The adoption of generative artificial intelligence (GenAI) has rapidly increased since the release of ChatGPT in 2022. However, there remain challenges to implementing GenAI solutions, especially for specialized business cases. According to Forbes, the technology behind text-based GenAI is based on large language models (LLM), which require extensive resources and a high level of complexity […]
The adoption of generative artificial intelligence (GenAI) has rapidly increased since the release of ChatGPT in 2022. However, there remain challenges to implementing GenAI solutions, especially for specialized business cases. According to Forbes, the technology behind text-based GenAI is based on large language models (LLM), which require extensive resources and a high level of complexity to create. As a result, few businesses have the knowledge or skills to directly work with LLMs.
In response to this, businesses now have access to pre-existing LLMs that are customizable and more accessible for technology teams with experience. Furthermore, there are several software platforms in almost every business function offering out-of-the-box GenAI features that do not require specialized skills or knowledge. However, the success and advancement of GenAI in the business world heavily rely on the quality of data used by businesses.
Invalid or low-quality data is a significant limiting factor in the adoption and effectiveness of GenAI. As the technology operates on a different level of speed, scale, and complexity, it has significant implications for data governance. Operational teams are required to curate and cleanse structured and unstructured datasets that support GenAI to deliver accurate and reliable insights. The use of GenAI models changes the way businesses deal with data quality issues, as they no longer predict data sets to be cleansed in advance.
The insights generated by GenAI models are displayed unpredictably, and the measurement and analytics teams may not be able to control the gateways through which end-users query data. GenAI grants access to a vast repository of data to make intuitive leaps to support user queries, allowing businesses to generate insights at a scale previously unachieved. Additionally, managing security, privacy, and consent requires businesses to implement new processes that don’t exist yet.
Managing data quality for GenAI use cases demands a different set of skills, and businesses are required to retrain their teams to manage new concepts. The role of operational teams in partnering with technical resource teams has become crucial in generating trustworthy responses from any GenAI tool. Data steward roles have to expand their ability to provide domain expertise and take on new roles as the arbiters of accurate insight generation. Data management has to move from cleansing the control of discrete data sets into the ongoing, active curation of conversations, both prompt and response.
In conclusion, data quality remains crucial to the success of GenAI adoption in business. GenAI may present new challenges in the collection, analysis and governance of data. To prepare for these challenges, businesses must invest in retraining their technology teams to work with large datasets, as well as redefining their internal processes to meet emerging GenAI data governance requirements.