Chief Technology Officer, Dun & Bradstreet.
Genetic artificial intelligence (AI) has the potential to change central aspects of how companies do business. Whether we’re taking on repetitive tasks or quickly analyzing terabytes of data, the hope is that technology reveals new efficiencies that make people’s jobs easier.
Unfortunately, few people have the necessary background to understand how genetic AI works. the inherent risks of large language models (LLMs) such as ChatGPT; and the potential for things to go wrong. Several publishers have recently learned this the hard way, allowing AI tools to “write” error-ridden articles without human oversight before publication. So these companies have become cautionary tales as organizations try to figure out how genetic AI should be used in their industries.
I have now seen that there is great value in having a safe place to experiment with this technology – an AI sandbox where novices and software engineers can interact with LLMs, discover new business cases and make mistakes without risking embarrassment or damage to the substance.
Cracking The Business Case For AI
Many executives are under tremendous pressure to develop an AI strategy for their companies. There is an expectation that everyone should go full steam ahead with AI. I have found that few leaders know what to do with such a transformative technology at this early stage. Most implementations will fall into one of two broad categories for companies.
First, genetic AI will have internal use cases that create new processes, help people work more efficiently, and reduce costs. Imagine an AI capable of “reading” invoices, recognizing when information is missing or incorrect, and monitoring suppliers to revise their documents. This allows professionals to be available for more complex tasks.
Second, many companies will want to include artificial intelligence in their own products. There are already several writing tools that have AI assistants that can create outlines, provide suggestions, and produce articles with minimal human supervision. This is sold to customers as a competitive element and a better way of doing business (although some publishers may now disagree).
AI can help companies explore both types of opportunities, but success isn’t as simple as asking ChatGPT to share an incredible new proposition to improve margins or increase sales. Employees must have experience with AI tools to write prompts and build apps that meet their expectations. This knowledge only comes with experience.
View failure as an option
The bad news is that experience with artificial intelligence is lacking in many organizations. While AI production tools are readily available, how they work is a mystery to many of those at the forefront of this technology. The sandbox AI shows its value here in two main ways.
Some companies offer AI training to their employees so they can understand and extend their technology beyond software developers and data scientists. If AI is to be deployed across the enterprise, it is imperative that people in every department share at least a basic understanding of it. Memorizing letters can go so far in building knowledge. allowing people to test what they have learned allows them to make mistakes, refine their opinions and improve their skills.
Sandbox AI isn’t just for newcomers. Software engineers and data scientists can push the boundaries of genetic AI during internal hack-a-thons. They can test practical applications of products or features in the early stages of development, eliminate bugs and deal with unwanted results. Giving workers the freedom to fail can be a powerful thing in building AI credibility.
How to start
There are several key elements to developing a safe space for AI innovation. The first is to have reliable data, as working with comprehensive and accurate information is the only way to understand how a product will perform in the real world. D&B.AI Labs provides such a dataset for us that matches our traditional strengths and core competencies.
Next, the AI sandbox must be secure. This certainly means that it is not accessible to the general public and may include licenses for internal users, so the work is private until it is ready to be shared. Many free AI tools that seem to offer a place for experimentation have privacy policies that suggest what you’re doing is anything but private, so it’s important to review the policies before deciding whether it’s the right choice for your company.
Third, the effort must be supported by education. This, like a lab portion of a college class, is where theory is put into practice. Workers of all skill levels should have access to experts who help them understand AI.
The last element is planning how the features developed inside the sandbox will work. To do this responsibly may require the help of external partners well versed in the intricacies of genetic AI. These experts would understand the framework underlying the LLM and the information it references, anticipating issues with data lineage or usage rights that could present problems when commercializing a tool. Offloading this oversight allows your employees to focus on making AI work for your company as opposed to navigating fundamental questions about the technology.
Exposure to the different GenAI technologies, platforms and engines so you can see which of the tools could lead to the best possible result is crucial. Different platforms, models and methodologies can all lead to different results. Experience comes through trial and error, and the ability to iterate with these different factors can achieve the best business outcome. This is so important that we’ve made it a core feature of our D&B AI Labs, where we work with our customers to deliver just that experience.
While technological change can be disruptive, most people come from a place of limited knowledge about artificial intelligence. This gives businesses the opportunity to invest in the professional development of their employees in a way that benefits both the individual and the organization. It will require training, time and space for people to learn from their mistakes.