Assigning of an interaction to an operator device

Abstract

An entity (14), method and systems for automatically assigning an interaction between a respective user device (10-1, 10-2, … 10-N) and a user assistance centre (15) to an operator device (12-1) are described. The entity comprises a monitoring unit (141) configured for carrying out a monitoring of a first interaction between a first user (10-1) and the user assistance centre (15), wherein the first interaction is assigned to a first operator device (12-1). A prediction unit (142) is present to provide, while a second interaction is in progress and on the basis of said monitoring, a prediction of switching said first interaction to an artificial or human operator device, wherein the second interaction is an interaction between a second user (10-2) and the user assistance centre. An assignment unit (143) assigns, on the basis of said prediction, the second interaction (10-2) to one among said first operator device and another one among one or more operator devices.

Description

[0001] DESCRIPTION of the invention having the title:

[0002] Assigning of an interaction to an operator device

[0003] FIELD OF THE INVENTION

[0004] The present invention relates to an entity, method and systems for assigning an interaction to an operator device.

[0005] BACKGROUND OF THE INVENTION

[0006] User assistance centres that users can contact in relation to various applications are known. Such assistance centres are often characterised by a considerable number of incoming calls from various users. Furthermore, such assistance centres are often equipped with various operator devices, each provided to a human operator to allow them to interact with a user. Assistance centres in which a user interacts with an artificial operator, implemented by means of artificial intelligence for example, are also known. Even in the presence of an artificial operator, operator devices operated by a human operator may also be present. Despite various technological developments in the sector, the inventors have recognised margins for improving functioning of the technology that supports these assistance centres.

[0007] SUMMARY OF THE INVENTION

[0008] According to one aspect, the object of the present invention is to solve problems linked to functioning, or at least to improve functioning, or at least one or more technological components that support functioning of the assistance centres.

[0009] Some aspects of the invention can be described as follows:

[0010] Al. Entity (14) for automatically assigning an interaction to an operator device (12- 1), wherein the interaction is one among one or more interactions between a respective user device (10-1, 10-2, ... 10-N) and a user assistance centre (15), the entity comprising:

[0011] - a monitoring unit (141) configured for carrying out a monitoring of a first interaction between a first user (10-1) and the user assistance centre (15), wherein said first interaction is assigned to a first operator device (12-1);

[0012] - a prediction unit (142) configured for providing, while a second interaction is in progress and on the basis of said monitoring, a prediction of switching said first interaction to an artificial or human operator device, wherein the second interaction is an interaction between a second user (10-2) and the user assistance centre;

[0013] - an assignment unit (143) configured for assigning, on the basis of said prediction, the second interaction (10-2) to one among said first operator device and another one among one or more operator devices.

[0014] A2. Entity according to aspect Al, wherein the monitoring unit (141) is configured for carrying out a monitoring of said second interaction; the prediction unit (142) is configured for providing a prediction of switching said second interaction to an artificial or human operator device, on the basis of the monitoring of the second interaction; the assignment unit (143) is further configured for assigning, on the basis of said prediction of switching said second interaction, at least one of said first interaction and said second interaction to one among said first operator device and another one among one or more operator devices of said user assistance centre.

[0015] A3. Entity according to aspect A2, wherein the prediction unit (142) is configured to provide said prediction of switching said first interaction and said prediction of switching said second interaction according to a first time interval and, respectively, according to a second time interval.

[0016] A4. Entity according to aspect A2 or A3, wherein the first interaction is between the first user and a human operator and the prediction of switching said first interaction comprises a prediction of switching the first interaction to an artificial operator, and wherein the second interaction is between the second user and an artificial operator and the prediction of switching said second interaction comprises a prediction of switching to a human operator.

[0017] A5. Entity according to one of the preceding aspects, wherein said prediction indicates a time instant subsequent to the given time instant in which switching of said first interaction from an artificial operator to a human operator, or vice versa, is predicted.

[0018] A6. Entity according to one of the preceding aspects, wherein said prediction indicates a probability of switching said first interaction from an artificial operator to a human operator, or vice-versa.

[0019] A7. Entity according to one of the preceding aspects, wherein said prediction indicates a probability of switching, within a time instant or within a predicted time interval from a reference instant, said first interaction from an artificial operator to a human operator, or vice-versa.

[0020] A8. Entity according to one of the preceding aspects, wherein said monitoring comprises estimating a load factor of said first interaction and / or second interaction, and wherein the prediction is provided on the basis of this load factor.

[0021] A9. Entity according to one of the preceding aspects, wherein the prediction unit (142) comprises an artificial intelligence unit configured for providing the prediction of switching.

[0022] A10. Entity according to one of the preceding aspects, wherein the prediction unit (142) comprises an artificial intelligence unit configured for providing the prediction of switching.

[0023] All. Entity according to one of the preceding aspects, wherein said first operator device (12-1) manages at least one interaction between a user and an artificial operator while an operator of said first operator device is engaged in an interaction with another user, and / or manages one or more interactions between a respective user and a respective artificial operator.

[0024] A12. Entity according to one of the preceding aspects, wherein the first interaction is between the user and a human operator who uses the first operator device or between the user and an artificial operator.

[0025] A13. Entity according to aspect A12, wherein assigning of the first interaction to the operator device allows the human operatorto monitorthe interaction between the user and the artificial operator.

[0026] A14. Entity according to one of the preceding aspects, wherein the operator device (12-1) is configured to passively manage one or more interactions at the same time, each between a respective user and an artificial operator and in which a human operator of the first operator device is not actively involved.

[0027] A15. Entity according to one of the preceding aspects, wherein the assignment unit (143) is configured to assign the second interaction to said first operator device if the human operator of said first operator device is not predicted as engaged in active management of another interaction at the moment of switching of the second interaction, or if the human operator of said first operator device has the capacity to manage the second interaction passively.

[0028] A16. Entity according to one of the preceding aspects, wherein assigning of an interaction in passive management allows control and / or monitoring of the interaction by the human operator, without the human operator participating actively therein.

[0029] A17. Entity according to one of the preceding aspects, wherein the interaction is between said user device and an artificial operator, and the prediction provides a prediction of switching from the artificial operator to a human operator.

[0030] A18. Entity according to one of the preceding aspects, wherein said monitoring unit provides an indication of functioning, and wherein said assignment unit is configured to switch upon the indication of functioning that satisfies a predetermined criterion.

[0031] A19. Entity according to one of the preceding aspects, wherein the interaction is between said user device and a human operator, and the prediction provides a prediction of switching from the human operator to an artificial operator.

[0032] A20. Entity according to one of the preceding aspects, wherein said assignment unit is configured to switch from the human operator to an artificial operator upon the monitoring indicating a suitability factor for managing the interaction by means of an artificial operator.

[0033] A21. Entity according to one of the preceding aspects, further comprising an interaction association unit configured to associate one or more interactions between one or more respective user devices and an operator device. A22. Entity according to one of the preceding aspects, wherein the prediction unit comprises a computing unit trained to provide predictions on the basis of previous interactions.

[0034] A23. Entity according to one of the preceding aspects, wherein an interaction is divided into two or more phases, each phase being associated to one between a human operator and an artificial operator, and wherein the prediction is associated to each of said phases.

[0035] A24. Method for automatically assigning an interaction to an operator device (12-1), wherein the interaction is one among one or more interactions between a respective user device (10-1, 10-2, ... 10-N) and a user assistance centre (15), the method comprising the steps of:

[0036] - carrying out a monitoring (S10) of a first interaction between a first user (10-1) and the user assistance centre (15), wherein said first interaction is assigned to a first operator device (12-1);

[0037] - providing (S20), while a second interaction is in progress and on the basis of said monitoring, a prediction of switching said first interaction to an artificial or human operator device, wherein the second interaction is an interaction between a second user (10-2) and the user assistance centre;

[0038] - assigning (S30), on the basis of said prediction, the second interaction (10-2) to one among said first operator device and another one among one or more operator devices.

[0039] A25. Method according to aspect A24, wherein the monitoring (S20) comprises performing a monitoring of said second interaction; providing (S30) a prediction of switching comprises providing a prediction of switching said second interaction to a human or artificial operator device, on the basis of the monitoring of the second interaction; assigning (S30) comprises assigning, on the basis of said prediction of switching said second interaction, at least one of said first interaction and said second interaction to one among said first operator device and another one among one or more operator devices of said user assistance centre.

[0040] A26. Method according to one of aspects A24 to A25, comprising further providing said prediction of switching said first interaction and said prediction of switching said second interaction according to a first time interval and, respectively, according to a second time interval.

[0041] A27. Method according to one of aspects A24 to A26, wherein the first interaction is between the first user and a human operator and the prediction of switching said first interaction comprises a prediction of switching the first interaction to an artificial operator, and wherein the second interaction is between the second user and an artificial operator and the prediction of switching said second interaction comprises a prediction of switching to a human operator.

[0042] A28. Method according to one of aspects A24 to A27, wherein said prediction indicates a time instant subsequent to the given time instant in which switching of said first interaction from an artificial operator to a human operator, or vice versa, is predicted.

[0043] A29. Method according to one of aspects A24 to A28, wherein said prediction indicates a probability of switching said first interaction from an artificial operator to a human operator, or vice-versa.

[0044] A30. Method according to aspect A29, wherein said prediction indicates a probability of switching, within a time instant or within a predicted time interval from a reference instant, said first interaction from an artificial operator to a human operator, or vice- versa.

[0045] A31. Method according to one of aspects A24 to A30, wherein the monitoring comprises estimating a load factor of said first interaction and / or second interaction, and wherein the prediction is provided on the basis of this load factor.

[0046] A32. Method according to one of aspects A24 to A31, further comprising providing the prediction of switching by means of a prediction unit.

[0047] A33. Method according to one of aspects A24 to A32, further comprising providing the prediction of switching by means of a prediction unit (142). A34. Method according to one of aspects A24 to A32, wherein said first operator device (12-1) manages at least one interaction between a user and an artificial operator while an operator of said first operator device is engaged in an interaction with another user, and / or is managing one or more interactions between a respective user and a respective artificial operator.

[0048] A35. Method according to one of aspects A24 to A34, wherein the first interaction is between the user and a human operator who is using the first operator device or between the user and an artificial operator.

[0049] A36. Method according to aspect A35, wherein assigning of the first interaction to the operator device allows the human operatorto monitorthe interaction between the user and the artificial operator.

[0050] A37. Method according to one of aspects A24 to A36, wherein the operator device passively manages one or more interactions at the same time, each between a respective user and an artificial operator and in which a human operator of the first operator device is not actively involved.

[0051] A38. Method according to one of aspects A24 to A37, wherein the second interaction is assigned to said first operator device if the human operator of said first operator device is not predicted as being engaged in active management of another interaction at the moment of switching of the second interaction, or if the human operator of said first operator device has the capacity to manage the second interaction passively.

[0052] A39. Method according to one of aspects A24 to A38, wherein assigning of an interaction in passive management allows control and / or monitoring of the interaction by the human operator, without the human operator participating actively therein.

[0053] A40. Method according to one of aspects A24 to A34, wherein the interaction is between said user device and an artificial operator, and the prediction provides a prediction of switching from the artificial operator to a human operator.

[0054] A41. Method according to one of aspects A24 to A40, wherein the monitoring provides an indication of functioning, and assigning comprises switching upon the indication of functioning that satisfies a predetermined criterion.

[0055] A42. Method according to one of aspects A24 to A41, wherein the interaction is between said user device and a human operator, and the prediction provides a prediction of switching from the human operator to an artificial operator.

[0056] A43. Method according to aspect A42, wherein assigning comprises switching from the human operator to an artificial operator upon monitoring, by means of an artificial operator, indicating a suitability factor for managing the interaction.

[0057] A44. Method according to one of aspects A24 to A43, further comprising associating one or more interactions between one or more respective user devices and an operator device.

[0058] A45. Method according to one of aspects A24 to A44, comprising providing a prediction, comprising determining the prediction by means of a computing unit trained to provide predictions on the basis of previous interactions.

[0059] A46. Method according to one of aspects A24 to A44, wherein an interaction is divided into two or more phases, each phase being associated to one between a human operator and an artificial operator, and wherein the prediction is associated to each of said phases.

[0060] A47. Computer program comprising instructions which, when the program is executed by a computer, causes the computer to execute the steps of a method according to any one of the aspects from A24 to A46.

[0061] A48. System comprising an entity according to any one of the aspects from Al to A23, and a user device for connecting with said user assistance centre.

[0062] LIST OF FIGURES

[0063] Figure 1 schematically illustrates an example of an assistance centre to which the present disclosure can be applied;

[0064] Figure 2 schematically illustrates an entity according to an embodiment; Figure 3 illustrates a flow diagram relating to a method according to an embodiment;

[0065] Figure 4 schematically illustrates a computer;

[0066] Figures 5 and 6 illustrate examples of the assigning of an interaction according to the teachings of the present document.

[0067] DETAILED DESCRIPTION

[0068] Of the various user assistance centres, an assistance centre 15 reachable by one or more users 10-1, 10-2, ..., 10-N, as exemplified in Fig. 1, is considered here by way of example. In such an assistance centre 15, one or more operator devices 12-1, 12-2, ..., 12-M are present. Communication between a user and the assistance centre can be by means of a telephone network (landline or mobile), the Internet, etc. When a user connects to the assistance centre (for example, by making a telephone call, a video call, multimedia call, by using a bot / chat, etc.), an interaction is established between the user and the assistance centre. The interaction therefore represents the exchange of information between the user (who, for example, makes a request, such as how to solve a problem, etc.) and an operator (who, for example, provides an answer or information in response to the user). The interaction can take place between an artificial operator implemented, for example, by means of artificial intelligence or a human operator who uses one of such operator devices 12-1, 12-2, ..., 12-M. Such an operator may be actively engaged in an interaction with the user, for example by interacting in a voice and / or video call, by interacting by text or by interacting using multimedia. Said operator may be managing one or more interactions in which they are not actively engaged in the interaction (for example, the operator device can be configured to passively manage one or more interactions at the same time), since the interaction is carried out between the user and an artificial operator; for the sake of simplicity, we will say that managing one or more interactions corresponds with a passive interaction between operator and user, in the sense that, for example, the operator monitors the interaction between user and artificial operator, but does not necessarily participate actively in it. In other words, for example, assigning an interaction to the operator device allows the human operator (i.e. the operator who uses the operator device) to monitorthe interaction between the user and the artificial operator. In that case, the operator who is passively managing the interactions can advantageously use an automatic supervision or monitoring system that provides an indication of functioning of a respective interaction between user and artificial operator. This indication of functioning indicates how well the artificial operator is interacting with the user, for example how accurately and / or efficiently the artificial operator responds to the user's requests. Based on the indication of functioning, the system can indicate whether it is necessary to switch to a human operator. The indication of functioning can be, for example, an icon (which, for example, provides traffic-light type colours) that indicates the degree of interaction / functioning. A human operator can be actively engaged in (at least) one active interaction with a user and, at the same time, be managing (passively) one or more other interactions between one or more other respective users and the assistance centre. The artificial operator, represented by a combination of hardware and / or software, can be executed on any one of the operator devices of the assistance centre, and / or on any device (or combination of devices) at the assistance centre or outside it (for example, on the cloud); independently of where the artificial operator (i.e., the code thereof) is executed, an interaction between a user and an artificial operator can be assigned to an operator device of the assistance centre, so that the operator device can allow management thereof, including control, monitoring, etc. Such a configuration poses a series of technical problems, as recognised by the inventors. For example, imagine a scenario in which, while an operator is actively engaged in an interaction, the indication of functioning indicates that it is necessary to switch as soon as possible, ideally instantaneously and automatically, a certain interaction from an artificial operator to a human operator. It becomes difficult to manage such a scenario. In such a scenario, solutions are conceivable for managing the instantaneous load of each operator and therefore switching an interaction based on a load of one or more operators, such as, for example, a measured or estimated one (i.e., estimated on the basis of measurements, for example). However, this solution is not optimal and not easy to implement in practice, particularly when the intention is to achieve implementations that are as automatic as possible. In this context, the inventors have recognised the possibility of predicting (through estimates calculated automatically) if and when a certain interaction in progress must be switched from an artificial operator to a human one, or vice versa. Based on this prediction, the inventors therefore propose assigning an interaction to a certain operator device, so that said device can manage it actively or passively, depending on the circumstances. It thus becomes possible to improve the switching or assigning of interactions (in particular, for interactions that are in progress) to certain operator devices rather than others, thus achieving a better functioning of the system and the components of the user assistance centre. To this end, an entity 14 is proposed for managing interactions between users and the assistance centre, in order to optimise administration of interactions between users and the assistance centre. What is stated above will now be illustrated with reference to the drawings, without limiting what has been recognised by the inventors. Interaction means an interaction in any instant subsequent to the one in which said interaction has been started, i.e. subsequent to the instant in which the connection between user and assistance centre has been established.

[0069] An entity according to a first embodiment will now be illustrated with reference to Figure 2. The entity can be implemented in a localised manner and be represented, for example, by a server, or in a distributed manner and be represented, for example, by a combination or several servers, or by a combination of one or more servers and a device available to a human operator. Such entity (14) is configured to automatically assign an interaction to an operator device (12-1). The operator is one among one or more operators of the user assistance centre, and the operator device can be a device (or also a terminal or client, such as, for example, a computer, a tablet, etc.) used by the operator. The interaction is an interaction comprised between one or more interactions between a respective user device (10-1, 10-2, ... 10-N) and a user assistance centre (15). In particular, such interaction is an interaction already in progress, i.e. already established and started, independently of how it has been established or started.

[0070] The entity comprises a monitoring unit (141), a prediction unit (142) and an assignment unit (143).

[0071] The monitoring unit (141) is configured for carrying out a monitoring of a first interaction between a first user (10-1) and the user assistance centre (15), wherein the first interaction is assigned to a first operator device (12-1). Such interaction with the assistance centre can be, for example, between the userand a human operator (in which case there will be active management by the human operator) or between the user and an artificial operator (in which case, the human operator will be managing the interaction, i.e. the human operator will be passive in the sense that they are not interacting directly with the user). Therefore, in this last case, the human operator is managing the user-machine interaction in order, for example, to monitor the progress and / or to take control thereof. The system is preferably capable of passing automatically from artificial operator to human operator when monitoring reveals a certain criticality. The passage can be programmed thanks to the "prediction", as explained below.

[0072] The prediction unit (142) is configured to provide, while a second interaction is already in progress (for example, at a given time instant or at different time instants while the second interaction is in progress) and based on the monitoring, a prediction of switching the first interaction to an artificial or human operator, depending on the state it is in while the prediction is being made. For example, if the prediction is made while the interaction is between user and artificial (human) operator, the prediction relates to when to switch to the human (artificial) operator. As already stated, the second interaction is already in progress, i.e. both the interactions (at the moment when the prediction is determined) proceed in parallel and have already started at respective previous time instants. It is therefore a case of determining a prediction relating to an interaction in progress while another interaction is also already in progress.

[0073] The assignment unit (143) is configured to assign, on the basis of said prediction, a second interaction (10-2) to one among the first operator device and another one among one or more operator devices.

[0074] In other words, a first interaction in progress between a user and an artificial operator device or a human operator device is monitored, while a second interaction is also already in progress. Based on the monitoring (i.e., the output provided by the monitoring), a prediction (or, in other words, an estimate) is made on when the first interaction is predicted to be in a state in which it can be switched to the human operator device or, respectively, the artificial operator one. Thanks to the prediction, it is possible to predict the availability of an artificial or human operator device at a certain, subsequent time instant, so that it becomes possible to better allocate the resources needed to conduct the second interaction. In this manner, it is possible to optimise use of the resources of the assistance centre. For example, if the second interaction is with an artificial operator and the moment at which it must be switched to a human operator is pre-established or known, and the first interaction is with a human operator, it is possible to determine whether the second interaction must be assigned to the operator device currently engaged with the first interaction (since, based on the prediction, it is possible to anticipate when it will be free) or to another operator device based on the estimation of switching of the first interaction; thanks to this measure, it is possible to plan the use of resources in advance, for example also reducing the waiting times during switching (since it is less necessary to calculate the resources available at the moment when switching must be carried out), and therefore to improve or at least resolve several of the drawbacks mentioned here above.

[0075] Preferably, the monitoring unit (141) is further configured to perform a monitoring also of the second interaction when this is already in progress between a user and an artificial operator device or a human operator device; correspondingly, the prediction unit (142) is configured to provide a prediction of switching to a human operator device or, respectively, to an artificial one, based on monitoring of the second interaction. In this scenario, the assignment unit (143) is further configured for assigning, on the basis of the prediction of switching the second interaction, at least one of the first interaction and the second interaction to one among the first operator device and another one among one or more operator devices of the user assistance centre. It is therefore possible to plan and assign the interactions dynamically and flexibly. Furthermore, for example, such assignment can be made even when neither of the two interactions has a pre-established and known duration, since the switching instant is estimated for each of them. Examples will be illustrated below with reference to Figures 5 and 6.

[0076] Preferably, the prediction unit (142) is configured to provide the prediction of switching the first interaction and the prediction of switching the second interaction according to a first time interval and, respectively, according to a second time interval. In other words, the prediction of the first switching can be repeated in time according to a first time interval and, similarly, the prediction of the second switching can be repeated in time according to a second time interval. Each of these intervals can be pre- established or variable, also dynamically. The two intervals can be the same, but not necessarily. Furthermore, the instants in which the predictions of the first switching and the second switching are made can coincide, but not necessarily. The time interval for each determination of the predictions can be chosen (and / or adapted) depending, for example, on how reliable each estimate previously made is considered to be, or can be chosen so as to have an updated estimation value. In this manner, it is possible also to take account of how the interaction evolves and therefore to obtain an updated (and therefore more accurate) value of the switching estimate. The assignment is therefore further optimised.

[0077] Preferably, in an example, the first interaction is between the first user and a human operator and the prediction of switching the first interaction comprises a prediction of switching the first interaction to an artificial operator; in the same example, the second interaction is between the second user and an artificial operator and the prediction of switching the second interaction comprises a prediction of switching to a human operator. In this example, it is possible to estimate when the first interaction will cease to engage the operator; also estimating when the second interaction will need a human operator, it is possible to decide whether to assign such interaction to the human operator who is, at the moment of the prediction, engaged in the first interaction, or whether, on the other hand, to consider assigning such interaction to a further operator, who will, for example, be more likely to be free (from active interactions) at the moment of switching of the second interaction. In fact, as will also be explained below, it is possible to extend the above to several operator devices, in order to increase the efficiency of the assistance centre.

[0078] Preferably, the prediction indicates a time instant subsequent to the given time instant in which switching of the first interaction from an artificial operator to a human operator, or vice versa, is predicted. In one example, the prediction is a predicted time instant, or an estimated / predicted value that indicates such a time instant. In other words, the prediction indicates when the switching is (likely) to occur.

[0079] Preferably, the prediction indicates a probability of switching the first interaction from an artificial operator to a human operator, or vice-versa. Even more preferably, the prediction that indicates the probability of switching indicates a probability of switching, within a time instant or within a predicted time interval from a reference instant, the first interaction from an artificial operator to a human operator, or vice-versa. That is, the probability refers to a probability in which, at a certain, future time instant, the switching can take place. For example, the entity can estimate that the switching will take place with a probability of 50% within the next N minutes (e.g., 5 minutes) and / or with a probability of 90% or more in the next 10 minutes; in another example, the probability can indicate that the switching will take place after 5 minutes (taking account of a tolerance interval) with a probability of 50%, and / or that it will take place after 10 minutes with a probability of 90% or more.

[0080] The operator device can optionally be configured to display an indicator for the first prediction and / or the second prediction. For example, the operator device can display an indicator of the probability of switching associated with a future instant, optionally for several future instants. For example, a bar can be displayed comprising three values indicating three future instants starting from the moment of display (e.g., 3min, 5min, lOmin), with a probability associated with each of those instants (e.g., 25% associated with 3min, 50% associated with 5min, 90% associated with lOmin). The probability can be represented by a colour in order to make the representation more compact on a screen. Such an indicator can be provided for the first interaction and / or second interaction (and more generally, forany interaction in progress that it is intended to monitor, for example for purposes of assignment).

[0081] Optionally, the monitoring performed by the monitoring unit (141) estimates a load factor of the first interaction and / or the second interaction, wherein the prediction is provided on the basis of this load factor. The load factor can be, for example, a weight and / or a complexity of the first interaction. The load factor can vary over time, depending on how the interaction is proceeding. In these examples, the monitoring determines the load factor and / or the complexity of the interaction on the basis of measurements made on the interaction. Measurements made on the interaction means an analysis of the interaction, as also explained below by means of examples. Once the weight and / or complexity has been determined, it is possible to estimate how long the interaction continues with the respective artificial or human operator. This estimate is therefore considered in assigning an interaction when the latter must be switched. Taking an interaction with a human operator as an example, it is possible to analyse the content of the interaction, from which analysis of the complexity thereof can be obtained (for example, simple question to be answered, or generic question that requires further analysis), from which, in turn, the duration up to the point when this interaction can be switched to an artificial operator is estimated. Rather than the content, or in combination with it, other measurements or analysis performed on the interaction can also be considered, such as, for example, the emotional state of the user (since an indication of dissatisfaction indicates that more time will be necessary to deal with the matter), the state of the interaction, etc. The emotional analysis can be obtained by analysing the tone of voice, physiological parameters of the user (for example, parameters observable or obtainable from an image of the face, parameters relating to the heartbeat, as detected, for example, by a smart watch, etc.) or other known techniques. The state of the interaction is an indication of how the interaction is evolving, for example compared to a previous measurement / observation, so it can be used to update the estimate of how much longer the interaction could last; in another example (as an alternative or in addition), the state of the interaction is based on analysis of the content to establish the estimated state of progress compared, for example, to a recognised pattern for that content, etc. In this example of a human operator, the load factor therefore indicates how much the resources of the human operator are engaged in the conversation, in particular for how much longer based on analysis of the interaction. On the other hand, taking an interaction with an artificial operator as an example, the content of the interaction is analysed and, from this, how many resources are being consumed by the artificial operator and / or how many resources are needed by the human operator to manage the interaction is obtained; these resources are therefore indicative of a weight that said interaction entails. The weight can be used by the assignment unit to establish, for example, whether such interaction can be managed by a specific operator device in view of the load present at a certain moment, or whether, given that, for example, said device is already overloaded, to switch said interaction to another operator (i.e., how many resources are used at said operator device).

[0082] Optionally, the prediction unit is configured to provide the prediction based on one or more phases into which the interaction can be divided. In other words, an interaction can be optionally divided into one or more phases, and the prediction can be provided for one or more phases into which the interaction has been divided. In particular, the prediction can be provided on the basis of an estimate for one or all of the phases into which the interaction is divided. In fact, for certain phases, an estimate of the duration of said phase is sometimes available; for example, once the phase type inside the interaction has been determined, the prediction relating to that phase can be determined on the basis of an estimate available forthe phase (phase type), forexample on the basis of historical data for that phase type.

[0083] Optionally, the prediction unit (142) comprises an artificial intelligence unit (Al unit, not shown) configured to provide the prediction of switching of the first interaction and / or the second interaction. For example, said artificial intelligence unit can use machine learning models to analyse historical trends and / or real-time data relating to interactions. Such Al unit can be trained on the basis of any combination of an analysis of the content of the interaction, the state of the interaction, the emotional state determined for the user during an interaction, etc. Such Al unit can therefore predict the probability that one or more interactions being managed change weight and complexity in the short term, i.e. can provide a prediction of switching. Data on the operator can also be taken into consideration. In one example, the Al unit is trained with a dataset of past interactions, and such dataset is structured so as to comprise indications on one or more of: the content of the interaction (for example, type of request made by the user); stage of the interaction; user's emotional level, preferably during performance of the interaction or at least in one or more time instants during the interaction, even more preferably at the end of the end of the interaction; operator's identifier; historical data on how the operator has managed past interactions; length of the interaction; weight and / or complexity of the interaction; length of each stage into which the interaction can be divided, etc. A duly trained Al unit therefore manages to provide an estimate of how much longer an interaction in progress may last before it can or must be switched. Optionally, the Al unit is configured to provide the prediction of switching for the first interaction and / or for the second interaction. Optionally, the first operator device (12-1) is managing one or more interactions between a user and an artificial operator while an operator of said first operator device is engaged (actively, for example through a voice, text, multimedia, etc., interaction) in an interaction with another user or is managing one or more interactions between a respective user and a respective artificial operator.

[0084] Optionally, the interaction is between the user device and an artificial operator, and the prediction provides a prediction of switching from the artificial operator to a human operator. The prediction therefore allows scheduling of assigning of the interaction to the operator (or to another operator, depending on the corresponding prediction). This allows the switching times to be optimised, thanks to scheduling of resources based on the predictions, thus avoiding resources from being underused.

[0085] Optionally, the monitoring unit provides an indication of functioning, and the assignment unit switches upon the indication of functioning that satisfies a predetermined criterion. For example, when the indication of functioning indicates that an interaction in progress between user and artificial operator has reached a state of non-functioning (for example, dissatisfaction and / or frustration of the user, inability to answer the question posed by the user, etc.), switching to a human operator is performed (automatically or manually). The prediction previously provided had, in any case, predicted, taking account of the usual tolerances, the instant at which this would have occurred and / or the probability with which this would have occurred at a certain instant; switching can therefore be performed more rapidly and efficiently. Monitoring of the indication of functioning can be performed as described, for example, in the Italian patent application with filing number 102023000023706 and filed on 9 November 2023 by the same applicant of the present one, also applicable to interactions with a human operator, correspondingly considering the information exchanged between user and human operator.

[0086] Optionally, the interaction is between the user device and a human operator, and the prediction provides a prediction of switching from the human operator to an artificial operator.

[0087] Optionally, the assignment unit is configured to switch from the human operator to an artificial operator upon the monitoring indicating a suitability factor for managing the interaction by means of an artificial operator. Switching can be immediate, orwithin a certain time that allows the operator to say goodbye or to complete the conversation. Switching performed by means of the operator device can be automatic or in response to a command given to the user device. The indication of suitability can be provided by means of an indication of functioning (see above), or provided by means of a command provided to the operator device.

[0088] Optionally, the entity comprises an interaction association unit configured to associate one or more interactions between one or more respective user devices and an operator device.

[0089] Optionally, the prediction unit comprises a computing unit trained to provide predictions on the basis of previous interactions. The computing unit can, for example, comprise the Al unit.

[0090] Optionally, an interaction is divided into two or more phases, wherein each phase is associated to one between a human operator and an artificial operator, and wherein the prediction is associated to each of said phases. That is, it is possible to perform a more specific prediction upon one or more phases, in order, for example, to schedule possible switches in various phases of the interaction. This increases the flexibility of switching, taking account of the resources available in each phase of the interaction.

[0091] What is stated above, with particular reference (but not limited to) the entity, also applies for other embodiments and examples described below (and vice versa), unless there are indications to the contrary. Repetitions will therefore be avoided.

[0092] With reference to Figure 3, an embodiment relating to a method for automatically assigning monitoring of the interaction to an operator device (12-1) will now be illustrated, wherein the interaction is one among one or more interactions between a respective user device (10-1, 10-2, ... 10-N) and a user assistance centre (15). The method comprises a phase S10 of carrying out a monitoring of a first interaction between a first user and the user assistance centre, wherein the first interaction is assigned to a first operator device (12-1). The phase S20 envisages providing (S20), while a second interaction is in progress and based on the monitoring, a prediction of switching the first interaction to an artificial or human operator device, wherein the second interaction is an interaction between a second user (10-2) and the user assistance centre. The phase S30 provides for assigning, based on the prediction, the second interaction (10-2) to one between the first operator device and another between one or more operator devices.

[0093] With reference to Figure 4, an embodiment will now be illustrated relating to a computer program comprising instructions which, when the program is executed by a computer, cause the computer to execute the steps of the above method or cause the computer to perform functions corresponding with those described here above with reference, for example, to the entity or further on with reference to examples. Figure 4 shows a block diagram of a device that can represent a computer. In one embodiment, the device 70 includes at least one processor 71, a memory 72 and an I / O interface 73. The processor 71 is configured to carry out instructions comprised in a computer program stored in the memory 72 in order to perform any one of the functions of the entity, the method or the examples as described herein. The memory 72 is configured to store the computer program. Furthermore, the memory 72 can also store additional information, which can comprise data, models trained to be used by the monitoring using Al, etc. During the carrying out of the instructions of the computer program, the processor 71 receives information from other devices via the I / O interface 73 and sends, as output, information to the other devices via the I / O interface 73. Such interface can also be used to exchange input / output with an operator. It is understood that the various elements of the device 70 need not necessarily be located in a single housing, since the box is purely illustrative. For example, the processor 71 can correspond to a server situated on a rack, the I / O 73 can correspond to a router and / or to a switch connected to the server and the memory 72 can correspond to a cluster storage system comprising several storage units. The I / O 73 can include a display and / or a connection to a display, an audio system and / or a connection to an audio system, etc. The device 70 of Figure 4 can also schematically represent an operator device or a user device, opportunely and correspondingly programmed.

[0094] In another embodiment, a system is provided that comprises an entity as described here above (including the optional functions thereof) and one or more user terminals. In a further embodiment, a system is provided that comprises an entity as described here above (including the optional functions thereof) and one or more operator terminals. In a further embodiment, a system is provided that comprises an entity as described here above (including the optional functions thereof), one or more user terminals and one or more operator terminals.

[0095] An example of functioning according to what is stated here will now be stated with reference to Figure 5, in which three interactions 11, 12 and 13 are represented (the number three is purely illustrative, as there could be only two or even more than three) with the passing of time (indicated on the horizontal axis with t). In the example in question, all three interactions are between a respective user and an artificial operator. The figure shows three artificial operators OA1, OA2 and OA3, which can be executed by the same distributed entity or localised and which can represent three instances of the same Al engine or instances of different Al engines (Al engine means an entity that executes a certain Al). In the example in question, the three interactions 11, 12 and 13 are being managed at an operator device OD1 (not shown; it can be represented by one of the devices 12-1 ... 12- M of Figure 1) operated by a human operator OU1 (not shown); the human operator OU1 is not engaged at instant tO in any active interaction with users. The entity 14 provides, at instant tO, a prediction of switching foreach ofthe interactions 11, 12 and 13. Such prediction can be represented by an indication on the time instant in which each interaction 11, 12 and 13 can or must preferably (meaning as soon as possible, for example if a human operator is not available) be switched to a human operator; in another example, as shown in the figure, the prediction indicates a probability Pl, P2 and P3 that each interaction 11, 12 and 13, respectively, will be switched at a respective time instant. In the case in which predictions Pl and P2 are more or less equal, the entity determines that one of the two interactions can be switched to the operator device DOI of the human operator OU1, while the other of the two interactions must be switched to another operator device DOn. Thanks to these predictions, the entity can therefore consider sufficiently in advance which other operator will be available and, for example by applying the same principle to the other interactions and to the other operator devices, which other device can take the active interaction, minimising the times in which the resources remain unused. In this manner, the entity is able to plan (perform a scheduling) as well as possible use of the resources of the assistance centre. In addition, preferably, the entity 14 can determine the load of the interactions 11, 12 and 13 (or 13 in combination with one of 11 and 12, after switching) being managed at the operator device DOI and determine whether it is possible to accept a new passive interaction and how many (in other words, for example, whether the human operator DOI has the capacity to manage a new interaction passively). In a time instant subsequent to switching, the entity can perform the same process with the interactions available at that moment; for example, assuming that the interaction 11 is switched to the operator device DOI and the interaction 12 to another device, the entity can provide a prediction of switching of the interaction 13, possibly in combination with a prediction relative to the interaction 11, in order to plan whetherthe interaction 13 can continue to be managed with the operator device DOI or whether it is better to switch it to another operator device. Such switching to another operator can be performed before the switching instant, for example based on the load factor of the operator devices. This allows the resources of the assistance centre as a whole to be advantageously balanced.

[0096] Another example is now illustrated with reference to Figure 6, in which three interactions 11, 12 and 13 associated with an operator device DOI operated by a human operator OU1 (the number three is also used here for illustrative purposes only) are shown. In the example, the interaction 11 represents an active interaction between a user and the human operator, while the interactions 12 and 13 represent interactions between respective users and artificial operators OA2 and OA3. The entity provides predictions Pl, P2 and P3 of switching for the interactions 11, 12 and 13, respectively. In the example shown in Figure 6, P3 indicates that the switching instant (from artificial operator to human operator) for the interaction 13 probably falls while the first interaction is still in progress, i.e. while the operator device DOI is still engaged in an interaction between user and human operator. The entity can therefore determine that it is more opportune to assign the interaction 13 to another operator and that, on the other hand, the interaction 12 can continue to be managed with the operator device DOI. Assignment of the interaction 13 does not necessarily need to be performed at the predicted switching instant P3, but can be performed before such instant, for example taking account of the load factor of other operator devices.

[0097] In the examples shown, the prediction can once again be provided at an instant subsequent to the instant tO (for example, at regular intervals), so as, for example, to take account of possible unpredictable changes in the interactions in progress (for example, call gets cut off, etc.); in this manner, it is possible to update the values Pl, P2 and P3 and therefore to increase the accuracy of assignment of the interactions. It is possible to provide the prediction once again or repeatedly even only for one or for a subpart of the monitored interactions 3 / o for which a prediction had been provided at a previous instant.

[0098] The examples discussed consider a single operator device DOI for illustrative purposes. However, it can be correspondingly applied to two or more operator devices of the assistance centre and to two or more of the interactions assigned to the respective operator device. In this manner, it is possible to assign interactions between operators, improving scheduling of the resources of the assistance centre.

[0099] In the examples, the prediction is provided on the basis of the monitoring. This can be executed continually, or at pre-established or dynamically determined time instants. It is conceivable to perform the monitoring at instant tO in which the prediction is provided (noting that the instant tO includes an interval close to this or that precedes it by a small amount of time, taking account, for example, of the times required for analysis of the data collected during monitoring).

[0100] Many of the embodiments and examples have been illustrated with reference to the entity. However, what is described can also be described or implemented in a corresponding method, and / or in a computer program to be run on a processing entity (for example, a computer) or on an entity having suitably configured means. As illustrated here above, the entity can be implemented in a single device, via HW / SW or a combination thereof, or on multiple units or interconnected devices (which are also HW, SW or a combination thereof). Furthermore, features such as monitoring units, prediction units, assignment units, can be replaced, equivalently, by monitoring means, prediction means, assignment means. Each of the units described here is intended as obtained by means of any software and / or hardware combination, in a localised manner on one device or distributed on several devices. Naturally, the description set forth above of embodiments and examples applying the principles recognised by the inventors is provided solely for the purpose of illustrating such principles and must thus not be understood as limiting the scope of protection of the invention claimed herein.

Claims

CLAIMS1. Entity (14) for automatically assigning an interaction to an operator device (12- 1), wherein the interaction is one among one or more interactions between a respective user device (10-1, 10-2, ... 10-N) and a user assistance centre (15), the entity comprising:- a monitoring unit (141) configured for carrying out a monitoring of a first interaction between a first user device (10-1) and the user assistance centre (15), wherein said first interaction is assigned to a first operator device (12-1),- a prediction unit (142) configured for providing, while a second interaction is in progress and on the basis of said monitoring, a prediction of switching said first interaction to an artificial or human operator device, wherein the second interaction is an interaction between a second user device (10-2) and the user assistance centre;- an assignment unit (143) configured for assigning, on the basis of said prediction, the second interaction to one among said first operator device and another one among one or more operator devices.

2. Entity according to claim 1, wherein the monitoring unit (141) is configured for carrying out a monitoring of said second interaction; the prediction unit (142) is configured for providing a prediction of switching said second interaction to an artificial or human operator device, on the basis of the monitoring of the second interaction; the assignment unit (143) is further configured for assigning, on the basis of said prediction of switching said second interaction, at least one of said first interaction and said second interaction to one among said first operator device and another one among one or more operator devices.

3. Entity according to claim 2, wherein the prediction unit (142) is configured to provide said prediction of switching said first interaction and said prediction of switching said second interaction according to a first time interval and, respectively, according to a second time interval.

4. Entity according to claim 2 or 3, wherein the first interaction is between the first user and a human operator and the prediction of switching said first interaction comprises a prediction of switching the first interaction to an artificial operator, andwherein the second interaction is between the second user and an artificial operator and the prediction of switching said second interaction comprises a prediction of switching to a human operator.

5. Entity according to one of claims 1 to 4, wherein said prediction indicates a probability of switching, within a time instant or within a predicted time interval from a reference instant, said first interaction from an artificial operator to a human operator, or vice-versa.

6. Entity according to one of the preceding claims, wherein said monitoring comprises estimating a load factor of said first interaction and / or second interaction, and wherein the prediction is provided on the basis of this load factor.

7. Entity according to one of the preceding claims, wherein the prediction unit (142) comprises an artificial intelligence unit configured for providing the prediction of switching.

8. Entity according to one of the preceding claims, wherein said monitoring unit provides an indication of functioning, and wherein said assignment unit is configured to switch upon the indication of functioning that satisfies a predetermined criterion.

9. Entity according to one of the preceding claims, wherein said assignment unit is configured to switch from the human operator to an artificial operator upon the monitoring indicating a suitability factor for managing the interaction by means of an artificial operator.

10. Entity according to one of the preceding claims, wherein an interaction is divided in two or more phases, each phase being associated to one between a human operator and an artificial operator, and wherein the prediction is associated to each of said phases.

11. Entity according to one of the preceding claims, wherein said first operator device (12-1) is managing at least one interaction between a user and an artificial operator while an operator of said first operator device is engaged in an interaction with another user, and / or is managing one or more interactions between a respective userand a respective artificial operator.

12. Entity according to one of the preceding claims, wherein the first interaction is between the user and a human operator who is using the first operator device or between the user and an artificial operator, wherein, preferably, assigning of the first interaction to the operator device allows the human operatorto monitorthe interaction between the user and the artificial operator.

13. Entity according to one of the preceding claims, wherein the operator device (12- 1) is configured to passively manage one or more interactions at the same time, each between a respective user and an artificial operator and in which a human operator of the first operator device is not actively involved.

14. Entity according to one of the preceding claims, wherein the assignment unit (143) is configured to assign the second interaction to said first operator device if the human operator of said first operator device is not predicted as being engaged in active management of another interaction at the moment of switching of the second interaction, or if the human operator of said first operator device has the capacity to manage the second interaction passively.

15. Entity according to one of the preceding claims, wherein assigning of an interaction in passive management allows control and / or monitoring of the interaction by the human operator, without the human operator participating actively therein.

16. Method for automatically assigning an interaction to an operator device (12-1), wherein the interaction is one among one or more interactions between a respective user device (10-1, 10-2, ... 10-N) and a user assistance centre (15), the method comprising the steps of:- carrying out a monitoring (S10) of a first interaction between a first user device (10-1) and the user assistance centre (15), wherein said first interaction is assigned to a first operator device (12-1),- providing (S20), while a second interaction is in progress and on the basis of said monitoring, a prediction of switching said first interaction to an artificial or human operator device, wherein the second interaction is an interaction between a second user device (10-2) and the user assistance centre;- assigning (S30), on the basis of said prediction, the second interaction to one among said first operator device and another one among one or more operator devices.

17. Computer program comprising instructions which, when the program is executed by a computer, cause the computer to execute the steps of the method according to claim 16.

18. System comprising an entity according to any one of claims 1 to 15, and a user device for connecting with said user assistance centre.

Images