Source: Hanelt et al. (2026) — “Opening the Network of Trust: How Domain Experts in Triadic Relationships Build Trust in AI-Based Counterparts,” MIS Quarterly
We often talk about trusting AI as if it happens directly.
A person meets a system. The system performs. Trust either appears or collapses dramatically, like a bad first date.
This paper suggests the trust story is more networked than that.
Domain experts can help other people build trust in AI-based counterparts. In other words, trust does not only come from the AI’s behaviour. It also comes through relationships with humans who understand the domain and can interpret what the AI is doing.
That makes sense.
Most users are not in a position to evaluate AI deeply. They can see outputs, but they may not know whether those outputs are meaningful, risky, or just confidently formatted nonsense. A domain expert becomes a translator. They help explain when the AI is useful, where its limits are, and what kind of judgment still belongs to humans.
This is especially important in professional settings.
If a doctor, analyst, engineer, or manager says, “This system is useful here, but be careful there,” that carries more weight than a vendor brochure saying “trusted AI solution” in a font that looks expensive.
Trust becomes triadic: user, AI, expert.
That is a much more realistic model of adoption.
It also means organisations should stop assuming trust can be created by training videos alone. People trust through people. Especially when the technology is complex.
My takeaway: AI trust is social infrastructure. Domain experts are not just users of AI. They are trust-builders, interpreters, and boundary setters. If companies ignore that role, they may deploy technically strong systems into environments where nobody knows how to believe them properly.