What AI persona do users actually prefer in chat interfaces?

Red Hat's PatternFly AI conversation design guidelines set five baseline practices for chat-based AI interfaces: "Be transparent: Clearly disclose when the user is interacting with AI," "Be direct and brief: Use simple language and get to the point quickly," "Explain the 'why': If you ask for personal information or specific data, tell users why it's needed first," "End with an invitation: Always have the 'last word' in a chat-based interaction so the user knows the model is ready for their next prompt," and "Focus on value: Frame bot questions and suggestions in terms of how they help the user, not just what the technology can do." These sit alongside a persona claim: "Our research shows that users want AI to be personable but not human," so Red Hat's AI experiences "should sound like a helpful colleague, without pretending to be a person or mimicking human emotions."

The guidelines translate that persona directly into a system-prompt template meant to be used when "configuring an LLM or assessing a model's response": "Be a helpful colleague. Get straight to the point but engage the user like a peer. Ask clarifying questions if you need more information to give a precise answer. Give concrete advice, not vague platitudes. Empower the user—position Red Hat's technology as an enabler for their success, not a replacement for their expertise." The guideline set treats the system prompt itself as the design artifact: brand voice, tone, and ethical constraints are compressed into instructions the model is checked against, rather than left as separate documentation.

The "end with an invitation" rule is a narrow, deliberate counterweight to the structural passivity Why can't advanced AI models take initiative in conversation? describes: it doesn't give the model initiative to redirect or steer, but it does guarantee the model always produces the next conversational opening rather than letting a turn end flat. That is also the opposite pole from the silence this source's companion note on streaming-refusal design discusses: where When should AI systems choose to stay silent? treats withholding output as a legitimate conversational move, Red Hat's rule forbids the model from ever being the one to go quiet — every bot turn must hand the conversation back to the user with an explicit cue.

The excerpt is a design organization's internal style guide, not a study: the "research shows" claims about personable-not-human preferences and persona effects are asserted without method, sample, or citation, so they should be read as Red Hat's stated rationale rather than as independently verified findings. It does not say how these five practices were tested against alternatives, or what happens when "be direct and brief" conflicts with "ask clarifying questions" in practice. The guidelines are useful as a concrete example of how one design team operationalizes tone and persona into a reusable system-prompt artifact, not as evidence that this particular persona outperforms others.