The Human Side of AI - part 2
The Question Isn't Whether You Use AI. It's Where the Human Stays.
By: Kara Maddox
“Uses AI” tells us almost nothing about how a business actually communicates.
We've spent years talking about AI adoption as though it's a switch.
Human.
Or:
AI.
But consider four employees.
One writes a customer email and asks AI to fix the grammar.
Another writes the substance and asks AI to reorganize it.
A third asks AI to create the first draft, then rewrites it.
A fourth lets AI generate and send the entire response.
Every one of them:
used AI.
But those workflows contain radically different amounts of human involvement.
And increasingly, research suggests that distinction matters.
AI involvement is a continuum
Recent leadership research tested different degrees of AI involvement while keeping the communication itself controlled.
When AI merely assisted with wording, observers did not show the same negative response found when AI generated the entire communication. High AI involvement reduced perceived empathy, credibility, and feedback quality (Sedefoglu-Ulucak, Ohly, & Schmelz, 2026).
Another preregistered experiment explicitly compared:
human communication
AI-assisted human communication
and
fully AI-mediated communication.
The researchers found meaningful differences in trustworthiness and authenticity across those production processes. In workplace email specifically, AI-assisted human communication was not significantly different from fully human communication in trustworthiness, while both outperformed fully AI-mediated communication (Sahebi, Formosa, & Bankins, 2026).
That's a much more useful finding than:
People hate AI.
They don't uniformly.
Instead:
How the human and AI divide the work matters.
Businesses already work this way
This isn't merely a laboratory distinction.
Interviews with e-commerce businesses found AI already embedded across marketing, customer service, content generation, data analysis, and operations.
But researchers found that many processes continued to depend heavily on human collaboration and supervision.
They described the dominant model as:
AI preprocessing + human fine-tuning.
That's important.
The emerging future may not be:
human versus machine.
It may be:
different configurations of humans and machines.
Human involvement can change the output itself
We see this outside business communication too.
Hitsuwari and colleagues compared human-created haiku with AI-generated haiku produced with and without meaningful human intervention.
The human-in-the-loop AI condition received higher beauty ratings than both unaided AI and human-only poems (Hitsuwari et al., 2023).
That's not evidence that customers will trust every human–AI communication.
But it demonstrates something important:
Human participation isn't merely ceremonial.
The configuration of human and AI work can change what ultimately gets produced.
Stop asking “Can AI do this?”
It probably can.
That's becoming less interesting by the month.
Instead ask:
What should AI do here?
Maybe:
retrieve
summarize
analyze
draft
compare
organize
Then ask:
What should the human still own?
Maybe:
context
verification
judgment
revision
responsibility
The correct answer will change from interaction to interaction.
The human doesn't need to do everything
This isn't an argument for keeping people busy.
Imagine a customer has:
17 previous interactions,
three purchases,
two refunds,
and six months of correspondence.
AI can summarize that history instantly.
Wonderful.
It can surface the relevant policy.
Great.
It can identify possible solutions.
Useful.
Then a person can decide:
What should we actually do?
AI didn't remove the human.
It removed work around the human judgment.
That's a much better use of automation.
Map the workflow, not the tool
For every customer-facing AI system, define:
AI's role: What exactly does the technology do?
Human's role: What meaningful work remains human?
Customer expectation: What does the customer reasonably expect from a person?
Consequence: What happens if the system gets it wrong?
Because “we use AI in customer service” tells us almost nothing.
The more revealing question is:
Where did the human stay?
Research note
This article is Part 2 of KJMdigital's The Human Side of AI series. Research supports the idea that degree and configuration of AI involvement can matter. The next question—whether customers respond specifically to downstream human review after AI generation—remains much less settled.
References
Hitsuwari, J., Ueda, Y., Yun, W., & Nomura, M. (2023). Does human–AI collaboration lead to more creative art? Aesthetic evaluation of human-made and AI-generated haiku poetry. Computers in Human Behavior, 139, 107502.DOI / article
Sahebi, S., Formosa, P., & Bankins, S. (2026). The AI penalty and disclosure paradox: Trust, authenticity and knowledge uptake in AI-mediated communication. Computers in Human Behavior: Artificial Humans, 8, 100304.DOI / article
Sedefoglu-Ulucak, D., Ohly, S., & Schmelz, J. (2026). Empathy in leadership communication: Experimental evidence from two vignette studies on AI's role in message improvement and observers' perceptions. Computers in Human Behavior, 185, 109096.DOI / article
Zhu, T., & Abd Rozan, M. Z. (2026). AI adoption in E-commerce enterprises: Insights into current practices and future directions from an interview study. PLOS ONE, 21(3), e0336416.DOI / article
Comments