Rethinking UX for Agentic Workflows: The Rise of Agentic Design Patterns
Over the past decade, UX designers have navigated unprecedented shifts, rapidly adapting to digital transformations across nearly every sector. Now, as we delve deeper into the complexities of automation, AI, and machine learning, we face a new frontier—Agentic Workflows. Unlike traditional automated systems that execute linear tasks, agentic systems involve autonomous agents capable of reasoning, adapting, and collaborating closely with humans. At Daito Design, we've witnessed firsthand the transformative impact of these systems in energy, marine, and industrial contexts, where reliability, efficiency, and safety are paramount.
Recently, Bijit Das's insightful Medium article on Agentic Design Patterns caught our attention, echoing many of our experiences and observations in practical deployments. This inspired us to further articulate our perspective, share actionable insights, and highlight critical considerations for UX professionals venturing into the world of autonomous agents.
Agentic Workflow example
Understanding Agentic Design
At its core, Agentic Design involves conceptualizing AI agents as active collaborators rather than passive tools. Traditional automation often focuses on linear, deterministic tasks, whereas agentic design accommodates dynamic decision-making, adaptability, and inter-agent communication. These digital collaborators can proactively identify problems, suggest solutions, and negotiate their roles based on changing circumstances.
This complexity introduces unique design challenges:
Visualizing the Agent's Mental Model: How can we effectively illustrate an AI's decision-making process and internal logic to build human trust and understanding?
Facilitating Trust and Control: When should users rely on an agent's judgment, and when must they intervene or override decisions?
Designing for Negotiation: How can agents effectively communicate and negotiate task distribution and priority among each other and with their human counterparts?
Agentic AI cluster
Core Patterns Shaping Agentic Design
Through our work and inspired by Das’s foundational article, we've identified several pivotal patterns shaping the future of agentic workflows:
Autonomy Gradient
Effective agentic systems aren't uniformly autonomous. Instead, they exist along a spectrum of autonomy, adjusting their level of independence depending on contextual needs. An agent might operate entirely independently in routine monitoring tasks but shift toward a supportive advisory role during complex operations.
Intent Signaling
Transparency about why an agent makes specific recommendations is crucial. Clear signals—like visual confidence indicators, explanatory text, or decision rationale visualizations—can significantly bolster human confidence and ensure appropriate user oversight.
Protocol-Centric Design
In agentic systems, clearly defined communication and negotiation protocols are as crucial as visual interfaces. Effective protocols guide agent interactions, decision-making workflows, and emergency contingencies, serving as the backbone for reliable agent collaboration.
Human-in-the-Loop Redesign
Traditional user interfaces like dashboards become less effective in agentic systems. Instead, UX designers need to develop mission-control style interfaces that empower human operators to oversee complex agent networks, stepping in only when anomalies or uncertainties arise.
Example AI Agent Cluster within the Maritime Industry
Case Study: Implementing Agentic Design in Offshore Maintenance
Consider a real-world scenario: maintenance management for offshore winches on marine vessels. Historically, this process relied heavily on periodic human inspections, manual logging, and reactive maintenance strategies. Crews would visually inspect equipment during shift changes, and any detected issues often required immediate, disruptive intervention.
At Daito, we approached this challenge through agentic design, breaking down the maintenance process into clearly defined collaborative tasks for AI agents and human operators.
Step 1: Continuous Monitoring
Agents continuously analyze sensor data, monitoring parameters like vibration, load stress, and operational usage. Human crews remain in passive oversight roles, receiving periodic summaries rather than continuously monitoring raw data.
Step 2: Anomaly Detection
When sensor data indicates abnormal trends—such as unusual vibrations suggesting bearing degradation—agents proactively flag these anomalies. Operators receive immediate alerts explaining the detected issue, enabling timely responses.
Step 3: Root Cause Analysis
Upon detecting an anomaly, the agent cross-references historical data, analyzing similar past events to determine the probable cause. Operators review the agent’s analysis, ensuring that the diagnosis aligns with operational experience.
Step 4: Task Planning
After diagnosing the issue, the agent proactively schedules inspection and maintenance activities, optimizing them around existing operational schedules, weather conditions, and crew availability. Operators retain full control, approving or modifying agent-generated plans as needed.
Step 5: Reporting and Learning
Once the physical inspection and maintenance tasks are complete, human feedback is incorporated into the agent’s models, refining future predictions and recommendations. This continuous learning loop steadily improves system accuracy and reliability.
Outcomes and Benefits:
Predictive Maintenance: Transitioning from reactive approaches significantly reduced unexpected downtime.
Operational Efficiency: Proactive scheduling minimizes disruptions and optimizes resource usage.
Enhanced Human Roles: Operators focus more on strategic oversight rather than repetitive tasks, improving job satisfaction and performance.
Essential Agentic Design Considerations
When designing these collaborative systems, several critical considerations emerged:
Clear Communication of Intent: Agents must transparently share the rationale behind their decisions, using intuitive and clear visual cues.
Adherence to Established Protocols: Ensuring that agents operate within established operational boundaries and comply with corporate and regulatory protocols.
Appropriate Autonomy Levels: Clearly defining limits on agent autonomy, particularly regarding high-stakes decisions, preserves essential human oversight.
AI Agent Orchestration with “Human in the Loop”
Unlocking Value Through Agent Orchestration
Building on these patterns, OneReach.ai's insights into AI Agent Orchestration add another critical layer to the conversation. In their article "Unlocking Enterprise Value with AI Agent Orchestration," OneReach emphasizes that AI agents, when effectively orchestrated, can automate complex tasks across systems, significantly improving both customer and employee experiences.
In industrial maintenance contexts, agent orchestration means connecting multiple agents across monitoring, diagnostics, scheduling, and reporting tasks. These agents collaborate in real time, accessing diverse data sources to make more informed decisions. By automating repetitive or complex tasks, agents free human operators to focus on strategic oversight and complex problem-solving.
To maximize value, organizations should:
Identify High-Impact Areas: Target workflows where agentic orchestration delivers significant operational gains.
Ensure Seamless Integration: Integrate agents into existing systems to avoid silos and maintain workflow continuity.
Invest in Change Management: Train staff to effectively collaborate with AI agents and adapt to these evolving workflows.
When orchestrated well, agents not only improve operational efficiency but also open new avenues for innovation, delivering scalable value across enterprises.
The Tangible ROI of Agentic UX
Industries implementing agentic workflows are already seeing measurable returns. For example, a recent maritime transportation study (MDPI) demonstrated significant gains in operational reliability and maintenance efficiency when employing AI-driven predictive strategies. Similar results have been echoed in sectors from manufacturing to energy, consistently revealing reductions in downtime, enhanced safety, and considerable cost savings.
The Future: From UX to Agent Experience (AX)
As these systems mature, UX design evolves toward a more sophisticated discipline—Agent Experience (AX). AX encompasses designing not just user journeys but also the journeys and behaviors of autonomous agents themselves:
How agents perceive and interpret the operational context.
How agents communicate uncertainties and rationale transparently.
How agents earn and maintain human trust through consistent and reliable behavior.
At Daito, we believe the next era of design will integrate human-centered principles deeply into agentic systems, creating cohesive human-agent teams that amplify both human potential and artificial intelligence.
Closing Thoughts
Agentic systems represent more than technological evolution—they signify a profound shift in how we interact with machines. As UX practitioners, our roles now involve actively shaping these interactions, ensuring they are intuitive, trustworthy, and beneficial. We invite our industry colleagues to explore this journey alongside us, exchanging insights, challenges, and solutions as we collectively define the future of agentic workflows.
The agentic future is already here—let's design it thoughtfully and responsibly.