Open source small language models offer organisations a strategic path to building AI agents that avoid vendor lock-in, enable explainability for regulated industries and provide operational independence from the three dominant LLM providers.