Designing AI-UX Patterns at PagerDuty

At PagerDuty, integrating Agentic AI into our products challenged us to reimagine UX design principles across every platform. Here’s how we created robust, practical guidelines, grounded in user trust and multi-surface consistency.

Why we created these Design Guidelines?

Initially, AI development was centralized within a single team. With multiple teams working on AI Agents, we needed to ensure consistency and prevent reinventing solutions for identical UX problems. To streamline innovation and maintain trust, we needed a unified design playbook that could be clear enough for fast experimentation, flexible enough for Slack, Teams, or Web, and robust enough to inspire confidence.

It also came down to trust. When you hand more control to AI, the outcomes get unpredictable. As a designer, I wanted users (and fellow teams) to feel confident in how our AI Agents work, what to expect, and how to recover when things go off track. These guidelines are our shared safety net: they help us move quickly, experiment, and keep user experience clear and reliable for everyone as we launch new AI Agents.

Learning from industry leaders

I studied leaders like Salesforce, ServiceNow, and Microsoft, analyzing how they onboard users, promote transparency, and handle errors. This research helped us distill universal challenges and adopt proven design patterns for the fast-evolving world of AI-powered products.

Component Placement and Behaviour on Different Platforms

Consistency matters. But context matters too. For each platform, we mapped essential components: prompt fields, feedback UIs, proactive messages, to adjust intelligently. For example: a ‘nudge’ in Slack appears as a button, while on Web it is be a persistent bar and bigger number of nudges are allowed.

We began by looking inward

We began by auditing our AI usage across PagerDuty and chat integrations. This deep dive revealed both strengths and blind spots. More than cataloging features, we focused on whether our AI truly empowered users, and how we could build clarity and trust into every touchpoint.

Designing for multi-platform requires guidelines that specify not just what components appear, but how their behavior adapts intelligently to the environment and user workflow.

Core elements of Pagerduty Advance AI-Guidelines

Proactive Messages
Agents initiate when urgent action is needed, balancing helpfulness and relevance so users never feel overwhelmed.

User journey of proactive messages being sent during Incident Resolution

How proactive messages can help during incident resolution?

Key Improvements on UX

• Eliminated manual message scanning

• No need to think where to start investigating o look for additional information

• Reduce time for new responders to become effective contributors

• Reduced manual checking and coordination

Key guidelines we established for AI Agent proactivity

Target appropriately
Ensure the message is relevant to the personas.
Choose appropriate times to send messages.

Make the message Actionable
If the message requires action, clearly state what the user should do next. Use nudges where possible.

Give users enough information
For them to understand why the action is needed and how to complete it.

Nudges
AI Agents coach users toward best practices non-intrusively, supporting decision-making rather than demanding it.

Example of 2 main places in Web UI where nudges can be displayed to coach users

Key Improvements on UX

• Eliminated decision paralysis

•  Provided one-click execution of complex procedures

• Reduced human error under pressure

• Reduced manual checking and coordination

Source Attribution
Every Ai Agent answer or recommendation includes clear references, boosting transparency and user confidence.

Key Improvements on UX

• Source attribution turns AI from a suggestion engine to a trusted advisor

• Clear sources enable faster, more confident decision-making

• Documentation links speed up implementation and reduce errors

• Reduced manual checking and coordination

Response Time (latency)
Progress indicators set clear expectations, while rapid feedback keeps users informed.

Show the AI's thinking, not just actions. Swap vague updates ('Analyzing data…') for concrete ones ('Searching logs from the past hour'). It builds trust, keeps users informed during wait times, and lets them work on other tasks simultaneously.

Benefits for the User

• Agents can handle complex tasks autonomously.

• Users receive real-time information as tasks progress.

• Background operations allow users to focus on other tasks.

Evaluation Mechanisms
Users can rate AI Agent responses, and we iterate quickly based on their feedback.

Collecting feedback about the Agents' responses and recommendations is crucial to identify areas for improvement and refine the AI Agents.
Although on a milestone version Agents won’t be learning directly from feedback being provided, we could within design and product use this feedback to keep improving Agents performance, for all the 3 surfaces: Slack, Microsoft Teams and Web UI, that you can see below.

UX Design Direction: Building on Solid Foundations

Our guidelines are rooted in clarity, collaboration, and adaptability. We've done a 360 Enablement Workshop, which considered Product, Design, and Engineering perspectives. We aligned on our definition of Agentic AI across Web, Slack, and Teams. This kickoff (April 16, 2025) wasn’t a singular event, but the beginning of a continuous journey , cataloging patterns, refining principles, and empowering every team to design confidently.

Final words of appreciation

Good work is always collaborative. I’d like to thank AI-First team for championing our shared vision, as well as to design and product leadership feedback which made our guidelines sharper and more valuable.

Looking Ahead

AI design at PagerDuty is evolving with every launch, every lesson, and every regulatory change. Our guidelines remain living documents , adapted and improved as we learn. If you’re facing similar challenges building AI-powered solutions, let’s connect and shape the future together.