What Makes a Good Task? Novelty, relevance and timeliness in Toodo

Right now Mark and I are spending the vast majority of our time at Toodo on ‘task quality’ - making sure the recommendations you get on what to do are always useful, and that the tasks Toodo suggests are high priority.

We all have an intuitive grip on the “things we have to get done”, but codifying what makes one task ‘better’ than another is extremely slippery. Consider which of the following is a ‘better’ task:

• On the surface the first task seems ‘better’ - more concrete, with a clear action and a deadline. The second task with an action to “review” seems vague, and maybe just bureaucratic busywork.

• Context is important, though. If ‘Mark’ is your CEO, and Project Orion is a war room about an emerging PR incident, it quickly becomes extremely urgent.

How to design tenets that would apply correctly to literally anything that could come into your in-tray?

When sorting through the attributes of a ‘good’ task, I had a flashback to the first time I ever presented to the C-suite while I was at LinkedIn. During the call I got (rightfully) schooled by the CEO about the quality of the customer insights I was providing:  “Good customer insights”, I was told, “are always 1) Novel, 2) Strategic, and 3) Timely”. To this day it is probably one of the best (most concise and true) pieces of professional feedback I’ve received.

The feedback settled somewhere deep in my core long-term memory and has never gone away. And it turns out that bringing it to bear on what makes a ‘good task’ is highly fruitful. 

When deciding to create or update a task for one of our users, the following three attributes tend to give us a very good handle on whether it will be a ‘good’ task for them to do right now:

• Novelty: Is this task based on some new information for the user (ie our website has just gone down)?

• Relevance: Is the task for them specifically, and is it directly related to their goals?

• Timeliness: Is this task urgent / is it important that they get started on it soon?

When you’re building an AI that helps people prioritize their work, it isn’t enough to just add into the system prompt, “tasks for the user should be novel, relevant and timely”. At the start, a model has very limited context on you, your work, your goals, your company, your colleagues etc. so it isn’t going to make good decisions on its own.

Toodo uses hundreds of signals to get a sense of a task’s novelty, relevance and timeliness - who a task is coming from, how it relates to your goals, whether you usually snooze similar tasks, how it relates to the work you’ve previously completed, and so on.

But for now, these three criteria—novelty, relevance, timeliness—form the foundation for how we think about task quality. They drive our product design and inform the AI that helps users cut through the noise and focus on what matters.