Cisco launches Motific hub to streamline generative AI deployment

Cisco is delivering a cloud-based service that gives enterprise customers a centralized hub for managing generative AI elements such as large language models (LLM), security controls, APIs and more. Called Motific, the service promises to help streamline and accelerate the creation, deployment and management of generative AI-based applications for the enterprise.

“Think of Motific as a one-stop shop where a ton of personas – large language models, ChatGPT etc. – come together with enterprise organization-specific policies and data sources to make it easier to create and deploy generative AI applications in a safe, secure manner,” said Vijoy Pandey, senior vice president of Cisco’s advanced research outfit Outshift. 

Organizations want to standardize and templatize the commonly used building blocks of genAI applications, such as assistants, APIs, foundation models, knowledge bases for retrieval-augmented generation (RAG), and policy controls, and Motific addresses those interests, Pandey said. RAG is an AI framework that typically increases the quality of LLM responses by including resources from a variety of locations in order to reinforce the accuracy, credibility, and contextual relevance of those responses.

Once in place, Motific discovers any generative AI resources in the enterprise and lets customers set up patterns and policies for specific applications. “For example, if a customer wanted to create a financial assistant for the CFO, they create the motif that would define the policies, identity and access controls they would need, as well as the LLMs available to them to develop that assistant, and Motific would control and monitor that development, Pandey said.

The idea is to reduce genAI application development time from months to days, Pandey said. Motific offers compliance controls for over-usage, overrun spending, and integration of organization-specific data sources, he said. It also tracks business processes and prompts usage intelligence with ROI and cost analysis, including consolidated monitoring audit trails and key metrics tracking of all user requests.

“Motific is using a ton of generative AI technology under the hood itself. For example, it uses a ‘prompt processing unit’ to provide insights that keep AI hallucinations and toxicity in the development process to a minimum,” Pandey said.

Enterprises can use Motific’s built-in policy controls – for handling compliance, sensitive data, security and access, for example – or customize them based on their own internal policies. On the security side, the first release of Motific implements the Open Worldwide Application Security Project (OWASP) tools and technologies for securing applications.

“Motific will let security teams know if there is a security problem and will point out where this is happening, then let them quarantine the impacted attack path,” Pandey said. “The system can also generate remediation code to help fix the issue.”

Motific will be generally available by June.