At Google's cloud conference, customers revealed the primary barrier to AI adoption is implementation complexity and "agent sprawl." While AI can accelerate discrete tasks, companies struggle to overhaul entire workflows. This creates new bottlenecks, as the tools' complexity outpaces firms' ability to integrate them.

The biggest resistance to adopting AI coding tools in large companies isn't security or technical limitations, but the challenge of teaching teams new workflows. Success requires not just providing the tool, but actively training people to change their daily habits to leverage it effectively.

Unlike software with discrete feature releases, AI capabilities are updated continuously in the background. Product teams must build mechanisms to constantly re-educate users on what the tool can now do, as its evolution is invisible to them and requires overcoming the 'blank page problem' repeatedly.

Despite the hype, advanced AI tools like autonomous agents won't reach scale in enterprises until late 2027, lagging startup adoption by 2-3 years. Even non-technical departments at major tech companies are still focused on basic chatbot usage, highlighting a significant gap in adoption speeds.

The constant leapfrogging between AI labs and shifting architectural paradigms makes enterprise teams hesitant. They fear backing the wrong technology and getting locked into a strategy that will soon be deprecated, leading to inaction.

AI agents like ChatGPT Work, built on coding assistant frameworks, are "software-brained." They focus on delivering a final product, ignoring the crucial iterative process of research, exploration, and learning that defines most knowledge work, creating a fundamental disconnect for non-coders.

By sunsetting easy-to-create Custom GPTs for a more complex developer-centric plugin system, OpenAI is sacrificing usability for a key user segment. This decision makes AI customization less accessible for non-technical professionals who relied on the simple, natural language interface of GPTs.

The transition from simple chatbots to autonomous, 'agentic' AI tools is a massive, underestimated leap. These non-deterministic tools carry risks of unintended consequences, like deleting files or codebases, and the average knowledge worker lacks the skills and mental models to manage them safely.

OpenAI's research shows a significant capabilities gap. While adoption is high, most workers use basic features like writing and search. Technical "power users" leverage advanced functions like custom GPTs, indicating a major need for company-wide training to unlock full productivity potential.