Chega a ser contraintuitivo, mas 95% dos projetos de IA nas empresas americanas falham em gerar impacto real. 42% foram abandonados só em 2025. Não falhou por causa da tecnologia. Falhou por causa da estratégia. Três erros que se repetem: - O primeiro: implementam onde aparece, não onde funciona. 50% do orçamento vai pra vendas e marketing. Os maiores retornos estão no back-office, operações, compliance, jurídico. - O segundo: constroem o próprio LLM quando deveriam usar o que já existe. GPT, Claude e Gemini estão sendo atualizados toda semana com bilhões de investimento. Construir internamente não é estratégia. É orgulho. - O terceiro: confundem piloto com transformação. A regra 10/20/70 da BCG é clara: 70% do sucesso vem de pessoas e gestão de mudança. A maioria das empresas inverte isso e fica presa no piloto para sempre. Fontes: MIT State of AI in Business 2025, BCG 10/20/70 Rule. #ia #carreira #fofocas #estratégia

It was the focus and tendences of my group of ex-Aloos Harbour. The discussion of this week is turning around an American technology company that lost a millionaire contract to implement IA in their operation. And the worst was that they found that they were in the airport. It was then that someone brought a study of the EMITIC that proves that this type of situation is not common, 95% of the project of IA in companies failed to bring real impact. 32% of these initiatives were abandoned only in 2025. And it didn't fail because of technology, it failed because of the strategy. I summarized here 3 super-common errors in the implementation of IA that these studies brought. First mistake, normally, implement IA where it appears, not where it works. The EMITIMAPIRO where the companies are focusing on this type of project. 50% goes to sell and market. But most of them are in back off if it offers juridical and financial operations. Implementations very successful in these areas are around 2 to 10 million dollars per year, only in the production of third-hand contracts. The gas companies where it is easy to show off to investors. It doesn't have a real return. Second mistake is more common. The companies build their own land. When they should use what already exists, develop a language model that costs millions of dollars and take years. While this GPD, Claude, Gemini are being updated weekly with billions of dollars of investment. The company here insists on building their own model is not being strategic. It is boring. And being the last mistake, the companies confuse the pilot with the transformation. The BCG studied implementation of IA in hundreds of organizations. And it arrived in the name of the 10, 20, and 70. 10% of the success comes from the algorithm. 20 data technology, 70 for people and change management. Most investors spend 70% off the tool. 10% thinking how people will work differently. The company of technology that we discussed in my group didn't say because I chose IA. It said because it implemented in the wrong place and never came out of the pilot. And thousands of companies exactly repeating these three mistakes at this exact moment. The question is, is yours one of them?