What do people actually want from an AI coworker — and does it change depending on the job?
What design principles do workplace users rank highest for AI agents?
This explores what people who use AI agents at work say matters most in how those agents are designed, and whether those priorities hold steady or change depending on the task and setting.
This explores what workplace users say they most want from AI agents, and whether those wants stay fixed or shift with the work. The most direct answer comes from a multi-method study that drew up eight UX principles. Business users put four at the top: staying in control as a human, reliability, awareness of context, and safety What UX principles do workplace users want in AI agents?. Users didn't treat these as nice extras. They saw them as the price of entry: an agent that misses any one of them won't get adopted, however capable it is.
These rankings make more sense next to what agents can actually do today. In a simulated company, leading agents finish only about 30% of workplace tasks on their own. They fail most at social interaction, at navigating professional software interfaces, and at domain knowledge Why do AI agents fail at workplace social interaction?. When an agent fails that often, wanting control and reliability isn't nervousness. It's a sensible response to how the tool really performs. Context-awareness turns out harder than it sounds, too. Unlike normal software, an AI's context (the prompt, the conversation history, retrieved data) keeps changing and disappears between sessions, so users can't learn it the way they learn a fixed interface How does AI context differ from conventional software context?. Related work argues that what makes an agent feel like a colleague rather than a chatbot is persistence: memory, reusable procedures, and seeing tasks through to the end What makes an AI system feel like a colleague rather than a chatbot?.
The twist is that these rankings aren't universal. A survey of 202 workers found that on tasks they saw as pointless busywork, they didn't want explanations or creativity. They wanted the agent to be fast, simple, and decisive What agent traits do workers prefer for meaningless tasks?. So the same person may want close control over meaningful work and just want the agent to get on with the drudgery. Design priorities seem to follow how much the work matters to the person, not a fixed list. This fits with evidence that real delegation to AI clusters in information-heavy jobs, following what the technology can do rather than how much people chat with it Where have workers actually delegated tasks to AI?.
Older design guidance helps here. Microsoft's 18 human-AI interaction guidelines are grouped by moment: at first contact, during use, when the system is wrong, and over time How should AI systems be designed across different interaction phases?. In that light, the workplace priorities line up with different moments. Control and safety matter most when things go wrong, and context-awareness matters most over time. Proactivity is a further tension. Agents are passive mostly because of how they're trained, and making them take more initiative risks becoming intrusive, which clashes with users' wish to stay in control Why do AI agents fail to take initiative?.
One thing users may not see coming is that individual preferences don't cover organisation-level problems. Executives increasingly see agents as coworkers rather than tools Does agentic AI fit the tool-or-coworker framework?. Consulting models picture humans supervising "above the loop" while governance struggles to keep up Can AI agents truly operate above the loop at enterprise scale?. Others argue an agent can't simply be loyal to one user. It also has to balance other people's interests and be visible and accountable Should AI agents serve one user or balance many interests?. The control a single user asks for may need to grow into accountability at the level of the whole organisation.
Sources 11 notes
A multi-method study identified eight UX principles for workplace AI agents, with business users weighting human control, reliability, context-awareness, and safety as practical necessities for user acceptance and effective collaboration.
TheAgentCompany benchmark shows leading agents achieve 30% task completion in a simulated workplace. Social interaction, professional UI navigation, and domain-specific knowledge are the three primary failure modes, with multi-turn task performance consistently dropping to 35% across enterprise settings.
AI interactions operate on a substrate of constantly shifting context—prompt, history, retrieved data, hidden state—that users cannot internalize like traditional UIs. This structural mutability demands a new design discipline centered on context engineering rather than interface design.
Research shows the chatbot-to-colleague shift depends on state persistence, bounded memory, reusable procedures, and task closure—design properties of the system architecture. Larger models alone produce transcripts that disappear; colleagues accumulate experience and maintain workspace continuity across tasks.
A 202-worker survey found that on high-bullshitness tasks, workers consistently favor speed, simplicity, and directness over explainability or creativity. Task-perceived meaning appears to drive agent behavior preferences, not fixed workplace priorities.
Show all 11 sources
Workers have committed AI tasks to structured workflows primarily in information-intensive occupations, following technical capability more than conversational LLM adoption. This gradient differs sharply from routine-task automation predictions and wage patterns reverse at advanced degree levels.
Microsoft researchers synthesized over 150 design recommendations into 18 actionable guidelines grouped by when they apply: initially, during interaction, when wrong, and over time. Validation with 49 practitioners across 20 products confirmed their applicability.
Research shows next-turn reward optimization structurally removes initiative from models, but proactive behaviors like critical thinking and clarification-seeking are trainable (0.15% to 73.98% with RL). The core challenge is balancing proactivity with civility to avoid intrusion.
A survey of 2,102 executives across 116 countries found 76% view agentic AI as a coworker rather than a tool. This duality creates operational tensions because agentic systems simultaneously substitute and complement labor, requiring both HR and asset management approaches that traditional either-or frameworks cannot accommodate.
McKinsey describes AI-first workflows organized around outcome-aligned agent squads with humans providing strategic oversight, citing bank case studies showing efficiency gains. However, industry governance lags deployment ambition, and agent interactions create risks beyond individual component reliability.
Research argues the traditional single-agent-single-user alignment framing is insufficient at scale. Instead, governance should rest on inclusivity (balancing diverse interests), visibility (rigorous tracking and transparency), and liability (developer and deployer accountability for harms).
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- Will AI Agents Free Us From Meaningless Work? A Human-Centered Analysis
- A Framework of User Experience Principles for Human-AI Agent Interaction in the Workplace
- Guidelines for Human-AI Interaction
- Future of Work with AI Agents: Auditing Automation and Augmentation Potential across the U.S. Workforce
- The Emerging Agentic Enterprise: How Leaders Must Navigate a New Age of AI
- Explaining AI Agents Through Execution Traces
- HumanAgencyBench: Scalable Evaluation of Human Agency Support in AI Assistants
- Unraveling Human-AI Teaming: A Review and Outlook