User Experience Is No Longer Just for Users
Designing software for humans and agents at the same time
For decades, software design optimized for a single audience:
Humans.
We built interfaces for human eyes.
Navigation for human cognition.
Workflows for human operators.
The field of Human-Computer Interaction (HCI) emerged from this assumption.
But something fundamental is changing.
As AI agents become increasingly capable, software now has a second audience:
Agents.
And unlike APIs or traditional integrations, agents interact with software in surprisingly similar ways to humans.
They search.
They navigate.
They retrieve information.
They perform actions.
They make decisions.
They collaborate.
This creates a new challenge for product builders.
We are no longer designing software solely for humans.
We are designing software for humans and agents simultaneously.
If HCI described the era of Human-Computer Interaction, perhaps this next era needs HACI:
Human-Agent-Computer Interaction.
The duality of modern software
Historically:
Human -> SoftwareToday:
Human -> Software <- AgentThe same system must now satisfy both.
A CRM must work for salespeople and sales agents.
A codebase must work for developers and coding agents.
A knowledge base must work for employees and retrieval agents.
A project management tool must work for managers and planning agents.
Every product increasingly has two users.
The human.
And the agent acting on behalf of the human.
User experience becomes agent experience
This doesn’t mean UX is disappearing.
In fact, UX becomes even more important.
Humans still make decisions.
Humans still review work.
Humans still own outcomes.
But now every UX decision has an AX consequence.
A folder structure isn’t just navigation anymore.
It’s retrieval context.
A document isn’t just something people read.
It’s context an agent consumes.
A table isn’t just a visualization.
It’s structured state an agent can reason about.
What humans see increasingly becomes what agents use.
Or, put another way:
What You See Is What Agents Get.
WYSIWAG.
This creates a fascinating convergence.
Good UX often produces better AX.
Good AX often produces better UX.
Clear organization helps both.
Clear naming helps both.
Good information architecture helps both.
Good provenance helps both.
Good search helps both.
The future is not UX versus AX.
The future is designing systems that are legible to both.
Agent harnesses are UX for agents
This is why so much innovation around agents feels strangely familiar.
Context engineering.
Memory.
Retrieval.
Tool calling.
Workflows.
MCP.
These are often described as infrastructure problems.
But another way to view them is:
They are user experience problems for agents.
Agent harnesses are becoming the navigation systems, search systems, menus, shortcuts, and affordances that agents use to accomplish work.
Just as UX designers spent decades making software easier for humans to navigate, we are now building systems that make software easier for agents to navigate.
The next frontier is collaboration
The most interesting challenge emerges when humans and agents begin working together.
Not interacting.
Collaborating.
The moment both operate on the same artifacts, entirely new requirements appear:
- Handoffs
- Shared context
- Provenance
- Review workflows
- Conflict resolution
- Permissions
- Audit trails
The challenge shifts from interaction to coordination.
And that may ultimately be the defining problem of the next decade of software.
Not how humans interact with computers.
Not how humans interact with agents.
But how humans and agents collaborate through computers.