Más allá de la prueba inicial: claves para escalar los proyectos de inteligencia artificial | Estrategias digitales

Recommended:

Nueve formas prácticas de optimizar tus costes IT Leer

Data cloud intelligent, clave para tu transformation digital In the report

Test info:
VMware Work from home Documento

The adoption of artificial intelligence (IA) by companies from all sectors is increasingly widespread and, in fact, it will be one of the IT segments that will grow the most in the next few years, according to MarketsandMarkets. This consultancy estimates that this technology will generate revenues of 407,000 million dollars in 2027, which represents a compound annual growth rate of 36.2%.

Sin embargo, according to atSistemas, despite the fact that it is a mature technology, the great majority of IA models remain in the initial testing phase, due mainly to the effort necessary to scale the model. “The lack of progress in the industrialization of these models supposes a great inefficiency both in terms of resources for the company and for employees who do not have access to technologies that could facilitate their daily work,” the consultant said.

Sus consejos para conseguir escalar los modelos o “industrializarlos”, son los siguientes:

Aumentar la calidad, trazabilidad y accesibilidade de los datos
Los resultados obtenidos por cualquier modelo de inteligencia artificial, machine learning o deep learning, están directamente relacionados con la calidad de estos. Por ello, si una organización tiene intentione de industrializar su IA, el primer paso para logarlo sería centrarse en sus bases de datos. Usually these are dispersed, and even in many cases in isolated silos, which makes it difficult to use large-scale AI models, because if the work team only has access to a subset of data, it will never be able to generate models that reflect the reality of the organization.

In this sense, the objective of the organization must be to generate data products with higher quality availability and easier access for operation, to avoid that each of the areas of the company seeks to create its own copy to cover its specific needs y evitar duplicidades. Once the information is centralized, it will be essential to guarantee the quality of the data, both from the structural point of view and from the business perspective. Es decir, que además de estar correctamente complimentados, tengan coherencia y aporten valor para los empleados y la empresa.

Apostar por la interacción multidisciplinar
Artificial intelligence is directly associated with innovation, and the key to innovation is in poder conjugar different perspectives on the same topic, so that between them they can be enriched and they can have a more complete vision. Por tanto, cuantos más groups trabajen juntos, mejores serán los resultados. This is also essential to achieve this industrialization, because the global contribution and collaboration of the different teams involved is necessary to achieve an optimal result.

Sin esta cooperation, es realente complicado construir modelos que aporten un valor added to la totality de la organization y se podría incurrir en el problema de generar ecosistomes isolados. It is also necessary that the collaboration of the different teams is produced in all cycles, from the intake of data to the implementation of the analytics, working in a synchronized manner so that the data is available in the shortest possible time, with good quality and ease access.

Crear entornos operativos
No sólo vale con disponer de buenos datos y las tools para extracter conclusiones utiles para la empresa, hay que preparer a la empresa para el uso de esos datos creando una estrategia global. The company must identify which of its operations must be modified to allow the adoption of new technologies, and above all to establish standards for the creation, testing and deployment of new AI models.

In this way, it will be much more easy to repeat and expand the new models as they are created. One of the essential parts of the strategy is establishing where the selection of algorithms and the development of models resides, in the technical and functional equipment, and in dotarlos de resources to achieve greater autonomy without losing control over the development.

.

Leave a Comment

Your email address will not be published. Required fields are marked *