While building Toss, I originally just had the idea of a meeting recorder that used AI to extract what you committed to on the call and offered you a very nice interface to have one click to complete those actions into the connected apps. However, I quickly realized that after a lot of calls there's both variable and invariable things that you want to do. So the variable things might be dependent on the type of call and it will change from call to call. For example, creating a linear issue or sending a customer follow-up a follow up email. These change from call to call and generally they're discussed on the call and are present in the transcript. Whereas after every call, you might want to always do something like sync the call to Notion or send a summary into Slack to your team. And these kinds of tasks won't be described in the transcript. So Toss handles both. Toss, the agent, can listen to your call and extract what you committed to on the call as action items and also run flows, which are natural language automations at the end of every call to sync the call to some destination or just carry out some kind of task that you always do at the end of every call, such as sending a summary into Slack.