Trevor was born out of pure passion for virtual assistants and personal productivity. During my MBA (2015-2017), my peers and I collectively envisioned a future in which each human would have a set of AI assistants to aid them in specific parts of their work. A vision influenced by “Just A Rather Very Intelligent System”. A planning assistant will surely be one of them, as a key to solving productivity in the digital age.

I had identified the following productivity challenges in the digital age:

1. Flexible Working Schedule

(Planning Challenge)

Flexible working promotes autonomy, which requires personal planning, time-management skills, work discipline, self-management and a sustainable work-life balance.

2. Remote/Hybrid Working

(Organisational Challenge)

Remote working requires purely digital collaboration, and enables unprecedented cross-border and cross-cultural collaboration using SaaS tools for Project management, Real-time Communication, OKR, Agile, etc. to align an organisation’s efforts with it’s high-level goals.

3. Digital Overload

(Mental Challenge)

Digital distraction, information overload, dopamine addiction and frequent context switching between multiple types of activities quickly deplete willpower, reduce focus and impair intrinsic motivation.

‍

Trevor v1 - Solving the Organisational Challenge of Remote Working

Initially, my collage peers and I decided to solve the Organisational Challenge by building a project management app comprised of Tasks which you can schedule in your calendar and organize in sharable Projects. The goal was for Trevor to employ Pattern Recognition AI to analyse each individual’s performance based on their tasks and project due dates, and provide valuable insights for team leads to act upon. Insights would include individual progress reports, overall project status, workload balancing, missed due dates, etc in the form of personal notifications from Trevor - “Sir, John is falling behind schedule for Project XYZ. Consider reassigning some of his tasks to Jane.”

We designed and launched the following futuristic project management interface as an iOS app.

The nativity of our team of 3 college students wore off quickly. We were essentially competing in the crowded market of with Asana-like tools, while also aiming to build an ML model before TensorFlow was easily accessible.

The iOS App had a handful of subscriptions and we managed an exit by selling it for a profit and pivoting to solving a less-crowded market - Personal Planning.

‍

Trevor v2 - Solving the Planning Challenge of Flexible Working

The intersection of Time management and Task management

Soon after the exit, I joined the Founder Institute Startup Incubator with the new purpose of solving planning. The pivot led me to discover an interesting gap in the market, which yielded impressive productivity benefits. The concept is known as Time Blocking (or Time Boxing) and essentially means scheduling your tasks in your calendar to allocate specific time for them among your meetings and events.

‍

Practical Benefits of Time Blocking

• Hyperfocus on the current task - scheduling a task primes the mind for action within a defined timeframe, improving focus, reducing distractions and enabling Deep Work
• Full overview of everything competing for our time - a unified view of tasks, meetings and events for both work and life
• Realistic daily planning - scheduling our tasks allows us to realistically measure everything we can actually achieve per day
• Best of both words (task & time management) - an endless but organized database of tasks and an actionable schedule in one seamlessly integrated system

Emotional Pains and Gains

Emotional pains of traditional planning:

• Overwhelm, Distraction, Procrastination, Decision fatigute, Multitasking, Guilt of feeling unproductive, Illusion of not enough time, stress of remembering, feeling stuck, etc

Time blocking induces feelings of being:

• Confident
• Relaxed
• In-control
• Focused
• Mindful

Time blocking answers the fundamental questions in productivity

1. What is everything that needs to be done? (to-do list)
2. What should I actually do? (Prioritize by scheduling)
3. What can I realistically do today? (making daily schedule)
4. What do I focus on doing now? (focus on current task in schedule)
5. How far have I progressed and how can I improve? (analyse schedule)
6. Am I investing my time in alignment with my goals? (analyse schedule)

Drag & Drop - The simplest form of Time blocking

During the Incubator program, I designed an interactive UI wireframe that managed to attract funding in the form of development effort for an MVP from Obecto. That’s also where I met my incredible co-founder Dmitry Yudakov.

I would mostly be responsible for product design - UI/UX & Marketing, whereas Dmitry handled Engineering.

The product Integrated EventKit to manipulate calendar and task data from the user’s device with the purpose of building a seamless integration between tasks and calendar events, allowing users to schedule and unschedule tasks at will.

Swipe between recommended tasks from the bottom and drag them into your calendar to schedule them.

‍

Schedule tasks by long-pressing directly from the Smart Inbox, which combines native OS tasks, along with 3rd party integrations, such as Todoist.

‍

Ask your Trevor to create new tasks using NLP services API.AI (DialogFlow).

Side note: I really enjoyed building conversational UI’s.

‍

Despite the mobile app’s positive feedback, we received an overwhelming number of request for a cross-device app on a bigger screen.

‍

Trevor V3 - Seeing the Bigger Picture

Building a version of Trevor that would scale responsively, be accessible on any modern browser and integrate with 3rd apps without being restricted to a single ecosystem, excited both us and our users.

Dmitry suggested a tech stack that allowed us to do just that:

• A Node.JS backend using Firebase’s Realtime Database
• A React.JS Front-end using Redux and Material UI’s components
• A Google Calendar integration + Moment.JS + FullCalendar.JS to render the user’s schedule

As Dmitry was building the Tech Stack, I was bootstrapping the UI/UX Design, Marketing, BizDev and Front-end Engineering with React, Material UI & (S)CSS

#1 — Early Draft of Trevor’s Web App (Nov 2016)

Key Ideas:

1. Drag and Drop To-dos from the Task Inbox on the left to the Calendar on the right, with the ability to unschedule it after if needed.
2. Separate the day into sections, allowing users to drag and drop to-do’s into each part of the day. Each scheduled task had the potential to teach Trevor’s AI to categorise tasks by their time of day individually.

Result:

1. Drag & drop was universally loved by our users (At the time, Outlook was the only popular time blocking solution)
2. Users wanted more precise control to allocate both a timeframe and a duration to each task for realistic daily planning.

‍

#2 Live version of Trevor Web (July 2017)

Key Ideas:

1. Design each Task as a Card, making it more intuitive to drag & drop it into a new minimalist schedule.
2. Provide users with fine control, as start and end time can be readjusted as a traditional calendar event would. Scheduled tasks can easily be unschedule back into the inbox or rescheduled either manually or using Trevor’s smart suggestions.

‍

‍

#3 — Current Design for Trevor Web (since Dec 2021)

Key Ideas: Modernise the design and create a unique floating UI, where the main content floats on top of our signature background gradient, inducing a feeling of calm focus. The Inbox is situated within the calendar box, promoting the idea of integration instead of separation.

‍

At this point, we built a product with a solid time blocking interface that has been live for over 4 years. It features:

• A unified inbox, allowing users to add tasks to an internal database or a 2-way sync with Todoist or Google Tasks (in development)
• Ability to manage task lists, including the ability to share lists with colleges, assign tasks and write notes
• Ability to add scheduled tasks by clicking or dragging within the schedule
• Ability to unschedule, reschedule, complete or partially complete scheduled tasks
• Ability to focus on a single task by clicking on it, which adds a backdrop and provides a big note-taking space
• Ability to integrate multiple Google Calendar accounts and to show/hide synced calendars individually
• Ability to easily keep an eye on overdue tasks which were not yet marked as completed
• Smart Scheduling (and rescheduling) logic designed to find available time slots for a desired task

All the positive emails, slack messages and the over 80% completion rate for scheduled tasks proved our concept but one major thing was missing - the Artificial Intelligence.

‍

The “AI” in Trevor AI

At any point in Trevor’s development, there were two prevailing questions:

1. How do you interact with an AI planning assistant?
2. What’s the most efficient way to create a daily schedule for each user that’s personalised, centralised and realistic?

The most natural interaction for planning your day with an AI Assistant

Our Node.JS architecture allowed us to launch a Facebook Messenger bot utilising our Smart Scheduling technology, perfectly suited to be empowered by future AI efforts.

The workflow was as follows:

1. User asks Trevor in natural language to add a new task (ex: Workout)
2. Trevor saves task to inbox and provides a small number of duration suggestions, which the user can override with a custom duration
3. Trevor scans the user’s schedule and provides smart (or AI) scheduling suggestions, based on a section of the day (morning, work, evening, etc)
4. Once the user selects an option, the task is scheduled in the user’s Google Calendar.

Unfortunately, soon after launching, Facebook disabled Accounting Linking in Europe due to GDPR issues, after which we decided to focus our development effort on the Web app to further reduce our dependence on closed-sourced ecosystems.

True AI Automation in Task Categorisation and Scheduling Suggestions

While the Smart Suggestions algorithm worked great for finding available time, we didn’t have enough data on the duration of individual tasks. Once we had that missing piece, we could automatically find the right amount of time for each task as the first step in automation.

Dmitry successfully developed and launched a TensorFlow model for predicting a task’s duration, trained on calendar event data. To increase prediction accuracy, we use a global model and a user-specific model that overrides it in certain cases. The model is also trained on the day of week, time of day, etc. which will allow us to dramatically increase the accuracy of AI scheduling suggestions, which are currently in development.

This is our latest step in the Trevor AI journey to solving Personal Planning. Our current goal is to feed Trevor a task that needs to be done. It can be a manually added task, or a task added by a team lead via a 3rd party system. The user would then automatically receive a scheduling suggestion from Trevor, perhaps directly in the native calendar app on each user’s phone. Once the suggestion is accepted, the task is scheduled and awaiting to be completed in a calm and focused state of Deep Work. Once we validate the technology, we can scale it to every calendar user in the world with the ultimate purpose of empowering ambitious humans to do great work for humanity and the planet.