![Cisco ACI vs. Cisco DNA Virtualization [EOL] report](https://www.peerspot.com/assets/media/images/icons/pdf-c33db68f.svg){kind=icon}
I believe Cisco ACI could improve by incorporating something related to agentic AI, where it can execute tasks based on particular evaluations or loop engineering, enhancing fault identification or suggestions based on each tenant's configuration. Additionally, I think they should ensure that the documentation is comprehensive. I appreciate that Cisco provides documentation, and their support engineers are helpful in case of any issues. In terms of governance and security regarding Cisco ACI's AI capabilities, I think what we call evals or evaluations is important, whereby the engineer can evaluate the functionalities of the AI in Cisco ACI, preventing it from making deletions or modifications on its own. Therefore, there should be evaluations or guardrails incorporated into the development capabilities. Regarding accuracy and reliability of output for Cisco ACI's AI capabilities, it hasn't met my expectations. What we need is something called RAG, or retrieval-augmented generation. If Cisco can implement that to utilize customer-specific data instead of general knowledge, while ensuring data privacy under GDPR laws, I think that will result in very good, reliable output.
