![\"\"](\"https:\/\/www.europesays.com\/uk\/wp-content\/uploads\/2025\/06\/iStock-1503658451.jpg\")
<\/p>\n<\/p>\n
When data software vendor Cloudera<\/a> hosted a lunch for just under 20 corporates at the Gartner Data & Analytics Summit<\/a> in June 2025, all of the participants put their hands up when they were asked if they were pursuing GenAI projects.<\/p>\n Then Cloudera\u2019s chief technology officer for Australia and New Zealand, Vinicius Cardoso<\/a>, asked how many were confident in their implementations.<\/p>\n \u201cThere was one hand,\u201d says Cardoso.<\/p>\n \u201cIt was all about trust. Trust in the data, that it was the right data, and trust in the models they were using.\u201d<\/p>\n In meeting this challenge of trust, Cardoso talks about two concepts, both beginning with the letter \u201ch.\u201d<\/p>\n Firstly, data needs to be \u201chuman centric,\u201d which means that it ultimately needs to be controlled by humans \u201cwith the intent to produce a good outcome for citizens, consumers, and for people in need.\u201d<\/p>\n Keir Garrett<\/a>, Cloudera\u2019s managing director for Australia and New Zealand, says that in seeking to build trust, many organizations at the conference were putting a priority on the idea of \u201cdata lineage\u201d with their AI projects.<\/p>\n \u201cI think this is part of the maturity curve,\u201d says Garrett. \u201cSo many organizations are on the AI journey, but they are waking up and saying, \u2018yes, we did pocket AI, and while some of it helped us make good decisions, also by chance we may have made catastrophic decisions because the data was the problem.\u2019\u201d<\/p>\n \u201cSo this is really encouraging because in fixing the data, they are looking at the governance and compliance, strengthening their teams in this area, and this is also putting a focus on where the data comes from,\u201d she adds.<\/p>\n Once this governance level is assured, and organizations can be confident that they are using the \u201cright data,\u201d the next step is around operational performance, and this is where the other \u201ch\u201d word comes in: hybrid.<\/p>\n With data coming from so many disparate sources, data management platforms should be unified by cannot afford to be monoliths.<\/p>\n \u201cYou don\u2019t have to retrain those models in mass data sets, but you enable them in new data sets, and this helps them stick to the facts.\u201d<\/p>\n<\/blockquote>\n Data is managed in the cloud, on premises, and at the edge. There is historical data, real-time data, and data from third parties.<\/p>\n It comes from sensors in the field, from social media and legacy databases, and much of its distribution and processing \u2014 certainly when used or GenAI \u2014 must happen in real time and is often best powered by Retrieval Augmented Generation (RAG) frameworks.<\/p>\n \u201cRAG helps the large language models to implement their learning with the data sets you have right now,\u201d says Cardoso.<\/p>\n \u201cYou don\u2019t have to retrain those models in mass data sets, but you enable them in new data sets, and this helps them stick to the facts, be very specific, and minimize AI hallucinations.\u201d<\/p>\n Life and death situations<\/p>\n In some cases, the technology can make a difference in life and death situations, examples that shine a light on AI\u2019s transformative potential.<\/p>\n\n