The AI introduced for alternative resource discovery in the summer of 2023 is now available and in use for all customers. Initial experiences have been positive, but a series of questions have been raised, which I would like to summarize here:
How does the AI identify individuals?
The AI analyzes the relevant group of individuals to determine their availability. Not all resources are checked, as it would be too time-consuming. Preferably, individuals from the same department as the overloaded resource are considered, assuming that they can perform similar tasks. The neural connection between individuals is strengthened or weakened based on user feedback, where users either accept or reject the suggestion. This iterative process improves the results over time.
Why do I sometimes not receive a suggestion?
The AI has likely identified individuals, however, these individuals are also occupied. This would not solve the problem, so the AI has not found anyone with the capacity.
Sometimes I receive only one suggestion, sometimes five?
The AI searches for a maximum of 5 individuals as alternatives. The person who best fits, based on the weighting of the network, is presented first. However, sometimes the AI only finds one or two individuals where the neural connection is sufficiently strong and they are available.
The first suggestion takes a moment, the others come immediately?
The AI identifies up to 5 suggestions in one go, which takes a moment. Only the first suggestion is then presented, but the other suggestions are already determined and are displayed very quickly afterward since no further calculations are needed.
If I assign a person with over 100%, I still receive suggestions that do not solve the problem, right?
That is correct, and it's a bug that we are fixing.
Can the training data of the AI be accessed?
Currently not, but we will introduce an app in Q1/24 that will showcase the relationships.
Can the training data be deleted?
Yes, it can be done by your administrator or for the cloud solution, through a brief request to us.
Are the training data used across companies in the cloud?
No, the training data is exclusively available per customer. There is no possibility of using the data across customers. This is also not practical, as each company has different individuals.
When will the next version of the AI be released?
In the first quarter of 2024, we will release an expansion. Alongside the suggestions, users will be able to manually choose an alternative resource. However, we still need to align the network to determine the extent of the AI's influence on such decisions.
Is it documented in the history when I replace a resource with the AI?
No, there is no distinction in the history regarding how the resource was replaced.
Are there currently known errors in the AI?
Except for the described bug above, which we are fixing immediately, no other malfunction is known.
How long does the AI take to provide meaningful suggestions?
This depends on usage. The more the AI is used, the better the suggestions become. Initial customer feedback suggests a timeframe of 3-6 weeks.
Does Can Do analyze the AI data?
We only analyze data from customers who have expressly granted us permission and are actively involved in the further development of AI through ideas and suggestions. All other customers are not evaluated.