Kanban Boards Getting AI Summaries and Anomaly Detection

We can hear you okay at the minute, Ana, give it a go and I'll jump in if we can't hear you again and then we can just go to the next one and come back to you.

Ana Loga 21:35

Okay, perfect. Thank you. Okay, so we integrated Kanban with assistance builder and we trained our assistant to analyse pre-computed metrics and from the client side in order to generate an overview for the board, highlighting any anomalies if they exist and recommend further suggestions.

So this feature can only be triggered by board owners and board admins because the generated AI summary can be sent to board stakeholders. To mention that for now, we only support sending the summary through feed, but after we will implement e-mail sending. we will be able to send it through e-mail as well. So having this board example, we can see it has a lot of data, which is also pretty complex.

So if we open the dashboard, we can see that the board has a lot of not started tasks. overdue tasks, some blocked task as well. The RAG status is critical. Also, the team workload is, the team members are also loaded up with a lot of tasks.

And if we generate an AI summary based on all the data. We can see that the the the assistant generated the three sections: overview, anomalies, and recommended steps. For the overview section, it starts with a sentence about the board purpose, which is derived from the board description, which is also fully visible here.

Then it analyses the stats for the current board state. It says how many tasks they are in total, how many are completed, indicating a low completion rate of 7.7%. A significant portion of tasks are overdue.

And it says how many days they are late. Then there are some more planning gaps, obvious, with five not started tasks, lacking due dates. It also specified that workload distribution is uneven, with this user handling twice the board average.

And it ends with a general statement. saying overall the board is struggling with overdue task bottlenecks and uneven workload requiring urgent intervention to improve its trajectory. Okay, and then we continue with the anomaly sections, which are grouped by severity. We have high severity, medium severity, and low severity.

So a few examples are for high severity are the tasks are being blocked and specifying the assignments of those tasks. then how many tasks are overdue, how many of them are high or critical priority, and also specifying the assignees for those. Also, if a task has been in progress for more than 10 days, it specifies its name and the number of days it's been stale. And also, if there is a bottleneck in a column, for example, for the to-do column, we have 16 tasks, which is more than twice the average task count, creating a bottleneck.

Then for medium severity, we have if a task has an approaching due date, it's specified and also its assignee is specified as well. And also, if a team member is over-assigned, it will also appear in the medium severity category. For example, this user is over-assigned with 14 tasks, which is double the board average of seven.

And then we have some low severity examples. For example, one in progress task has no due date set. And another one that one parent task remains open despite all subtasks being completed and specifying the name of the task as well.

All the anomalies category groups are based on our pre-computed calculations and from the client side. So these are stating that the assistant just make an analysis over them. And based on this, it recommends some suggestions.

For example, it suggests that we should address the overdue task by prioritising higher critical ones. Then it suggests we should resolve the bottlenecks in the to-do column. by moving the tasks, also to assign the three unassigned tasks on the board, set due dates for all tasks, especially those in progress, and review block tasks and collaborate with assignees to remove obstacles. From this view, we can either copy the data summary or share it with the board stakeholders.

So if we click on this and we cheque the stakeholder, it's this user and on this browser M log on that user, we can see that the summary has been sent through feed. So the notification was received. The only thing is that the feed items are limited in have a character limit and the summary may not be fully visible if it's long.

But that's why we are waiting to have an e-mail center. So where on e-mail, the summary will be fully visible for the stakeholders. And the feed notification will serve more as a reminder for the stakeholders.

Okay, and this is all regarding the summary part. Now, if I go back, I can show how we set up the Kanban as a connector. And I can quickly show this from the researcher point of view.

So I created a project where I added the Kanban as a connector. and I prepared some questions. So for example, if I ask it to create a summary for all the boards that I currently have access to, and I specify what I want to know for each board, it should create it. Yeah.

It. Okay, so here we can see a summary for all my boards. It says I have 11 boards, three boards visible, and this is true because I have three boards visible and the rest of them are hidden.

I can see the overview for each board. For this example, which I previously showed for the AI summary, we can see that the task completion rate is 7%, exactly as we had it in the dashboard. Average progress, 21%, RAG status. how many stakeholders we have members and specifying my current role on the board.

And let's... Continue with the next prompt. For example, if I want to know how many tasks I have assigned in that board example across all columns.

It says I have 15 tasks assigned. Let's also cheque this information. If I philtre by my assignee, then we have 8, 6, and one.

So it's a total of 15 tasks. It's also making an overview over town. Yap.

And let's also test the word creation and task creation. So, I tell it to create a board and do tasks inside the board. Okay, the board should be created.

Yes, and also the two tasks. These are just a few examples of using Kanban as a connector, but we can continue of experimenting with it. Regarding the demo, this would be everything.

So I am now passing to Andrea from Survey. If there are any questions, please pop them in the chat.