A few thoughts on enterprise adoption of Generative AI in 2025
Here are four thoughts on enterprise adoption of Generative AI in 2025 - would love to hear your perspectives and what I might have missed!
1. GenAI features will continue to be integrated into many vendor products and services. The governance and control mechanisms that many organizations have been putting in place to manage associated risks will be tested in 2025 and beyond.
2. Gradual and cautious adoption of GenAI based task assistants/copilots will continue - e.g., helping with coding, testing, writing, summarization, analysis, presentation, visualization, etc. I should mention the continued adoption of chatbots in support as well.
3. Agent development and adoption will continue to be slow as developers continue to grapple with the lack of reliability of the underlying models, in addition to other challenges such as explainability/transparency, data quality, privacy, security, and regulations. Building guardrails and controls around LLMs is like chasing a moving target given the rapid evolution of and growth in model capabilities. In summary, I would be (pleasantly) surprised to see significant agentic implementations in critical business functions.
4. The higher costs and the types of problems that advanced reasoning systems like o3 can solve mean that they are likely to be used for high-end scientific research before they are used for seemingly ‘simple’ day to day applications. This follows Moravec’s paradox: what’s simple for humans is often very hard for AI and vice versa.