Users are incredibly creative when it comes to breaking apps. So we're going to take a leaf out of the book of one of the most under-appreciated characters in his world, Wreck-It Ralph. After all, if Ralph didn't break things, there wouldn't be much of a game. And without proper QA, there won't be much of a business solution either.

Except now we've handed part of the wrecking over to the agents. A Copilot doesn't fail neatly. It waffles, it hallucinates, it confidently reaches for the wrong tool, and it gives you a slightly different answer every time you ask the same question. "Did the button work?" is a much easier question than "was that answer good enough, often enough?"

Together we'll channel our inner Ralph and learn how to put our apps and our agents through their paces before they reach the real final boss, the customers. Because users will stress-test anything you give them in ways we couldn't possibly imagine, and a chat box is an open invitation.

Join this fun and slightly chaotic tour through testing in the Power Platform. We'll explore the realms of functional, integration, regression and non-functional testing, then carry that thinking into agentic territory: building test sets and expected responses, grading non-deterministic answers on quality, groundedness and completeness rather than exact wording, running multi-turn conversations, and validating that your agent's plan actually picked up the right tools instead of going full Turbo and wandering off into someone else's game.

Along the way we'll revisit a few testing horror stories (yes, including test scripts in Excel 😱), discuss how to advocate for better testing and evaluation during sprint planning, and highlight some platform "power-ups" that help analyse solutions, trace issues and simulate user interactions: Test Engine, the activity map, the Copilot Agent Kit, agent evaluations in Copilot Studio, and wiring evals into your pipelines as a quality gate so nothing ships on vibes alone.

Packed with practical tips and relatable maker pain, this session will leave you with a new appreciation for breaking things on purpose, because sometimes the best thing you can do for your users is be a little more like Ralph.