How does your company plan on investing in generative AI over the next 3 years? In which ways has your company implemented AI? Model Investment Deploying nd Customizing AI Use Cses Just as the leading preferred models leading commercial closed-source improved operational efficiency Coding copilots are becoming Organizations can optimize genera- (RAG) - enhancing the model’s are closed-source commercial models are closely tied to leading is the key driver behind adopting mainstream with technical users tive AI models for specific use cases knowledge by integrating infor- models, planned investments in cloud-service providers. Enter- generative AI. Generative AI use being early adopters of solutions through the following techniques: mation from external sources these categories of models reflect prises can draw down from cloud cases reflect this shift in priorities. like GitHub Copilot, CodeLlama, during the generation process. usage trends. 72% of organizations spend commitments through use of The leading use-cases for generative and Devin. Model vendors have • Prompt-engineering - guiding plan to increase investments in partner models (e.g., Amazon and AI adoption are computer program- responded to demand for content the model’s output through Teams are likely to maximize their commercial closed-source models. Anthropic, Microsoft and Open AI). ming and content generation.. generation with prompt templates carefully crafted input prompts AI investments by adopting these A lower percentage of organiza- that guide users to effective content techniques. For organizations that tions plan to invest in open-source Last year, organizations referenced creation questions for functions • Fine-tuning - training the model already fine-tune their own models, models at 67%. While open- the ability to develop new products including Marketing, Product Man- on domain-specific data 39% saw improved performance on source models provide organiza- or services as the leading reason agement, and Public Relations. domain-specific tasks compared to tions with greater control, many to adopt generative AI. This year, • Retrieval-Augmented Generation out-of-the-box models. 18 19