Brex initially invested in a sophisticated reinforcement learning model for credit underwriting but found it was inferior to a straightforward web research agent. For operational tasks requiring auditable processes, simpler LLM applications are often superior.

For complex cases like "friendly fraud," traditional ground truth labels are often missing. Stripe uses an LLM to act as a judge, evaluating the quality of AI-generated labels for suspicious payments. This creates a proxy for ground truth, enabling faster model iteration.

At Qualtrics, the text analysis platform initially relied on complex syntactic and keyword-based rules. The advent of transformer models like BERT and their powerful embeddings rendered these older techniques obsolete, representing a fundamental paradigm shift in natural language processing capabilities.

Building reliable AI agents for finance, where accuracy is critical, requires moving beyond pure LLMs. Xero uses a hybrid system combining LLM-driven workflows with programmatic code and deep domain knowledge to ensure control and reliability that LLMs inherently lack.

LLMs fail at core enterprise tasks like demand forecasting on structured data. SAP is developing "Relational Pretrained Transformers" (RPTs) to apply the foundation model concept to tabular data. This aims to democratize predictive modeling, which currently requires specialized data scientists and doesn't scale.

Standard LLMs fail on tabular data because their architecture considers column order, which is irrelevant for datasets like financial records. LTMs use a different architecture that ignores column position, leading to more accurate and reliable predictions for enterprise use cases like fraud detection and medical analysis.

IBM's CEO explains that previous deep learning models were "bespoke and fragile," requiring massive, costly human labeling for single tasks. LLMs are an industrial-scale unlock because they eliminate this labeling step, making them vastly faster and cheaper to tune and deploy across many tasks.

The concern that AI will surface the same deals for everyone is unfounded. A competitive edge comes from using a complex infrastructure with multiple, specialized Large Language Models (LLMs) for data extraction, validation, and structuring. This sophistication ensures a differentiated output compared to simpler AI tools.

The era of simply scaling up Transformer-based models is ending. AI21's Jamba model, which combines Transformer and Mamba architectures, points to a new innovation wave focused on hybrid designs. This shift aims to improve efficiency and specialized capabilities like long-context processing, moving beyond the 2017 paradigm.