AI's Leap: From Assistant to Executor in Enterprise

The Agentic AI Frontier

The report compellingly illustrates the accelerating adoption of AI agents in enterprises, moving beyond simple assistance to actual task execution. The introduction of 'frontier firms' as a metric for deep AI integration is particularly insightful, highlighting a significant performance gap (8.3x output tokens) compared to typical firms. This gap is directly linked to the adoption of advanced capabilities like Plugins and skills, which connect agents to company context and tools. The data showing rapid growth in agentic AI usage across diverse knowledge work functions, such as legal and sales, is a strong indicator of its broad applicability. Furthermore, the finding that early-career employees are leading AI adoption challenges conventional wisdom and presents a clear opportunity for knowledge transfer within organizations. The emphasis on connecting agents to context, tools, and establishing robust governance is a practical and actionable agenda for any enterprise looking to scale AI effectively.

However, a key limitation is the reliance on 'output tokens' as the primary proxy for 'depth of use.' While a useful metric for the volume of work, it doesn't fully capture the value or quality of the AI's execution. A firm might generate many tokens for less critical tasks, while another generates fewer but highly impactful tokens for strategic work. The report could benefit from qualitative assessments or case studies that delve into the business outcomes achieved by frontier firms. Additionally, while the article mentions the need for governance and human review, more detail on the specific structures and processes that frontier firms employ for this would be valuable for organizations seeking to implement similar frameworks. The potential for AI to generate incorrect or biased outputs, especially when connected to company data and tools, remains a significant concern that requires continuous vigilance and sophisticated oversight mechanisms, which are only implicitly addressed.

Key Points

• Enterprise AI is rapidly transitioning from 'assistance' to 'execution,' with AI agents increasingly performing substantive, delegated work.
• A significant 'frontier gap' exists, with top AI-adopting firms generating 8.3x more output tokens per user than typical firms, a gap that is widening.
• Frontier firms disproportionately use advanced capabilities like Plugins and skills, which connect agents to company context and tools, enabling deeper workflows.
• Agentic AI adoption is spreading rapidly beyond engineering into knowledge work functions like legal, sales, and marketing.
• Early-career employees show higher AI usage than senior employees, presenting an opportunity for broader adoption through visibility and knowledge sharing.

---

📖 Source: From assistance to execution: How enterprises put AI to work