Introduction
With the explosion of generative AI onto both consumer and enterprise productivity tools, UiPath was quick to realize its unique position in the market and we knew we had to act fast to release a proof-of-concept we could demo to our shareholders and to our customers offering our vision for integrating generative AI into our existing automation suite.
The UiPath Autopilot was the original name given to this product and enhanced the existing 'Assistant' product with a contextual, chat-based experience.
This product aimed to develop a chat-based assistant for non-technical employees of large companies to improve their productivity and encourage use of automation. This AI-powered tool streamlines workflows by offering contextually relevant automations such as translations, ticket booking, and article summarization, ultimately simplifying business operations and allowing users to focus on core tasks without technical expertise.
The Problem
With the increasing demand for productivity, business employees often face significant challenges in navigating and utilizing the technological solutions given to them for their day-to-day operations. The need for a seamless, intuitive solution that minimizes technical barriers and maximizes efficiency was evident. The core problem was the lack of accessible, user-friendly tools that could offer instant, contextually relevant solutions — such as automating routine tasks or providing AI-powered assistance for complex queries. This gap hindered productivity and emphasized the need for an innovative solution tailored to the workflows of non-technical business users.
My Role and Contributions
At the outset of the project I was the sole designer given the responsibility of visualizing early concepts and bringing them to life. My role began with the creation of wireframes, which served as the project's architectural blueprint. These wireframes were crucial in visualizing the user flow and suggested an interface design, laying the foundation for what would eventually become a robust, flagship product for us. The creation of higher fidelity prototypes followed, enabling us to demonstrate the practical applications of the assistant and secure buy-in from key stakeholders, including presenting to the CEO and CTO of the company. This early work was pivotal in setting the direction for the product and establishing a concrete vision that guided subsequent development efforts.
Transitioning from the conceptual phase to a more defined development stage, I collaborated closely with another designer and together we embarked on a rigorous competitive analysis, participated in a week-long design sprint dedicated to refining our concepts, and meticulously defined user interactions. Our collaborative efforts saw the creation of a comprehensive set design assets that guided the engineering team in their development work. Throughout this process, our adherence to an agile, week-to-week approach facilitated a dynamic and responsive environment, allowing for rapid iteration and improvement of the design based on feedback and evolving requirements.
Examples of final design assets that were prepared for our front-end engineers to build in code.
Methodologies and Processes
Initially, we conducted a thorough competitive analysis to understand the landscape and identify unique opportunities for differentiation. This research provided valuable insights into the strengths and weaknesses of existing solutions, enabling us to avoid common pitfalls and aim for innovation in user experience and functionality.
Subsequent to the competitive analysis, we engaged in a week-long design sprint with several PMs, engineers, and product owners involving rapid ideation, prototyping, and testing. One outcome of this sprint was a getting a lot more buy-in as everyone felt like their voices were heard and ideas at least considered before we moved forward. By working closely with the engineering team and leveraging our iterative design process, we were able to refine the product's user experience and interactions, ensuring alignment with user needs and project goals.
During the design sprint we set up a virtual affinity diagram and then sorted ideas into meaningful categories.
Challenges and Solutions
The project presented a set of distinct challenges that required innovative solutions to ensure both functionality and user satisfaction. Among the most pressing issues was the necessity to differentiate between quick actions, pre-built automations, and just-in-time automations. Just-in-time automations were experimental and thus less reliable, so was important to establish a clear, user-friendly system of categorization without overwhelming the user with excess information.
Three types of actions were possible and the action 'cards' for each type needed to convey their differences while still maintaining visual cohesion.
Effectively communicating the system's status and the confidence level of responses also presented a considerable challenge. Users need to be aware of the processing status of their requests and the reliability of the actions taken by the agent. I enhanced the user interface with visual indicators and thoughtful content design to effectively communicate the system's status and response reliability, aiming to build trust in the system.
These wireframes were meant to communicate at a glance the flows representing high, medium, and low-confidence responses.
Outcome and Reception
While the project had not been fully deployed by the time of my departure, the feedback from its demonstration at the company's annual conference was overwhelmingly positive. This presentation served as a pivotal moment, showcasing the potential impact and value of the chat-based assistant to both leadership and attendees. The enthusiastic reception highlighted the project's innovative approach and its capacity to meet the identified needs of non-technical business users.
Although specific metrics to measure the success of the project were not available, the positive response and interest generated among the demo's audience provided a strong indicator of its potential effectiveness and utility. This outcome underscored the importance of closely aligning product development with user needs and expectations, as well as the value of iterative design and testing in creating solutions that resonate with end-users.
Insights and Reflections
The project brought to light a fundamental insight about user interaction with automated technologies: users are generally indifferent to the type of technology enabling their actions, be it AI, machine learning, or any other form of automation. This revelation was crucial in understanding that while the technological underpinnings of each action were not a primary concern for users, setting clear expectations regarding the performance and reliability of these actions was paramount.
Using "try" as the CTA was our way of clearly communicating the experimental nature of AI-powered automations.
We recognized the need to effectively communicate that some actions, especially the experimental just-in-time automations, could take longer and might be less reliable compared to quick actions or pre-built automations. This understanding led us to implement user interface design elements that subtly conveyed the confidence level of an action's outcome, and potential wait times, thus managing user expectations and enhancing the overall user experience. Reflecting on this project, it became evident that the clarity of communication and managing expectations were key factors in fostering user trust and satisfaction, regardless of the complexity of the technology involved.
Giving users the option to show or hide the intermediate steps during an AI-based automation allowed us to use the principal of progressive disclosure to foster user trust and understanding.