Human-AI Experience Guidelines

Designing a standardised and scalable approach to Human-AI interaction.

The Strategic Challenge

Workday's strategic pivot to an AI-first product rollout presented a high-risk situation for the company and its user experience. The core problem was not just implementing best practices across hundreds of products, but the absence of a clear conceptual framework to guide designers in this new technological landscape. This lack of guidance was causing delays and leading to inconsistent approaches across product experiences. I identified this critical gap and highlighted it to our VP of Design.

This shift marked a fundamental change in how users interact with technology, moving away from instructing the computer on what to do. The new paradigm required users to specify the outcome they desire, a concept known as "intent-based outcome specification". This raised critical questions we needed to answer about how to convey to users how AI systems behave, how to explain what an AI system can do, and how to create trust by learning from changing user expectations.

The Mission

To address this challenge, I championed the creation of a comprehensive resource hub. My objective was to provide a standardized and scalable method for Human-AI interaction that would enable 1000’s of Product and Technology Workmates across various roles to design impactful experiences. The project mission was to create hands on tools that helped teams quickly grasp the complexities of crafting appropriate Human-AI experiences, preventing the risk of inconsistencies and delays.

The Solution: Human Centered Heuristics

I defined a simple, human-centered set of guidelines that everyone could follow, from experienced AI practitioners to beginners. The guidelines are divided into three phases, corresponding to a user’s journey and is supported by four sets of guidelines relevant to that phase. Every guideline includes multiple real-world examples, making it easier for first-time users and expert practioners alike to comprehend the subtle aspects of Human-AI interaction techniques.

Before Interaction: Set the right expectations.
During Interaction: Ensure users feel in control.
Over Time: Build trust with users over time.

My Role & Leadership

As the Principal UX Designer on this project, I was responsible for:

Strategic Research & Problem Definition: I proactively identified a critical gap in the company’s UX framework for AI and defined the business risk of inaction.

Led the Definition and Creation: I championed the creation of a simple, standardised set of guidelines that put design in the driving seat during early stage ideation and when auditing UX with a clear set of heuristics.

Driving Adoption: I partnered with Product Marketing and Responsible AI teams to promote the guidelines and establish them as a company-wide standard, helping to define standardised development protocols that are now part of our Responsible AI framework, I also facilitated internal learning and external customer events, helping develop trust in Workday’s approach to AI

Secondary Research & Synthesis
Writing Guidelines & Designing Examples
Publishing to ‘Canvas’ - Workday’s Design system
Internal and External Promotion - learning, and customer events.

The why: navigating a paradigm shift in UX

As we navigate opportunities and hurdles in developing user-facing AI, prioritising human-centered User Experience (UX) Design becomes increasingly important. Placing users at the core of product development ensures we’re building intuitive and meaningful experiences, supporting users that are navigating powerful new technology for the first time. This approach requires the application of evidence-based best practices from the outset that build trust, and empower users to leverage Generative AI’s full potential.

Users now require support for entirely novel workflows that help them work in parallel with the newfound superpowers AI provides. Jakob Nielsen, describes this as “intent-based outcome specification.” The user no longer tells the computer what to do. Rather, the user tells the computer what outcome they want.

We’ve got some big questions to answer as this technology is rapidly adopted for commercial use.

How can we convey to users how AI systems behave and the potential for diverse outcomes from identical inputs?
How can we explain how to use an AI system effectively when users have little experience of what it can do, and how well it can do it?
How do we learn from users about their (changing) expectations to create a personalised experience, built from their behaviour and preferences so we can tailor the AI system’s interactions with them?

1. Research

Bringing the guidelines to life required extensive secondary research of scholarly publications, emerging and existing research paper in the field of ML and AI - a project I undertook over 2 months initially to advance my own understanding. Before long i realised how helpful this information was going to be for my workmates - you can see the sources I used here

2. Drafting & Pattern matching

With a rough set of guidelines outlined i then went hunting for patterns and examples to match with each guideline.

3. Review, iterate and improve

Now with examples to share I could road-test the guidelines with product teams in design reviews. By reviewing a broad spectrum of interaction scenarios against the guidelines I could spot gaps in my approach.

4. Refine and Share

Working with the very talented Amreen Ukani we started to bring better meaning and readability to the guidelines, pulling and pushing them to make them better. I also had the pleasure of working along side an amazing visual designer Pilar Valencia who helped define some visual styling for use in external and internal presentations. Once this was all done I took the guidelines on a roadshow around Workday teams, sharing them at various Townhalls, and design meetings to help teams operationalise them as part of thier design process. They were also featured at Workday ‘Rising’ - Workday's annual conference with more than 30,000 in-person and online attendees. Shortly afterward They were also featured on the Workday engineering blog with a request to share them to the world on our Canvas Design System. I also created an accredited learning course for all Workday Employees that is now mandatory for all UX Designers at Workday Worldwide.