I work for an organization with over 10k employees. I want to educate them on prompt engineering and risks associated with AI usage (e.g., IP considerations, training on company data, hallucination, etc.). What tools are other organizations using or are you building content in-house?
Sort by:
There are plenty of free resources (like on-demand webinars) and trainings provided by various vendors and hyperscalars around responsible AI policies and considerations, and also for prompt engineering. You could review them and decide if they are sufficient to train the employees. However, guidance or your company's policy around Responsible AI and AI usage would then need to be drafted based on your business domain, so that the employees have clear guidance on what they can do and what they shouldn't do with AI, to comply with your industry and local regulations. To support your policies and trainings, you also need an organizational structure (like an AI Governance Council) that would be responsible for your AI policies and acceptable usage. Hope this helps. Glad to have a chat if that helps.
I agree with Goldy Aloyslous, having a Gen AI council to start looking at governance is important. Creating some type of community page on SharePoint or other similar service for cross pollination of use cases and ideation is also important.
As we are still very early in the lifecycle of AI development there is risk that needs to be managed well. I recommend looking at what tooling your company is using then look to see what they have, for example we are using Azure Open AI and we have the ability to monitor for "undesired" searches such as self-harm, hate speech etc.
I think over time the tooling will develop to have larger and larger contextual abilities natively however until that day you may need to look into RAG architectures as these open up a lot more use cases for your organization with more inherent context into your company. Consider what data sources should be accessible to AI such as company policy, financial standards, competitive landscapes.
Training will be very important consider creating a video training series with topics such as appropriate use of company data, Hallucination, "Human in the loop", prompt engineering here is a good guide: https://github.com/brexhq/prompt-engineering.
One other point. I think the governance committee needs to balance the risk with the opportunity. Consider having a centralized generalized chat tool for the whole company, that way you can control how the data entered into this chat tool is controlled and safeguarded but, recognize that this tool will not serve all use cases, and they may need to approve specialized tooling for specific teams. For example, a chat bot for legal review or something for generating photo and video content might need a more specialized approach and you won't be able to develop an internal tool for all these use cases and keep pace with the market.
Make sure to focus your efforts for any custom development on what is core to your business.