LKS Next – Deliveron
LKS Next
Sector: Professional activities, Scientific activities, Technical activities
Business Case
The maintenance staff, who are in charge of carrying out repairs, often go to the customer simply to check the problem they are facing and return to the head office to obtain the necessary materials. This wastes a great deal of time.
Objetivos
Develop repair manuals based on the information provided by the customer when reporting the incident. This allows the maintenance staff to take the necessary materials the first time, thus saving unnecessary trips.
Use case
After the client notifies the technical support service (SAT, in its Spanish initials), the conversation is transcribed to text format, and its respective embedding (position within a vector space) is calculated. Once the embedding is obtained, the warning is associated with a repair manual generated for this type of fault (faults with a similar embedding). Consequently, the maintenance staff may take the repair materials suggested in the manual.
Infraestructura
On Premise
Tecnologías utilizadas
AI technologies that generate written or spoken language, images or videos (generative AI) Voice recognition Text mining
Datos utilizados
Conversations between customers and the SAT, database of old fault reports (devices, type of fault, date of fault, date of sale of the device, etc.) and texts with information on repair processes.
Recursos utilizados
Assignment of personnel with expertise, servers to host the application and access via services to language models.
Dificultades y aprendizaje
It is understood that from now on unstructured data such as free text, voice, images, etc. are a very valuable resource and that it is important to adapt them in order to be able to avail of them.
KPIs (impacto en el negocio y métricas del modelo)
Once the maintenance technician arrives at the repair site, he/she determines whether the generated manual has been useful or not, providing feedback to the model that generates such manuals. This will reduce the error rate over time and consequently fewer trips will have to be made for repairs.
Financiación
Clients and public funding
Colaboradores
No
