It’s fascinating to hear which tasks people want to delegate to AI.
Some time ago, during a discussion about GenAI ideas, I asked one question that completely changed the room:
“What happens if the other person also comes with an AI partner?”
Silence.
That silence matters – because it reveals an unspoken assumption behind many GenAI use cases: we design automation as if there will always be a human on the other side.
GenAI doesn’t only change workflows. It changes relationships – because many automation ideas assume a human counterpart… until that counterpart shows up with their own AI.
And when both sides delegate, we can easily end up with AI-to-AI conversations, while humans are left outside the loop.
That changes:
- how trust is built,
- how conflicts are resolved,
- and who is accountable when things go wrong.
Imagine:
- A supplier negotiation where both sides use AI to draft and optimize responses.
- A performance review prepared by one AI and answered by another.
- A customer complaint handled by automated agents on both ends.
And consider the implications.
Research shows two opposite risks in such environments:
- People may over-rely on automated recommendations.
- Or they may reject the system entirely after one visible mistake.
So when we build LLM systems, the question isn’t only: “What can we automate?”
It’s also: “What kind of human collaboration do we want to protect?”
Checklist: building GenAI without losing the relationship layer
- Clear business goal – Are we optimizing for cost, speed, quality, risk reduction, or experience?
- The other side – What changes if they also delegate to AI? What happens in a fully AI-to-AI interaction?
- Human-critical moments – Where is the relationship itself the value (negotiation, feedback, conflict, care)?
- Accountability model – Who owns the final decision? How fast can a human step in?
- Trust design – Are limits, uncertainty, and responsibility transparent?
- Collaboration principle – Does the system support human judgment – or silently replace it? If both sides automate communication, who remains in control – and who carries the risk?
When AI shapes the dialogue on both ends, responsibility does not vanish – it shifts. The question is: are we designing for that shift?
Leave A Comment