Given the recent demise of Level Money, I was forced to search for a new personal spending tracking app. After some very brief Googling, I discovered Penny. Penny brands itself as not just a budgeting app, but a “personal finance coach.” This brand is embodied through a fascinating interactive UI that mimics an SMS conversation between you and “Penny” — the personable finance coach in your phone. Accordingly, the app foregoes a traditional menu and nests the main interactions within a hierarchy of commands modeled as a texting conversation.
Navigating Under Constraints
Text-based customer service bots are not novel, and in most cases are fairly painful to use: they are slow, not terribly interactive, and are generally just the triaging gatekeepers between you and the human you’d prefer to be talking to.
Penny, however, is built from the ground-up around this texting-like interface, and by imposing some simple constraints on user actions, manages to deliver an effective core solution with a pleasing and differentiated user experience. It’s an intriguing form of skeuomorphism, but it works.
By foregoing a typical menu and modeling actions as conversation, the creators are constraining the set of actions a user can take at any given time. Eligible actions are presented as texting short-cuts that are then entered into the conversation after you select them, much like Apple Watch preset replies or Allo’s Smart Replies.
Every time you open the app it will open in the “texting” interface and the Penny bot will greet you. Occasionally, the bot will interject an insight about your spending, and less frequently an offer, into the conversation before you can continue. The user can then interact in response to those situational prompts, if any, or continue to use core functionality such as requesting a chart of spend month-to-date or by category.
A few actions do utilize a free-text input, for instance, to ask, “How muchhave I spent at Starbucks?” However, this seems to be limited to some regex-type filtering and aggregation as more complex questions stump Penny, e.g., “What’s my average purchase amount at Starbucks?” which will again just report your total spend at the merchant.
A Little Personality Goes a Long Way
Impressively, the creators of the Penny bot have managed to imbue it with a bit of personality. The bot isn’t a dry automaton, but rather, feels quite like a casual human on the other side — cheering support when you achieve a spending goal, and sparingly sprinkling emojis and gifs into conversations.
Moreover, the UI imitates a typing animation that plays while the bot is ~crafting its response. The duration of this animation roughly corresponds to message length such that a single emoji is completed much more quickly than a full sentence. The average speed of response is user-configurable and at the fastest setting her comments are almost instantaneous.
Although I have enjoyed using Penny to date, I am concerned about the UI’s inflexibility for future feature elaboration. It’s natural that as the app matures they’ll wish to provide additional functionality, which, if maintaining its current UI conceit, will risk exploding the complexity of the hierarchical menus, the full functionality of which can be difficult to expose transparently, or access quickly.
I suspect the creators’ aspiration is to have an increasingly intelligent AI-agent guide the user as opposed to creating more and more elaborate elective menus, but in some sense, the brilliance of Penny is that it pretends it’s smart when it’s really quite dumb. A system of simple rules for interaction goes a long way in creating a very useful and usable solution where many many people fail at creating natural and effective AI-based interaction.