On my radar: the AI conversation is moving from “Which tool should we use?” to “What work should this system actually own?”
That is a much bigger shift.
For the last several years, most workplace AI use has been reactive. Ask a question. Draft a document. Summarize a meeting. Generate ideas. Useful, but still centered on a human initiating each interaction.
Agentic AI changes the model because systems can take on multi-step work, use tools, maintain context, and move a task forward over longer periods of time.
The change is already visible
Microsoft’s 2026 Work Trend Index, based on a survey of 20,000 AI users across 10 countries plus productivity data, found that advanced users are more likely to redesign workflows, establish shared AI standards, and document agent handoffs and quality expectations.
Only 26% of AI users surveyed said their leadership was clearly and consistently aligned on AI.
Source: Microsoft 2026 Work Trend Index.
This is not just a better chatbot
OpenAI has described the shift as moving from short, self-contained interactions toward delegated, long-horizon tasks. Its enterprise research similarly shows a widening gap between firms that use AI mainly for assistance and firms that connect agents to tools, context, and repeatable workflows.
Sources: OpenAI, How agents are transforming work; OpenAI Enterprise Signals.
Why most organizations are not ready
The obstacle is not necessarily model access. Many companies already have access to powerful AI.
The harder questions are organizational:
- Which work can be delegated?
- Who owns the outcome?
- Where does a human need to approve?
- What data can the agent access?
- How is quality evaluated?
- What happens when the system is uncertain?
- Who redesigns the workflow when the old process no longer makes sense?
The org chart may become an AI strategy document
This is why I keep returning to the language of an AI staff.
It forces a better conversation. Instead of collecting tools, you define roles. Instead of random automations, you define ownership. Instead of assuming “AI will help,” you decide exactly where it participates in the work.
What I am watching next
I am watching three things closely:
- Governance moving closer to workflow design. Companies will need practical rules at the point of work, not only high-level AI policies.
- Agent-to-agent coordination. Multi-agent systems will make role clarity even more important.
- Software economics changing. Gartner has estimated that agentic AI could put as much as $234 billion in enterprise application spending at risk by 2030 as agents increasingly operate across systems rather than requiring humans to use every interface directly.
Source: Gartner, July 1, 2026.
The companies that win will not simply have more AI. They will know where AI belongs in the work.
That is why this is On My Radar.