CTO eating its own c(h)at food

After several years as the co-founder and CTO of a startup (Konverso), building an enterprise bot platform and ready-to-use bots for our customers and partners, I realized I was not actively using a virtual assistant for my own good! What would be the result ? Can I really find value for a personal virtual assistant that would help me in my daily tasks? This is the challenge I decided to take on at the end of this summer.

The environment

We are a team of about 20, including roughly 14 developers. We are super active, motivated, working on intense two weeks sprint cycles with a large number of enhancements and bug fixes. We build an enterprise-ready chatbot platform focusing on the IT world.

We use the Atlassian suite, leveraging Jira for the ticketing and the sprint tracking, Confluence for our various internal wikis (for Developer, for Product Documentation, etc). We also work with the Office 365 collaborative suite.

Our environment includes Linux VMs, managed through Ansible (host management) and Terraform (Azure deployment library) to run our DevOps activities on resources and services hosted by Microsoft Azure.

Defining the Use Cases

So… as a team we decided to build our own bot to help us in our daily activities. We agreed on starting with the following features:

• A dashboard of our key tasks as we open the chat, displaying links to our tickets in actionable status (to-do, in progress)
• Quick actions to Resolve or Comment on tickets, to get a list of tickets
• Search within our various internal wikis, with a display of the results embedded in the chat
• Being able to start and stop our VMs
• Ability to create Outlook and Teams meetings, to search in our O365 files on SharePoint and OneDrive
• Search for technical support from Microsoft, Google, StackOverflow public knowledge bases
• Propose technical tips and process reminders

That will do for a start. More use cases will probably be built and described in a future post.

Building it

The above was done over a couple of weeks, as a side task by the team. It was overall easy since we build bots for enterprise and most of these connectors were already available, the workflows for ticket management were in place and the conversational search on enterprise knowledge base was available. Building this from scratch on a GAFAM platform would have been much more involving.

And the result is?

Well, first of all, the most challenging thing is adoption. We have our usual tools, on classic web interfaces, we know them well. To win, the bot needs to be even more accessible, reaching it should be super convenient, with access through various interfaces, simple and quick. What makes it so valuable is that it is always there, ready to use, with no need to open yet another app. So we shared it via Skype, Teams, and web chats on our internal portal. It has also been required to discuss it daily in our team meetings, to keep the topic active, to demonstrate and discuss features, to push all team members to use it, to make small changes and surprise enhancements to the rest of the team.

So… it took several days to start using it, to make it a habit, and then it really worked out!

One immediate benefit I see is time saving. Modern web applications tend to have interfaces that overtime get nicer and nicer, which often translate to slower and slower. You add up the time to find the link or shortcut, the time to open the page, to authenticate, to navigate to find what we want! It really seems like we forgot, while using them day after day, how these graphical application interfaces add time everywhere. To search in a screen loaded with information, locate and move the mouse (yes, it is true, it can be painful on large or multiple screens), all these activities are time-consuming. To talk and express what you need and get it done immediately is truly a nice experience.

Another aspect I found extremely pleasant is the « aggregation » and « noise filtering » aspect, with numerous applications accessible in one place and a strong focus on the key features, the most used functions of our software. This filtering of functions through a number of described use case is also a good way to enforce particular procedures and work-flow. And because the conversation dialogs are much more effective to build and maintain, resources-wise it becomes very affordable.

It has only been a couple of weeks since we started this side project, but already, it fully convinced me. Working on a real personal use case actually worked. The most difficult thing, just like anything, is to figure out what you really need, make it simple and effective, do not try to overestimate what AI is capable of. I Am ready to sign a check to my own company for licenses! Next question, what is the value of such a productivity assistant. Maybe a topic for a future post. Stay tuned fellow Chatbot-curious readers.