What should CEOs do about ChatGPT and Generative AI?

Many enterprises, including industry giants like Amazon, Apple, JPMorgan Chase, Meta, and Verizon have banned ChatGPT due to concerns over sensitive data leakage.?

CEOs face critical decisions on how to harness AI’s potential. This post will offer practical recommendations to help you navigate AI adoption confidently and securely.?

The data leakage concern

The foremost worry for business leaders is the risk of sensitive data falling into the wrong hands. Companies must safeguard valuable information like employee compensation, product plans, and expansion strategies at all costs. OpenAI, the developer of ChatGPT, acknowledges this concern. At a recent tech conference CEO Sam Altman emphasized, “We don’t ever train ChatGPT on data submitted through an API.”?

However, two challenges persist for OpenAI:?

• Building trust: convincing business leaders to trust this assurance is crucial for widespread AI adoption.?
• Communicating clearly: OpenAI’s technical and legal jargon make it challenging for CEOs to grasp how client data will be protected.?

What early adopters are doing?

Early adopters have followed one of two distinct paths:?

1. Spending a fortune on AI: some enterprises are building expensive private data islands, which are often prohibitively expensive and resource-incentive.
2. Embracing unnecessary risks: others are taking the plunge, yearning for potential rewards and accepting the risk of possible consequences, such as data being exposed to the public.

Recommendation for CEOs to dip their toes in the AI waters

Instead of standing still, there’s a way to strike the right balance between innovation and security. I recommend the following strategy, which is 10x cheaper and 10x safer.?

Select a suitable dataset, one that’s not too big and not too small, not too sensitive and not too plain. Allow for effective experimentation without compromising critical information.

• Don’t boil the ocean (of data)! Instead of overwhelming yourself with vast amounts of data, start with a small, manageable dataset for experimentation. Contrary to recent trends of accumulating data in expansive repositories like data lake houses and warehouses, a more focused approach yields better results.
• Don’t play with fire. Avoid tinkering with data that, if exposed, could lead to severe consequences. Steer clear of compensation data, legal contracts, and details related to acquisitions and layoffs. No one needs a potential public relations nightmare.

Avoid on-premise environments to control costs. Instead, use a cloud-based large language model. You can choose from many alternatives of two types:

• A proprietary one like ChatGPT, Google’s Bard, Antropic’s Claude, and others
• An open-sourced one like Meta’s LLaMa, Dolly, StableLM, ChatGLM, StarCoder and others (These are less expensive but require development resources or consultants.)

Use chatbots to replace dashboards. If you are wary of stale dashboards and unending PowerPoint presentations, consider leveraging bots as a faster and more dynamic means of accessing information. They can quickly surface relevant insights.

Practical example: Insights on revenue-generating initiatives

To illustrate the practical applications of AI in decision-making, consider the following fictitious example based on our experience at Biztera.?

The CEO wants to get insights on the initiatives she’s approved in Biztera. Engaging in a conversation with a bot, she gets answers on anticipated revenue and weighs past performance.

Conclusion

You can harness the power of AI while ensuring the safety of your sensitive information. As Marc Andreessen, co-founder of Andreessen-Horowitz put it, AI tools like ChatGPT are like “eager puppies wanting to make you happy.” If the power of AI can help your employees be more productive, that’ll undoubtedly make you happy.?