preloader

Authority Magazine: 5 Things I wish I Had Known About AI

By Paul Davis, Metropolis Corp


In November of 2023, Paul Davis, the AI engineer behind Metropolis Corp's leading AI-enriched solutions for Unified Communication Analytics and Call Center Dashboards, was interviewed by Authority Magazine about his knowledge of AI and how he felt it had changed the Unified Communications industry. Read the full article for an in depth dive into how AI has changed CX for SaaS products.

In addition to his thoughts on how things would change CX in the future, he discussed the five things he wish he had known when he started out learning AI several years ago. As an early adopter of ChatGPT, here are a few tips Paul shared:


Number 1: Proper prompting is an art.


Crafting the perfect prompt is an art. It's not just about asking the AI a question; it's about framing it in a way that guides the AI towards the answer you're looking for.


Number 2: Data Quality is Junk IN = Junk OUT


The power of AI comes down to the data you feed it. Machine learning models learn from data, so the quality and diversity of the data you provide will directly impact the model's performance.

From a data perspective, junk in = junk out. It is best to have better quality data than a large quantity of less useful information. It is the same from a prompting perspective; your AI response is only as good as the data you've trained it on. It is extremely important to prime your prompt.


Number 3: Track your interactions


Track every interaction you have with AI. Master the art of prompt templates, categorizing them, utilizing agent personas to accomplish tasks. This way you can continue to adapting and tweaking the prompts that work best for you and you don't have to start from scratch.

In my role as a Solution Architect, I often use a collection of pre-defined prompts to automate tasks. For instance, if I need to generate a risk assessment report, I have a specific prompt that I use to get the most accurate and relevant information.


Number 4: Provide feedback


Be nice and give your AI positive feedback. Giving it the thumbs up when it provides you information you need, and a thumbs down when you were looking for something else will train the AI model and improve your future responses as it trains.

When initially training our customer support AI, we found that providing positive reinforcement when it generated good responses significantly improved its accuracy and reliability.


Number 5: Evaluate


Talk to the AI as you would a peer and don't be afraid to humble yourself. Remember that if you're not getting the response you thought you'd get, it's probably not the AI hallucinating, it's how you're communicating with the AI.


In two years, we can expect to see a remarkable change in how businesses communicate and interact. While a lot of jobs will be phased out, new opportunities will be created too. So, be prepared for the exciting changes that lie ahead as AI continues to shape the future of our world, and start learning how to not just use it, but master it.

Related Topics