Before we made this system, we had to find out the problems we are facing to solve. We spent a great deal of time exploring the existing CALO system, how busy professionals manage tasks and collaborate among each other, and how an intelligent agent could take off some of their burdens. We first took a good look at the existing system and buried our noses in theoretical papers. Then to understand users in the context of their work, we conducted a series of user interviews / observations.
The existing CALO was a monolithic window that contained almost every components of CALO. The interface was cluttered and inflexible, making it hard to find the information users were looking for. Below are some of the major problems the existing system had that we aimed to fix.
- Not based on user research
- The fundamental problem we perceived in the existing system was the lack of user research. It is crucial for a complex system such as CALO to be based on thorough understanding of the user needs.
- Made for AI developers / researchers
- One of the main problems the existing CALO had was, despite its sophisticated AI algorithm, the interface was made to be most useful for AI developers and researchers to figure out what was working and not working with the system. The system was slow and cumbersome to be used by regular users and the potentials of the AI was not brought apparent to the users.
- Fragmented components
- The existing CALO had many of the components that we have incorporated in our redesign, but they were highly fragmented. The to-do manager was detached and some components were rarely used. Information integration is essential to take advantage of a powerful AI such as the one for CALO.
- Interface is not consistent
- The user interactions of the existing system did not follow consistent standards or consistency within the system, making it hard for users to form a clear expectation of their actions.
From the literature, we learned that existing technologies for information management, such as email and instant messaging, are overloaded and take on multiple responsibilities and tasks. CALO solves this by directly pulling information from places like email and displaying them in their intended form such as tasks, scheduled meetings or reminders. We also discovered that current collaborative technology interrupts people's work rather being helpful, and there are AI techniques that mitigate this effect substantially.
Through careful research, we have developed an understanding of how CALO should support the work of busy workers.
Users We Interviewed
To understand the needs of overburdened knowledge workers, we looked at two user groups: assistants and executives. Executives fit the demographics of busy professionals who face the complexities of dealing with multiple projects and people at any given time. Assistants were important resources for us because of their relationships with and importance to the work of primary target user group, executives.
Highlights of Interview Findings
- Constant Interruption
- Interruption is an integral part of their work; instead of reducing these interruptions, we focused on creating a system that would support workflows resilient to interruption.
- Waiting for others
- Assistants must track short sequence of actions that require responses from other people, making the prioritization difficult. CALO Stardust serves as a repository for these short, pending tasks so not all information needs to be stored in their heads.
- Trust Over Time
- In general, assistants' relationship with executives is one of increasing trust and responsibility over time. Instead of being explicitly trained, assistants' functions and autonomy increase organically over time. CALO Stardust leverages this concept to its AI algorithm.
- Decentralized Information
- Information applicable to the executive is typically spread across many different repositories in many different forms. CALO seems well suited to collecting information from a diverse number of repositories and present them in a useful manner.
- Communication is Fundamental to their Work
- These days, executives' work is more service-based that require communication among parties who are often physically separated. Components of CALO that support collaboration should be more integrated to improve their communication flows.
We aimed to create a solution that solves the problems of existing system and the breakdowns users face while supporting their natural workflows and maximizing the benefits of AI.
Read more about the future directions of CALO Stardust's on our roadmap page.