Generative AI, one year later: moving beyond the hype
This is guest blog post by, James Fisher, Chief Strategy Officer, Qlik
Almost one year ago, ChatGPT exploded on the scene and quickly took AI mainstream.
Since then, an endless number of articles have been written, reports published, and products launched. Along with that flurry of activity came a fear of the unknown paralyzing many organisations. Should you act quickly to avoid being left behind, or proceed with caution to avoid major missteps? Is it even possible to drive value for your business with generative AI today? And is this all just all hype, anyway?
Well, the hype may be real, but so is the value potential for your business. But if your company has not quite figured out a path forward, you are not alone: only 39% of organizations have formalised an AI strategy, according to a new survey commissioned by Qlik.
So, where do you begin? It starts with your data. Generative AI relies on vast amounts of training data to learn patterns, relationships, and structures, serving as the bedrock for creating meaningful outputs. So, laying a strong, trusted data foundation is where it all begins.
Start by assessing the readiness and needs of your data fabric. Are you able to bring data together from multiple sources? Can your data can be organised and trusted? Can the insights from your data be consumed in context with other analytics? Are your systems are connected so they can bring the AI ouputs into your operational systems and processes to trigger action?
If the answer is “no” to most or all of these questions, rest assured that here again, you are not alone: only 20% of organizations from our survey indicated that their data fabric supports generative AI very or extremely well. At the same time, the majority recognise that generative AI places more demands on data; 73% expect the amount of data moved or managed on their current analytics to increase.
Here are my recommended actions to help get your data fabric ready.
- Establish a trusted data foundation
Data privacy and security is paramount to a successful implementation of generative AI; in our our survey respondents rated this as top concern.
Ensuring your data foundation is secure and ready for AI involves efficiently integrating data from diverse sources into a secure repository, continuously updating it in real-time to enhance model performance, and optimising data transformation based on your target system’s requirements. Ensuring data quality through the use of high-quality data during training is critical for reliable and accurate AI outputs. Additionally, implementing strong data governance practices, including policies and technology for responsible data management and protection, is essential to ensure the ethical and effective utilization of data in generative AI.
- Leverage the power of all AI
A data fabric that is AI-ready also includes an anal ytics experience that is enhanced by AI. Here all AI matters: traditional AI and generative AI. They have different use cases, but both hold immense value and huge untapped potential and are a rising tide for each other, so it’s important to look at your entire AI strategy as one.
Today traditional AI has a wide range of very effective applications, including automated insights (“how did my sales last quarter compare to the prior year?”), predictive modeling (“what will my sales look like next month and what’s driving it?”), intelligent alerting (“alert me if you detect a customer’s spend exceeding a certain threshold”), and natural language processing for tasks like text classification (“extract and categor ise names of people, organizations and contract value”). Generative AI is a fantastic new capability, but it is still an emerging technology that is best suited for use cases around content generation and summarization or extending the capabilities of traditional chat bots.
- Consider your Language Learning Model (LLM) options
Finally, as you consider your options for implementing generative AI in your organization, my recommendation is that you use proprietary data to minim ise security risks. In our study, the option most considered by respondents (68%) is using public or open-source models refined using their own proprietary data. A smaller percentage (48%) are also considering building a model from scratch.
Whichever option you choose, you will need to prepare that model with data curated for generative AI before you design the model, and there are considerations as well on deployment.
Through it all, there is good news. While there has been much focus on generative AI, it’s still early days. Yes, generative AI will be transformative, but we are in the infancy of its potential and adoption. At the same time, AI is more than generative AI: traditional AI remains prominent and a value driver. And if you’re looking for a great partner on your AI journey, Qlik can help: we have long-standing leadership with AI across a great portfolio of data integration, data quality and analytics.