Abstract
While AI is promoted as a useful tool, its wider implications are important, under-recognized and contested.
Complex Customer Relational Management (CRM) systems can be difficult to use, especially for many older users. DealCompass seeks a more usable system. To reduce this cognitive load, an LLM-powered chatbot was prototyped to abstract complex backend operations into simple, natural language commands. Using Python, Uvicorn, and Llama 3.2, the system executes basic CRUD operations for key database tables. This streamlined a multi-click, multi-field manual entry process into a two-step interaction, improving LLM JSON accuracy from 20% to 80%. With further work, the methods can be made more robust, schema discrepancies can be resolved, and the chatbot could handle complex, multi-tool queries.
The fight to influence AI’s development is not a binary issue, but a complex competition over policy. Big Tech is winning this policy war not by defeating regulation, but through legislative loopholes. Tech leaders and other proponents of AI characterize AI development as a geopolitical competition and use their political influence to promote this framing. Their critics, however, citing civil rights, demand informed consent. They compete for influence in Congress, state legislatures, and regulatory agencies. Through its overwhelming influence, Big Tech marginalizes its critics. The consequence includes state laws such as the Texas TRIAGA Act, which offers perfunctory ethical guardrails while opening “regulatory sandboxes” to unfettered AI experimentation.
