Discover how the revolutionary AI Split node can transform your automation flows, making them smarter and more productive by gradually directing them towards the best performing paths. AI at your disposal for increasingly efficient automated flows.
Let's start with an example
A fashion eCommerce wants to understand which, between two versions of the same communication, converts more: Newsletter A - with a focus on images and graphic style or Newsletter B - with a focus on text and product description.
We can in this case use the new AI Split node.
In this case you will need to configure 2 exits, but for your more challenging automated flows you can go up to 7 exits with 7 different paths.
The automated flow self-evaluates itself and progressively moves towards the best path, the one that performs best.
How it works
At the initial stage, the node divides the incoming sessions randomly but equally distributed among the exits (e.g. if the node has two exits, the sessions will flow 50% from each exit).
After a sufficient number of sessions starts to evaluate how the automated flow is going, the percentages with which sessions are distributed among the exits are gradually adjusted to converge the majority of sessions towards the exit that maximises the objective.
How to set up
The AI Split node is designed to be simple and intuitive to configure:
- Number of exits: how many paths do you want to evaluate? The more exits you enter, the more paths you will have to think about and configure. You can start with 2 exits for your first AI automated flow, evaluate how it works and then you can modify it or copy the flow and modify the node.
- Recalculate exits: choose according to what you want to evaluate the automated flow. The objective can be either automated flow success or number of conversions.
By automated flow success we mean how many sessions ended by positive exiting an exit node. By number of conversions we mean the number of goal conversions that were made by contacts from the communications sent by the flow. - Conversion goal: if you have chosen to evaluate the success of the automated flow by the number of conversions, you have to choose which goal to evaluate.
Back to our example
If we come back to the example described above, we set:
- Number of exits: 2 corresponding to the two newsletters A and B.
- Recalculate exits: we set the number of conversions, as we want to know which of the two communications generates more conversions, so purchases.
- Conversion goal: we choose the goal that represents purchases.
At the beginning, the system will distribute the two exits equally, sending each newsletter to 50 per cent of the contacts. Then, after an appropriate period of time to evaluate the results, the system will start sending the newsletter that generates the most conversions to a higher percentage of contacts.
This results in increased automated flow efficiency, as more and more contacts will receive the best performing communication, optimising our chances of conversion.
As time goes by
The AI Split node will continue to learn and adapt to the performance of the two communications, automatically adjusting the distribution between the node's exits.
You will see the current distribution percentages within the node at all times.
Advantages
The advantage of the AI Split node is in its ability to learn and adapt. Unlike a static flow that can be ineffective if you have chosen an underperforming path, the AI Split allows you to continuously maximise the conversion or success capabilities of your automated flows.
With this self-learning system, you may also discover new and valuable information about your contacts, for example, who respond better to one particular style than another, allowing you to further refine your approach even in non-automation communications and interactions.