Large language models (LLMs) are powerful but they are not infallible. Like all technology, they have limitations and weaknesses that compound as data sources become more complex and volume increases. Insights teams using LLMs for insight generation need to be aware of these factors as ignoring these limitations pose a serious risk: outputs become irrelevant, unverifiable and impossible to act on – or worse, inaccurate insights are shared within the business. On Thursday, July 30, 2026, Justin Rehayem, Senior Director of Spotlight Delivery and specialist in omni-channel AI solution architecture, will explain the critical LLM limitations insights teams must account for and how to successfully integrate LLMs into their workflows.
In this session, you’ll learn:
- Five limitations of large language models that must be taken into account when using them for insight generation.
- How to overcome the key limitations of LLMs when analysing customer feedback at scale to prevent hallucinations and analysis bias.
- How to use LLMs to enrich unstructured customer data with AI-generated tags, themes, sentiment and metadata to unlock deeper and more actionable insights.
Justin will showcase that when AI is applied within its boundaries, organisations can consistently produce insights that are verifiable, actionable and accurate, even across the most complex listening signals.