RunSignup | GiveSignup is embracing the power of AI in 2025 to enhance the way we support race organizers, timers, ticket event directors, and participants. By integrating AI-driven tools, we aim to streamline customer service, provide faster and more accurate responses, and deliver smarter insights to help our customers and users succeed. As AI continues to evolve, RunSignup | GiveSignup is dedicated to leveraging these advancements to create a more seamless, efficient, and user-friendly experience for all. Last week we turned on our first AI Chatbot, “Ron Synup”. You can interact with Ron here! We are still very much so in Beta with this AI Chatbot. We will continue to learn and train the bot as time goes on.
Quick First Week Lessons
The first week has been a solid success. There have been 212 interactions with the AI Chatbot. Here are the quick results so far:
- 36% of those were entirely successful – meaning the customer did not need to interact with a human.
- 12% of questions were not answered well by the Chatbot.
- 52% were answered, but the customer still wanted human interaction.
We are encouraged by these early reports and impressed with the level of depth the AI Chatbot can provide our customers. We think these numbers will improve as we take steps to improve the Chatbot, and as people gain confidence in using AI powered Chatbots.
A Brief Background on RunSignup | GiveSignup Support
Our frontline support team (internally we refer to this team as the Info@ team) consists of 5 full time employees, Andrew Sigwart, Allison Carignan, Dave Hunt, David Drew and Elvir Ramic, who manage our info@runsignup.com, info@ticketisignup.io and info@givesignup.org email inboxes 7 days a week. Between the 5 of them they answer questions ranging from “How to reset my account password?” to “How can I setup referral tracking?” to RaceDay Scoring support questions and everything in between. The diversity in experience and knowledge on the Info@ team is what makes the team so efficient in solving our customers’ issues. Paired with our incredible Account Management team, we are able to assist all of our customers and end users with extreme efficiencies.
In 2024 the Info@ team received 26,786 tickets and sent 41,011 emails out to customers and users with an average first response time of 31:50. As we look into 2025 and beyond our volume will only continue to grow as RunSignup and TicketSignup continues to take market share. Through the first 2 months of 2025 our ticket volume has increased by 14% and email sent has increased by 21% – approximately that same as our growth.
The Release of AI Chatbot Ron Synup in Beta
Our first foray into an AI Chatbot is through Freshdesk, our support ticketing platform. We use Freshdesk to respond to customers and users, and log all of our issues. The reason we decided that our first chatbot was going to be through Freshdesk was because our extensive knowledge base articles are all housed in Freshdesk. This allows us to leverage the 300+ how-to articles to train the bot. The Info@ team also just spent the last year going through and updating all of the articles!
In addition to training the bot with all of our How To articles, we can also set up specific Q&A’s. We are constantly reviewing conversations and seeing the types of responses sent by the chatbot. If tweaks need to be made to the answers, we’ll go ahead and add specific Q&As.
We can also upload URL links, and .docx, PDF or .txt files to train the bot.
Another reason for opting for Freshdesk’s chatbot is its ability to quickly transfer chats to our Info@ queue and get the user to a real person. When setting this up, we’ve opted to give users the option to be transferred to a real person after every message the chatbot sends.
Below you can see a chat was started at 3:55 PM regarding a locked account. The user wanted to be escalated to an agent which the chatbot initiated at 3:56 PM. Four minutes later at 4:00 PM, Elvir responded to the user and was able to unlock the account.
Improving the AI Based on Early Interactions
In just under a week of having Ron live on our How-To page, we’ve had 212 chats come in.
We’re able to see which articles are being referenced the most, we’re then able to see the breakdown of the specific questions that triggered that article to be used. From this we’re able to determine if the bot referenced the correct article and if not we can make verbiage updates to our How To articles or we can make a specific Q&A.
Here we can see the question that was asked was “how do I change my default URL?”. The bot referenced Add Custom Race URLs which is not the correct answer. For race website URL’s, only admins can make updates so this would’ve required a real person to correct this.
So we’ve set up a specific Q&A to address this question. We’ll let the user know that they can email info@runsignup.com and right after the bot sends this message, we’ll also ask them if they would like to be transferred to a real person.
We’ve also accounted for scenarios where the bot doesn’t know the answer. There can be nuanced questions that we don’t have a How-To article created or a specific Q&A setup. In these situations we automatically transfer the chat to our info@ team to handle.
Detailed Results and Observations
Diving into some of the numbers we can see of the 358 questions that were asked over the course of the 212 chat conversations, 87.99% of the questions the chatbot did have an answer while 12.01% of the questions the chatbot did not have an answer to. Some the types of questions the chatbot did not have an answer for were race related questions. Such as “when will the medals be shipped?” or “what time is packet pickup?”.
When asked by the chatbot if they wanted to be transferred to an agent 37.75% of the time the user wanted to be transferred while 20.08% of the time the user answered No to being transferred. The remaining ~42% of the time the user did not select either option.
Something to note is that because we are asking the user if they want to be transferred after every message the bot sends this doesn’t mean that 37.75% of the 212 chats are being transferred. In fact we’ve had 130 chats transferred into our Info@ queue so we’re seeing 63.7% of the chats being pushed to a real person. This doesn’t necessarily mean the bot was wrong 63.7% of the time. Oftentimes we’re seeing the chatbot had the correct answer but sometimes the user still wants to talk to a real person. Here we can see the correct answer was given by the chatbot but this was still escalated to our team.
Summary
As we move forward with 2025, we will continue to improve this current chatbot. You will see more chatbots on our websites. We will continue experimenting with different options to see other AI offerings. This ultimately helps not only our productivity but also you as the race director or end user.
Some More Examples
