From assistance to execution: How enterprises put AI to work

Source: Enterprise Technology & Strategy Briefing

The Challenge

Many organizations remain stuck using artificial intelligence purely as a conversational assistant—asking for brainstorming advice rather than delegating multi-step execution. While leaders purchase enterprise AI access, a widening execution gap separates organizations that merely experiment from frontier firms that achieve deep operational impact. Furthermore, high-value workflows often stay siloed in individual employee habits or engineering departments. Leaders struggle to connect AI agents securely to internal data, governance protocols, and non-technical teams, missing out on massive productivity gains across core operations like marketing, legal, and program delivery.

Core Findings

Drawing on enterprise data, the research reveals that top-decile 'frontier firms' generate 8.3× more output tokens per active user than typical organizations, driven by a pivot toward agentic execution. Codex accounted for 64% of combined enterprise output tokens, reflecting the shift toward executing complex, multi-step tasks rather than simple query-answering. Agentic workflows are exploding outside technical teams: weekly active users grew 108× in legal, 41× in sales, and 26× in marketing. Advanced tool usage—such as connecting agents to internal data and custom playbooks via plugins—is more than twice as common at frontier organizations. Additionally, administrative data demonstrates that early-career workers use AI far more intensively than senior executives, presenting a bottom-up adoption pattern.

Strategic Takeaway

To transition from passive assistance to active execution, leaders must shift their focus from tool procurement to workflow integration. Identify early-career power users and turn their individual habits into documented, shared team playbooks. Equip agents with direct access to company context, tools, and repeatable instructions while embedding strict permissioning and human-in-the-loop review guardrails. Expanding agentic automation across non-technical functions—such as operations, fundraising, and administrative workflows—delivers outsized efficiency gains without requiring deep coding expertise.

Deep Dive Q&A

What separates frontier AI organizations from typical enterprise users?

Frontier organizations generate over eight times more output per user because they move beyond answering ad-hoc questions to automating multi-step execution. They connect AI agents directly to organizational tools, data, and standardized playbooks.

Which non-technical departments are seeing the highest growth in AI agent adoption?

While software engineering pioneered agentic adoption, enterprise data shows rapid growth across knowledge work, led by legal (108× growth), sales and recruiting (41× growth), and marketing (26× growth).

Who is driving bottom-up AI usage inside organizations?

Administrative usage data indicates that early-career employees send significantly more messages and build deeper AI habits than senior leaders, making them vital internal champions for scaling practical workflows.

Related Research & Implementation Playbooks

Cross-disciplinary frameworks and practical blueprints connected to this insight.