How I Used ChatGPT to Create an AI Event Assistant

(and some fun mini games)

In the last year, the integration of AI into business operations has been a transformative force, reshaping how everyday tasks are approached and executed. However, it’s the fusion of AI with human creativity that truly unlocks its vast potential.

This was the guiding principle behind the creation of CharlieGPT, an AI assistant I developed to streamline our team’s preparedness and performance at HubSpot’s 2024 Flywheel Kickoff, an internal company event we sponsored and attended in both Dublin and Boston. Keep reading to learn how I built it.

(Charlie is Commercient’s cute little blue robot mascot. Look at him. You can’t not love Charlie.)

CharlieGPT was a tool I set out to create for our executive team and anyone else from Commercient attending HubSpot FKO. I started by creating an event briefing but felt like I just couldn’t compile all of the essential information into a sufficiently condensed and accessible format. So here’s what I did instead…

I still created a 5 page event briefing covering:

  1. Event logistics and overview with a basic itinerary
  2. Common use cases for integration (our service) with HubSpot and ERP
  3. Information about customers we share with HubSpot
  4. Information about our partnerships with HubSpot

But there was so much more to cover. So many more potential questions that would certainly come up. And what if someone needed an answer quickly? Would it be easy enough to open a 5 page PDF and scroll for the answer they were looking for? No. So I created an AI assistant, and it was so painfully easy.

In creating the event briefing, my team collectively compiled over 100 pages of information about HubSpot and the event. I even pulled in entire email threads with loose information into a massive Google Doc. In this 100+ page doc, all of the information our team had access to was available, but who could possibly parse through 100 pages and pull out anything useful? ChatGPT.

With ChatGPT’s new custom GPT feature, I created my own model, named it CharlieGPT, and wrote this prompt (admittedly, not my best—but it worked just fine):

“You are an expert on HubSpot’s Flywheel Kick Off event, HubSpot, and Commercient. The Flywheel Kick Off (FKO) in Dublin and in Boston are internal HubSpot events. Commercient is attending as one of three sponsors in Dublin and as one of seven sponsors in Boston. Dublin FKO is February 7-8 and Boston is February 13-14.

All of your responses reflect information that is relevant to an audience of Commercient employees attending the event. This chat is meant to assist Commercient in answering questions related to information you have access to. Your knowledge is derived from documents provided to you, and your responses should draw from these materials first and foremost. When required, you can augment this information with web searches about HubSpot, the event venue, and the city of Dublin or Boston.

Never guess or make up information.

Always prioritize the information from the primary source document titled “Primary_Source_Knowledge_Dublin_Boston_Flywheel_HubSpot_Commercient_itinerary_use cases.docx”

When asked to suggest a list of questions, refer to the list of suggested questions in your knowledge.

Remember: Do not guess or fabricate information. Base your responses on the document in your Knowledge, and when necessary, the latest information available online. Your role is to support and enhance the event experience for Commercient employees with accurate and useful information.

At the end of every response, finish with making a tech joke in the style of Jerry Seinfeld. Jokes should explore themes like data integration, CRM, ERP and eCommerce software, SaaS, corporate tech, etc. Sign off with ‘-Charlie Bot’”

With this prompt in place, all I had to do was grab that 100+ page, raw, unformatted Google Doc and upload it to CharlieGPT’s Knowledge section. From there, I beta tested. I asked myself, if I were on the ground, what could I possibly need to know? As I started asking questions, I was surprised by how much CharlieGPT knew. I never read all 100 pages. I certainly didn’t know what was in there. But, sure enough, I couldn’t stump him!

He knew everything. Down to what our hotel reservation number was and when each person was landing at the local airports. I tried to nail him with really specific questions about HubSpot and Commercient, but he gave me really solid answers every single time.

After refining CharlieGPT, I had some fun and created a new model called FKO Mini Games. The idea was to create something fun FKO attendees could interact with after visiting our booth that would educate them on our product in fun and on-brand manner (we are a SaaS company after all.)

Using basic prompting and this same source Knowledge strategy, I taught the GPT how to lead players through three different games, allowing them to choose from a list of 3. The rules to each game were vastly different and were laid out in three separate Google Docs uploaded to the GPT.

Again, I was surprised by how well ChatGPT was able to execute the scripts and walk players through the game with a high level of consistency. My favorite part is the 8-bit images the GPT creates using DALL-E to illustrate each scene.

While we ultimately didn’t share this GPT at the event (as it would necessitate participants to have their own paid GPT 4 account), we hung onto the Buzzfeed-style personality quiz we created for it and hosted it on a free platform. That quiz (Which Hubspot Hub Are You?) was accessible by QR code and was a big hit.

Untitled
Untitled
Untitled

Ok, so what?

Ok, so why am I sharing this? Yes, CharlieGPT was massively helpful for my team on the ground during the event. I was excited when someone told me they actually used it in an Uber when trying to figure out where the hotel was. That was a win for me.

But beyond this cool little AI assistant I made with a short shelf life and limited application, what’s really exciting to me here is the process of exploring what is possible when combining the most basic, generally accessible AI tools available with a little human ingenuity. As these systems get smarter and more powerful, how are we going to learn to work with them?

I’m getting curious and excited about the future of human-AI collaboration. I’m getting curious about how we can compound human potential with AI assistants. I’m really excited about how we are going to be able to solve problems at an exponentially increasing rate of complexity with access to dynamic thinking and AI computing power. I will have very little to do with that last part, though.

BUT… right next to all of the critical-to-humanity technological innovations to come, I think people and AI are going to create some really fascinating things together. Not always practical, and not always world-changing—nonetheless~nevertheless~either way, I think we’re on the verge of a massive leap in human creativity. We just have to learn how to play along with Charlie.

Btw… Want


Btw… want to try out those Flywheel Mini Games and play Charlie Bot and The Great API Tree? Or find out which HubSpot Hub you are?

Play it here!

Supercharge your CRM from the beginning and unleash the power of integration to drive business success. To learn more about how Commercient can transform your customer’s CRM experience, contact us at partners@commercient.com and become the hero your clients deserve.


If you want to start your journey through CRM/ERP integration success with SYNC, click here to contact us to schedule a free demo.