As a software engineer, I love finding applications for new technology, so when OpenAI first debuted their GPT-3 API, an artificial intelligence large language model (LLM), I was immediately encouraged by the quality of its output and breadth of capability. As Vice President of Online Ordering at Paytronix, I also wanted to explore how GPT could help our clients.
Now, with the release of ChatGPT, LLMs have gone from niche techie tool to mainstream adoption in record time. OpenAI officially launched it in November 2022, and anyone can try it out and get familiar with how it works.
I got my hands on a beta version back in 2021 and experimented with it both as a feedback response vehicle and for its language translation abilities. I had tried other AI LLM options, but OpenAI’s models were the only solution that offered what Paytronix needed.
How ChatCPT Can Help
The beta version of GPT-3 confirmed my suspicions that it could facilitate great customer communication for the Paytronix guest surveys and feedback module, an integral part of Paytronix Online Ordering. Responding quickly to guest concerns keeps them engaged, lifting overall customer lifetime value and preventing negative public reviews on platforms like Google and Yelp.
The Paytronix Online Ordering platform automatically engages guests within two hours of their order. Guests rate their experience with star ratings and written comments, and the results are immediately sent to a store manager, who also can see the guest’s lifetime value. Designed to help store managers measure and mediate guest interactions, this process identifies guests with unhappy online ordering experiences. A quick response is often the difference between keeping a guest returning and losing them for life.
Previously, store managers would respond to an unhappy guest by using either generic, pre-written emails or customizing one for their experience. Neither situation is ideal, since one isn’t personalized, and the other is time-consuming.
Here’s where GPT really shined. By creating prompts based on the guest’s experience, GPT generates a personalized, on-brand message that can be quickly reviewed and sent.
As this technology continues to advance, the process of reviewing and responding to guest issues will move further along the automation curve. The responses will get better and the process of sending them will get even faster.
Solving the Integration Puzzle
While much of the hype around ChatGPT has focused on the public version, Paytronix worked with OpenAI to license a premium version of GPT, the best solution for embedding in commercial products like Online Ordering. Once we could access the GPT application programming interface (API), Paytronix had access to OpenAI’s GPT-3.5 model and all its bells and whistles.
The next step in making this a usable solution was “prompt engineering,” which requires engineers to write prose, not code, to test the AI chatbots. At Paytronix, our prompt engineers wanted to create relevant, personal responses that would resonate with guests. If the model generates only generic apologies, it adds little value over the already-built templates. Through prompt engineering, we can create specific, on-brand guest responses that a store manager would be happy to send.
This is a case where content truly matters. When our engineers built prompts using direct quotes from guests, the responses hit the right notes. As we continue to develop these capabilities, we can further automate the response process, saving store managers even more time—and in some cases, totally automating the entire guest interaction.
Beyond the Email
While useful for guest engagement, GPT offers far more potential applications, and our teams continue to explore ways to add value through AI. We have seen promise with language translation, instantly making a store’s menu available in a multitude of languages, or generating copy for menu item descriptions, for example.
In addition to generating responses and content, GPT can also summarize and categorize raw information, like guest commentary. By adding TL;DR (Too Long; Didn’t Read) into the prompt, the AI can generate a synopsis of a month’s worth of guest feedback in seconds, providing an actionable summary to store management.
Of course, automation is dangerous without oversight, especially at scale. While keeping a human in the loop, AI gives that person superpowers, and as we prove out AI safety, we can drive towards appropriate, scenario-specific automation.
At Paytronix, we remain committed to providing world-class solutions for our clients, using the best available technology to drive success, and I am personally excited to see the continued advancement and potential in AI. In some ways, we’re just scratching the surface. I can’t wait to continue the journey with our team and our amazing clients.
Tim Ridgely leads development of Paytronix Online Ordering. In that role he drives the online ordering technology and experience, working hands on with customers to help maximize first-party digital sales, leverage data and AI, and to drive new revenue sources.
