AI Guardrails and Evals | 7th Chicago Meetup

​Most testers use AI coding tools in a familiar way: open the chat, ask for something, copy the result, and move on. ​That works, but it is only one way to use them. ​At The Test Tribe’s 7th Chicago Meetup, Maneesh Maddala will explore how Skills and plugins can make AI more repeatable by packaging instructions, context, and tooling around everyday testing tasks. Session: Guardrails & Evals ​Speaker: Maneesh Maddala, Principal QA at Morningstar ​As AI becomes part of software products and engineering workflows, how do we make sure it behaves as expected? ​This session explores two important aspects of working with AI systems: guardrails and evaluations (evals). It looks at how teams can approach defining boundaries for AI behaviour, assessing outputs, and thinking about quality beyond whether an AI-generated response simply works. ​The discussion will also touch on writing guardrails and evaluations, and how these approaches fit into testing AI-powered applications. ​What you'll take away: * ​Understanding guardrails and evals: Explore their roles in managing and assessing AI system behaviour. * ​Defining expected AI behaviour: Understand how guardrails can help establish boundaries for AI-generated responses. * ​Evaluating AI outputs: Explore how evaluations can be used to assess whether AI systems meet defined expectations.