Peter Elger talks with Sven Johann about AI as a service - commodity web services offered by large cloud providers, which make it very easy to use AI in your application just by using an API. They start the discussion with the typical AI use cases companies in finance, retail and so on have and which of those are already as a commodity available and which are not. They then discuss how those APIs are used, from very simple API calls like text-to-speech to slightly more complicated ones like chat bots or recommendation engines. They continue with understanding how you come up with AI requirements and how to (regression) test your AI service, which is often different than doing the same with non-AI code. They clarify how much AI you need to know to use those services, the cost models of them and how to get your AI to production. The conversation then moves to lower level AI services like AWS SageMaker or Google Tensorflow and how you combine them to create novel pieces of AI relatively quickly. They close the conversation with when to use specific packaged AI solutions and when to create your own AI to push the envelope.