI workloads, marketplace operations, general platform services, less-sensitive/aggregated data.	Less direct control over data residency for sensitive information.
Private Cloud / On-Premises	Sensitive client data, attorney-client privileged communications, jurisdiction-specific compliance, high-security enterprise environments.	Higher fixed cost; less elastic than public cloud.
Serverless	Event-driven microservices, fluctuating workloads (e.g., document-processing spikes), cost optimization.	Pay-as-you-go; tied to event-driven design patterns.