On that platform I developed an artificial intelligence assistant capable of responding to the technician from the installation itself using real information from the equipment and the available documentation, instead of just generating a generic response.<\/p>\n A technician may find themselves in front of a UPS with an alarm and have all the information necessary to resolve it: equipment data sheet, installed batteries, previous interventions and hundreds of pages of manuals.<\/p>\n The problem is locating The right information in a few seconds<\/strong>, especially during an intervention.<\/p>\n The need was clear: to allow the technician to ask questions naturally and receive an answer based on the actual information available to that team.<\/p>\n Need:\u00a0<\/strong>Reduce the time spent consulting manuals and historical records during an intervention.<\/p>\n <\/p>\n The key decision was not to build a generic chatbot.<\/p>\n The assistant consults the information following a hierarchy of sources:<\/p>\n 1. Team and client<\/strong> 2. Technical documentation<\/strong> 3. Web<\/strong> In that context, Claude generates the response indicating diagnosis, procedure, reference values, and source used<\/strong>.<\/p>\n The system was integrated directly into EnerviApp so that the technician can consult the assistant without leaving their workflow.<\/p>\n The application combines structured MySQL information with pre-processed technical documentation and contextual retrieval before sending each query to the model.<\/p>\n The result is a tool capable of answering questions such as:<\/p>\n \u201c10 kVA UPS at customer's premises, low battery alarm after discharge test. What should I check first?\u201d<\/p><\/blockquote>\n and automatically relate them to that specific equipment, its configuration, its history, and the corresponding manual.<\/strong>.<\/p>","protected":false},"excerpt":{"rendered":" AI applied to UPS\/SAI technical service to convert documentation, history and equipment data into useful answers directly in the field.<\/p>","protected":false},"featured_media":4986,"template":"","portfolio_category":[166,169],"portfolio_tag":[279,278,280,276,277,275,194],"portfolio_industry":[178,176],"class_list":["post-4829","portfolio","type-portfolio","status-publish","has-post-thumbnail","hentry","portfolio_category-diseno-web","portfolio_category-inteligencia-artificial","portfolio_tag-api","portfolio_tag-claude","portfolio_tag-ia-aplicada","portfolio_tag-javascript","portfolio_tag-mysql","portfolio_tag-php","portfolio_tag-rag","portfolio_industry-energia","portfolio_industry-servicios-profesionales"],"acf":[],"_links":{"self":[{"href":"https:\/\/germanmallo.com\/en\/wp-json\/wp\/v2\/portfolio\/4829","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/germanmallo.com\/en\/wp-json\/wp\/v2\/portfolio"}],"about":[{"href":"https:\/\/germanmallo.com\/en\/wp-json\/wp\/v2\/types\/portfolio"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/germanmallo.com\/en\/wp-json\/wp\/v2\/media\/4986"}],"wp:attachment":[{"href":"https:\/\/germanmallo.com\/en\/wp-json\/wp\/v2\/media?parent=4829"}],"wp:term":[{"taxonomy":"portfolio_category","embeddable":true,"href":"https:\/\/germanmallo.com\/en\/wp-json\/wp\/v2\/portfolio_category?post=4829"},{"taxonomy":"portfolio_tag","embeddable":true,"href":"https:\/\/germanmallo.com\/en\/wp-json\/wp\/v2\/portfolio_tag?post=4829"},{"taxonomy":"portfolio_industry","embeddable":true,"href":"https:\/\/germanmallo.com\/en\/wp-json\/wp\/v2\/portfolio_industry?post=4829"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}![\"\"](\"https:\/\/germanmallo.com\/wp-content\/uploads\/2026\/07\/inteligencia-artificial-SAI-tecnico-Enervia.jpg\")
<\/figure>\n01 \u2014 Understand<\/h4>\n
The problem wasn't finding information. It was finding it in time.<\/h5>\n
\n02 \u2014 Choose<\/h4>\n
AI, but with context and sources.<\/h5>\n
\nModel, battery configuration, history and previous interventions.<\/p>\n
\nMore than 10,000 fragments of manuals<\/strong> processed and indexed by page and section.<\/p>\n
\nOnly when internal documentation does not contain sufficient information.<\/p>\n
\n03 \u2014 Build<\/h4>\n
From technical data to a useful answer in the field.<\/h5>\n