- Models. Whether it’s a large language model predicting tokens or a machine learning model forecasting sales, you need something that captures how the world behaves. Some capabilities and tools for this component include workflows, processes, algorithmic system models, rule-based models and embedded AI applications within vendor products.
- Data. Models learn from data. If you’re using a pretrained model as is, great. If not, you’ll need to supply your own data to make it relevant to your context. Some capabilities and tools for this component include storage, compute, governance, observability and machine learning operations.
- Context and ecosystem. Foundation models don’t automatically know your business, systems or what tools they can use. You must teach them to operate within your world. Some capabilities and tools for this component include partners, application programming interfaces, agentic and other frameworks, and Model Context Protocol or other protocols.
- Compute. None of it works without scalable computing to power inference and automation. Some capabilities and tools for this component include cloud or on-premises compute, memory storage, CPUs and graphics processing units.
