The workflow-first framing is the part that maps cleanly into clinical settings. For an OR-adjacent agent, I would want the bounded task to name four things before it acts: what state it can see, what action changes, who owns the exception, and where the system stops when the record and room disagree.
Great article Ayten. Really sets out how companies should move forward with AI and we will definitely use it within our organisation. Thank you!
I know you do a lot of work on AI as well. Similar experiences?
Love this. Such deceptively simple steps for SMEs (and large corporates too, but that’s a story for another day).
Reading it, I found myself wondering whether the same four steps apply to individual leaders.
Where is AI already “in the room” influencing how we think, prioritize and decide?
What is the one bounded decision we’re prepared to let it own?
What would have to change in our personal workflows for AI to become part of the work rather than another screen to check?
And if AI is making more recommendations—or even decisions within guardrails—how do we think about ownership and accountability?
What I like most about this piece is that it shifts the conversation from what AI can do to what work we’re actually prepared to let it own.
A thoughtful and practical contribution to the agentic AI conversation.
I suspect it will be a while before agents take over decision in the boardroom.
The workflow-first framing is the part that maps cleanly into clinical settings. For an OR-adjacent agent, I would want the bounded task to name four things before it acts: what state it can see, what action changes, who owns the exception, and where the system stops when the record and room disagree.