The Question Most Organizations Aren't Asking Yet
By Martin Knudsen - Chief Executive Officer
06/24/2026
The question most organizations are still asking about AI is some version of "how do we get more of it into our work." It's a fair question. It just isn't the one that produces a different result.
A different question is sitting one level underneath it, mostly unasked. What does it mean to design work for AI participation, instead of designing AI to fit human-shaped work?
That question is harder to hold for a reason. It doesn't point at the technology. It points at the operating model.
The first question has a known shape. The second one doesn't.
Asking how to get more AI into existing operations is a comfortable question. It has familiar answers. Which tools. Which use cases. Which teams pilot first. The shape of the answer matches the shape of how organizations already make decisions.
Asking what work would look like if it weren't designed for humans alone to do it is a different kind of question. It doesn't have a tool answer. It has a structural one.
What would change. Where decisions sit. Who owns what. Where escalation routes go. Where work hands off, and where it doesn't have to.
The unasked sub-questions are mostly about decisions.
Step into that question for a moment, and a handful of smaller ones come with it.
When AI is making a recommendation that becomes an action, who owns the decision? When does the system get to act on its own? What happens when something goes wrong, and who notices first? How much should anyone rely on it, and when should a person step back in?
None of these are technical questions. They live in how teams operate, in how work gets coordinated, in how trust is allocated across people and new kinds of systems.
In a lot of organizations, that part hasn't really been designed yet.
The technology has moved faster than the structure around it.
The structure underneath was already undefined.
What's interesting is that adopting AI doesn't create this problem. It exposes it.
A lot of the way work was structured was never fully defined in the first place. Who really owns this kind of decision. When this kind of escalation should happen. What the team does when the answer is unclear. The slack of human judgment was carrying it.
AI doesn't fit into that slack quietly. It surfaces it. The places that worked because someone was quietly making sense of them are exactly the places that don't survive a workflow where AI is now a participant.
That's why the question feels uncomfortable. It isn't really about AI. It's about the parts of the operating model that have been undefined long enough that defining them feels like new work.
The answer isn't a tool. It's a design decision.
Asking the question seriously means a few things.
It means looking at where AI is being used today, and noticing that most of those places are the easy ones. Drafting, summarizing, generating, answering. The places where the human is still the decision-maker, and AI is helping faster. The harder places sit further back. Where AI would actually do the work. Where the loop closes without a human pulling it shut.
It means accepting that the answer isn't a tool selection. It's a design decision. About what the work is for, what the human actually contributes, and what the system is now responsible for.
It means the conversation moves out of IT and into the operating model.
The question is on the table.
Most organizations aren't asking it yet. They're still inside the first question, getting better at it, scaling it, optimizing it. There's nothing wrong with that. It just has a ceiling, and the ceiling is starting to be visible from inside the room.
The next move starts with a different question, sitting one level under the one most rooms are still asking.
The work itself is on the table. It hasn't been for a while.
Chief Executive Officer, Arkane Digital
Martin works with enterprise leaders to define AI operating models that align strategy with execution, helping organizations move from experimentation to scalable adoption.