Over the last year Head for Data have been building an AI-native support platform, integrating large language models into an operational workflow used by real engineers handling real customer requests.
The interesting part wasn’t choosing the model – it was everything else.
In this talk Colin Parry will walk through the engineering decisions that don’t make the headlines: protecting sensitive data before it reaches an LLM, designing reliable event-driven architectures, collecting user feedback, testing AI systems whose behaviour changes over time, and deploying changes safely into production.
Along the way he will share practical lessons from a real client engagement: including the mistakes they’ve made, what surprised them most, and why successful AI systems are usually won or lost outside of the model itself.
Whether you’re building AI products, leading technical teams or simply curious about what production AI really looks like, you’ll leave with a clearer understanding of what it takes to move beyond proof-of-concepts.
