Designing Meaningful Agentic AI Experiences
The rise of Agentic AI has revealed new UX challenges alongside new opportunities. Careful integration within existing workflows determines whether it will enhance your users' experience or hinder it.
As a Strategic Design Studio, here are our thoughts on it.
What is Agentic AI?
AI Agent
A system designed to complete specific tasks, often reacting to instructions or input.
Agentic AI
A system that takes initiative, sets goals, and acts autonomously across tasks and workflows.
The 4 UX Challenges of the AI-pocalypse
Challenge 1 Trust and Transparency
The biggest challenge lies in enabling users to feel confident and safe with the technology. Trust takes time, consistency, reliability and transparent communication. It can be unclear how Agentic AI is processing, interpreting, and acting upon data, leading to unpredictable moves that can result in harmful outcomes while handling critical information.
Users need visibility into the actions and decisions AI makes behind the scenes to trust it.
Our Principle
Transparency doesn't mean showing everything: it means showing what matters.
The Opportunities
How can the design of Agentic AI integrated workflows provide transparency, consistency and reliability to users?
- Embedded conversational, spatial and temporal paradigms to visualize, track, explain and take action on AI output.
- Communicating system status, confidence levels and uncertainties before, during and after decisions and interpretations are made by AI.
- Branches and safe sandboxed environments for AI to explore and solve its queries safely, as well as revert states and data preservation when agents cannot complete tasks or make mistakes.
- Transparency and traceability over how user data is gathered, handled and processed, and over sources used to inform decisions and processes.
Challenge 2 Control versus Automation
The second challenge lies in establishing balance between AI autonomy and human oversight in critical workflows and tasks. The unpredictable decisions and irreversible actions that Agentic AI systems sometimes make can lead users to feel powerless and damage their hard-earned trust.
Users need to retain creative agency and decisional power over how their workflow and tasks get carried out.
Our Principle
AI shouldn't replace human decision-making, it should empower it.
The Opportunities
How can the design of Agentic AI support users efficiently while preserving their sense of agency and decisional power?
- Built-in progressive teaching and delegation phases: from very specific and contained tasks, to overseeing larger workflows. This includes feedback loops to help AI understand its performance and manage risk.
- Intent-based alignment and contextual adaptation to the teams' ways of working: user-defined areas of interests, goals and constraints, so that AI embeds itself in their workflows, not the other way around.
- Clear cues, control touchpoints, escalation triggers and fail-safe mechanisms that users can easily identify and take action on.
Challenge 3 Cognitive Overload
The third challenge lies in enabling users to manage multiple agentic AIs and increasingly complex workflows under tighter deadlines. Multiple agents operating interdependently and taking decisions across different touchpoints concurrently can be hard to track and assess in a timely manner.
User need to retain situational awareness and understanding of the bigger picture to make informed decisions.
Our Principle
Effective AI experience doesn't reduce the amount of tasks, it reduces the mental burden they create.
The Opportunities
How can the design of Agentic AI preserve users' situational awareness in increasingly complex workflows without overwhelming them?
- Progressive disclosure of information and personalization of the granularity of details provided, allowing drill-downs to deeper explanations when sought.
- Intelligent, contextual and personal cues that proactively surface the right information or trigger at the right moment through internal prioritization.
- Layered interface paradigms for navigation, aggregated views for overall system health and information processing, with appropriate feedback loops and cues.
Challenge 4 Human and AI Collaboration
The fourth challenge lies in designing interfaces accessible to both human users and AI systems together, when human need for visual and intuitive navigation patterns can conflict with AI requirements for structured data and slow-down machine processing.
Users need to be able to benefit from collaborating with AI without either specific workflows suffering from it.
Our Principle
Future interfaces won’t only cater to human needs. They will also cater to AI.
The Opportunities
How can UX Design enable both Agentic AI and Humans to collaborate efficiently and be more than the sum of their parts?
- Progressive autonomy that allow collaboration to mature naturally, with AI taking on more responsibility as human confidence grows.
- Providing users with context-specific insights and bidirectional feedback loops to enable them to get the most out of AI, and for AI to continuously learn and adapt.
- Hybrid interfaces and processes that align more closely with the users' end goal to collaborate (brainstorm ideas vs create visuals, vs write outlines), beyond the familiar conversational paradigm.
AI is transforming entire industry workflows. With or without good user experience.
As designers, we have the ability to shape how technology impacts people.
We're here to ask the right questions today to determine better outcomes for tomorrow.
Did this analysis resonate with you? Are you looking to integrate Agentic AI into complex workflows?
Get in touch