For decades, firms have quietly built and refined proprietary machine learning algorithms, models tuned to their own data, their own risk appetite, and their own edge. Then large language models arrived, general purpose, fast to deploy, and trained on the whole internet rather than a firm's private history. The two approaches were built for different problems, yet increasingly they're being asked to work side by side, or even replace each other. This raises a question that's no longer theoretical: should LLMs sit on top of existing systems as an interface layer, work alongside them as a second opinion, or in some cases replace them outright? Each path carries real trade offs, in explainability, latency, cost, and how much control a firm retains over its own models. And the stakes extend past any one firm's balance sheet. When enough market participants lean on similar general purpose models, questions about correlated behavior and systemic risk stop being hypothetical too. This conversation matters to practitioners deciding what to build next, to firms weighing where their competitive edge actually lives, and to anyone thinking about what market stability looks like when more of the infrastructure underneath it starts to look the same. There's no consensus answer yet, which is exactly why it's worth having the conversation now.

Time-to-value is the metric that kills more PLG ambitions than any other. Teams obsess over sign-up flows and onboarding sequences, but the real problem is almost always deeper: they have not defined what the first win actually looks like for their user, or they have defined it in terms of their own business goals rather than the user's job to be done. This panel brings together practitioners who have rebuilt activation from the ground up and are willing to talk honestly about what they got wrong first. Panelists will share specific examples of activation changes that moved the needle, the data they used to identify the problem, and the organisational friction they had to overcome to fix it. Expect a practical, example-driven conversation rather than high-level principles. This is a session for product people who are in the weeds on conversion and want real benchmarks and approaches to take back to their teams.