At Penta, we’re leaning into all things Generative AI and helping our clients do so as well. As a data native firm, we’ve been in the NLP/machine learning business for well over a decade, and even our product engineers were floored by the capabilities of the likes of GPT4, Bard, Claude, and others. While our internal product team is working on exciting innovations like tailored news summaries that use AI to identify what matters, our communicators and strategists are focusing on ways GenAI can help us be more thoughtful, creative, and productive in our day-to-day work. Since we’re all about knowledge sharing at Penta, we wanted to share the five things we’re thinking about when deploying AI in communications:
Top-down encouragement: People fear change. Most people, especially if they are busy, aren’t going to spend their spare time figuring out how they can leverage GenAI to improve their productivity at work. It has to be encouraged from the top, or else adoption and innovation will stagnate.
Understand the (very real) limitations: What is so remarkable about GenAI is the outputs are really confident but can be totally wrong. These so-called “hallucinations” could include citing case law that doesn’t exist or attributing quotes to people who never said them. This means users of GenAI tools need to be particularly aware of these shortcomings and never treat any GenAI-created output as final without reviewing, editing, and fact-checking it.
Establish clear guardrails: In light of these limitations, it’s essential to ensure your team can benefit from the power of GenAI while minimizing the risks. Successfully doing so requires organizations to be clear about the rules of the road. Your team should know what kind of tools are available and which ones need heightened scrutiny; what information should not be shared; and what the expectations are for ensuring the accuracy of outputs. These are game changing tools, but to use them effectively in the near-term, organizations should be explicit about the dos and don’ts.
Train your staff: From an understanding of the tool’s capabilities to prompt construction best practices, arming your staff with the right information is crucial for a smooth and efficient AI journey. Through experimentation, we’ve learned that what and how you ask any GenAI tool for help really makes a difference. Crafting a great GenAI prompt is all about setting the stage for success. Here are some key components you can mix and match to achieve that when using your AI-powered chatbot of choice:
- Define the task: This is the foundation of your prompt. Be clear and specific about what you want your GenAI tool of choice to do. Instead of saying “write a tagline for a new healthcare app,” try “write a 15-word catchy and playful tagline for a new app that detects early signs of disease.”
- Indicate a voice: Defining the voice of your request will make the response more aligned with what you are looking for. Do you want the response to be serious and academic, or playful and informal? Specifying the desired tone will elevate the quality of your responses.
- Establish an audience: Depending on what your ask is, it might be helpful to include the intended audience in your prompt. For example, you could say, “write a 500-word oped on the economic benefits of free trade for a California audience.”
- Set constraints: Adding restrictions like word limits and format type can lead to more precise results.