Seneca: A Personalized Conversational “Planning” Assistant

Why Read This from an HCI Perspective

This article treats AI not as a mere automation tool, but as an interactive system that helps users organize their thinking. It’s an attempt to combine the ‘record-keeping’ of task apps, the ‘principles’ of paper planners, and the ‘flexibility’ of conversational AI—making it especially interesting for HCI/UX practitioners and researchers. In particular, by focusing not only on what users say, but on what they actually need, it prompts us to rethink product design and evaluation criteria.

CIT's Commentary

The core of this piece is not ‘smarter AI,’ but ‘a better interface for intervention.’ The design—where conversational AI asks questions to refine a user’s plan while also leaving behind structured lists and records—is a compelling combination in real products. However, this approach may trade away the fast ‘type it now and be done’ experience in exchange for convenience. That’s why it’s important to design the depth of questions, the frequency of intervention, and when the system should stop. From an evaluation standpoint, it’s also valuable that the focus is on process metrics such as the realism of the plan and alignment between goals and values, rather than just completion rates. These metrics must be rigorously validated even when building automated measurement tools with LLMs; otherwise, AI may make measurement easier while also making it more unstable.

Questions to Consider While Reading

• Q.How can we define the boundary between an intervention users feel is ‘helpful’ and one they feel is ‘interfering’?
• Q.In real products, how should we measure plan realism or goal–value alignment, rather than just goal attainment rates?
• Q.When conversational AI and structured task management coexist, how can we reduce the problem of users using only one of them?