LLMs, content design and more human-sounding conversations
As AI shifts from fun toy to useful tool, agencies are often at the forefront of proposing new usages. Starting in 2022, I worked on several AI projects that ranged from conversation design to a straight-forward “chatbot” to successfully building a meta AI machine.
More on that below, but as I was able to jump in feet-first and build AI-enabled products, I spent time reflecting on my experience, how copywriting can evolve and the path forward as we build AI tools that are beyond search engines and simple assistant tools.
Check out these articles:
AI work
The enthusiastic AI trainer
One of my clients (a chain of fitness centers) tasked us with building an AI trainer that was more than a chatbot. It had to sound like the brand, like a trainer, create fitness plans and workouts and adjust on the fly. It needed to be a trainer.
I worked with the team consisting of design, strategy, engineering and program to create an MVP (minimum viable product) of the AI trainer. I developed a voice and tone for the LLM, wrote out tens of “happy path” conversations and golden queries, collated content (videos, workout plan examples, etc), tested with the live system and produced a sizzle reel to show it all off to investors.
Along the way, we confirmed the need for not only a copywriter, but a copywriter who could speak with all the different departments. A hub to the spoke of the project, if you will. In the end, we presented a functional AI trainer example that sounded like the brand, offered enthusiastic encouragement, provided workout plans and celebrated the successes for anyone who worked with the trainer.
Cordial search
The commonality with a lot of AI projects I’ve worked on has been flexibility and never was that more the case than with the building of a white-label dynamic search AI. Originally intended for a high-end watch company, this project imbedded natural-language conversational search to help users who may not know what they wanted.
Like with any white-glove service, this AI system needed to address users in much the same way an employee at a brick and mortar location would. The system needed to be exceptionally knowledgeable, confident, not arrogant, and make the user feel like they’re receiving all the attention. I accomplish that by spending a lot of time reading ‘Unreasonable Hospitality: The Remarkable Power of Giving People More Than They Expect by Will Guidara‘ and learning from high-end hospitality.
From there I built out the content library, tagging each watch and compiling a consumable database for the AI to learn from with descriptions pre-established by the brand. As well as creating 50+ full conversation examples and golden queries.
The end result was a system that responded to uncertainty with guiding questions that learned and built a profile of the user’s wants, eventually suggesting that the user go into a brick and mortar location to make the purchase.
AI pitching AI projects
I pitched this project to the agency I was working with. It became an individual initiative to help streamline the generation of pitches for AI work.
Along with developing consumable data for the AI system (collating content like successful pitch decks, developing the journey map and building out both golden queries and “happy path” example dialogues), I built an in-depth LLM.
The goal wasn’t just to create an AI system that “talked the talk” like a traditional AI chatbot but also produce pitch decks in the room, while pitching to clients.
