What special considerations are needed when implementing Generative AI tools at the enterprise level?
Sort By:
Oldest
Chief Strategy Officera year ago
Would like to know the views of other leaders in the community about the issues faced and mitigating actions undertaken in their respective field of work ? Associate Director of Data Science & Analytics in Healthcare and Biotecha year ago
Cost. Training these models is extremely expensive. A large internal LLM can cost tens of millions of dollars in compute. And that is assuming your organization has the proper infrastructure to support it, and the right expertise to optimize it. CIO in Insurance (except health)7 months ago
For organizations operating in highly regulated environments (e.g. healthcare), governance and compliance is a critical consideration.For all enterprises, training at all levels is critical. These tools require a fundamental shift in thinking especially critical thinking. Developing skills in prompt engineering, Gen-AI oversight, specifically how to detect when the LLM/tool is straying from the task or hallucinating.
VP - Chief Architect in Banking7 months ago
Every organization is a mix bag of vendor supplied technology and home-grown stuff. When it comes to vendor saturation (e.g. supplied platforms/tools/apps) we're soaking at around 90%-95%. The majority of our tech vendors are planning to implement some form of Generative AI in their next release. So, the consideration becomes... how to manage the saturation of AI capabilities across the vendor technology landscape. All require one critical thing to work.... data. It's more than policy controls or contract management. It's like we need AI to solve for the AI saturation. LOL
Business and technology leaders will need to collaborate and strategize on the best approach which is ‘human-centric’ :
Focus on and address questions of responsible and ethical use of AI (e.g., data bias, proprietary information, etc.)
Decide whether you will use large publicly available models with some APIs and consideration towards cybersecurity or initiate smaller very niche models for specific use cases which are fine tuned to meet your requirements.
Include the team by educating and clarifying that this implementation will augment but not replace human effort. The idea is to empower the employees and serve the stakeholders better more effectively and efficiently.