Watch the above video or the below instructions to understand workflows better.

Any conversation in the world can be thought of as a set of goals that are to achieved one after the other, and the next goal to be achieved is based on the intent of the user at the end of the current goal.

At the heart of Atoms lies a powerful graph-based system called the workflow which enables breaking a conversation down into sub-goals or nodes, and based on the intent or branches move to the next goal/node.

The above image describes a simple workflow of atoms.

It contains 3 default nodes, 2 end call nodes and 5 branches.

In this example, we show a simple Insurance lead generation workflow.

  1. In the first node Introduction and Offer, the agent introduces itself and tries to see if the customer is interested in buying home insurance.
  2. In case the customer is not interested, the agent will proceed to the Disinterest Closure node. In this node, the agent simply thanks the customer for their time and cuts the call.
  3. In case the customer is interested, the agent will proceed to the Qualify Home Insurance Need node. Here, the agent checks if the customer currently has home insurance or is looking actively. This identification of intent is extremely important.
  4. In case the customer says they are not looking or already have something, then the agent goes back to 2.
  5. In case the customer is looking, then the agent proceeds to the Collect Basic Qualifying Data node. In this node, the agent collects the name and a suitable time to call back.
  6. Finally, the agent proceeds to confirm the call back time and cut the call.

We hope that you get a brief understanding of how a workflow graph works. Once you get the hang of it, you will be able to build infinitely long and complex workflows for any usecase in the world.

While it appears complex at first, when you realise that nodes are given goals and not exactly what has to be spoken, it becomes extremely intuitive and simple to build the workflow.

In the next sections, we will dive deeper into nodes and branches and best practices around them.