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.
Because LLMs are non-deterministic like humans, it's more effective to integrate them using existing human-centric processes. Give an agent an email, permissions, and "onboarding" so it can navigate the organization like an employee, rather than building complex new software interfaces.
A major architectural shift is underway: instead of embedding AI features into a product, companies should treat AI as an external agent that uses the product via a CLI or API. This simplifies integration and better aligns with AI's capabilities.
The idea that AI code generation reduces demand for engineers is backward. It leads to more complex systems, which in turn creates more challenges around system upgrades, downtime, and security incidents, ultimately requiring more engineering oversight and expertise.
When a major platform like Salesforce prioritizes headless APIs, it's a bellwether moment. It signals a recognition that AI agents will become primary "users," driving demand for API-first access and creating a new wave of automation use cases.
The divide is not just about enterprises being slow. It's because engineers in the Valley have high technical aptitude, can debug their own tools, and work with verifiable code—conditions that don't exist for most knowledge workers.
When boards pressure CEOs for AI, the result is often a centralized, consultant-led project disconnected from operations. These initiatives fail because they lack alignment and nobody understands how they work, creating skepticism for future efforts.
While headless APIs are ideal, many websites and apps actively block headless browsers to prevent scraping. This forces AI agents to interact with the standard graphical user interface to complete tasks, just as a human would, rather than relying on APIs.
An AI agent cannot simply use a human's credentials. It requires its own identity, permissions, and access controls for security and traceability. This means SaaS companies will likely charge for agent seats, creating a significant new revenue stream.
When companies measure AI adoption by counting tokens used, it creates a perverse incentive. Employees and their teams create agents to perform pointless tasks simply to boost their metrics, leading to fake productivity and problematic artifacts.