We set out to create an AI assistant for work, helping workers being more productive in their job.
Have you ever had the feeling that you could better plan, prioritise and structure your workday? Have you ever thought you were spending too much time handling the back and forth of meeting scheduling and dealing with administrative tasks? If you do, we have a lot in common.
The problem is that too much time is wasted in today’s workplace. Low-value tasks consume individual’s valuable time, and getting overwhelmed can be quite easy for modern office workers. This is especially true for managers, when 30% to 60% of their time is spent in coordinating meetings (BCG Survey).
The sum of repetitive ad-hoc administrative tasks has become a pain, therefore indicating that time management isn’t just a personal challenge but an expensive organisational issue.
With the rise of messaging platforms like Slack, Kik, Google Hangouts or Microsoft Teams and the improvements in Natural Language Processing (NLP) technologies, we saw an opportunity to tackle the challenges of today’s workplace and help companies save a lot of money.
That’s why we set out to create Ophi, a chatbot that saves the time of office workers and managers by automating and organising their workflows in one single place. In short, offering you the possibility to resolve problems and plan your day through a quick chat.
With Ophi, we planned on entering the era of conversational offices, the idea being that everyone need and deserves an automated assistant for her or his workday.
The project started with doing user research on today’s current pain points within the workplace. We managed to build a better understanding of what would make sense to automate first and identified the first features to build, revolving around time-management and meeting scheduling.
This is how we started to prototype Ophi to work in Slack and focused first on synchronising team meetings and individual calendars in a smart way. Our tech team developed the chatbot using Natural Language Processing technologies (NLP) and microservices architecture. In total we developed eight different services like Slack integration, Google Calendar synchronization, analytics tool, NLP and more. All services were connected to our central application using NATS (high performance messaging system).
At the same time, the UX team explored the best practices of Conversational User Interfaces (CUI) in order to ease and drive adoption within the market.
Through a cycle of user testing and iterations, Ophi got faster and smarter to help office workers achieve their tasks and coordinate meetings.
The feedback from early alpha testers was encouraging and the idea seemed to resonate with a lot of people, especially top managers, our target user. We decided to start recruiting beta-testers to get a better idea whether or not Ophi could work as an independent company.
Ophi got more and more traction over weeks, with 370 people signing up to be beta testers our A.I assistant for work. The team got accepted to pitch the solution at the 2017 Chatbot Summit in Berlin, and everything was on track to spin-off a new venture.
Nevertheless, and even though we observed rapid improvements in the field of NLP, we started to understand that the technology wasn’t mature enough for us to achieve our vision. To design and develop a personal assistant that would save people and companies time and resources will still require a lot of technological improvements in parsing words and sentences, handling users’ mistakes and making sure the context of the conversations is still the right one.
Therefore, after a long reflection, we decided to park the project.
Looking back at the whole journey, it was a good introduction and learning experience for the team that enabled us to understand what the current possibilities and limitations of NLP and chatbots are. Along the journey, our developers and data scientists got to explore cutting-edge technologies in the field of NLP and AI, while designers got a chance to work with conversational interfaces.