Will event organizations begin to build their own registration and ticket websites using AI tools – AI DIY? With the release of Claude Code and continued maturation of the Vibe Coding tools, this is now a viable question to ask. Even public SaaS companies like Salesforce have dropped an average of 30% over the past year because of fears of customers moving to AI DIY.
TLDR: Very limited today, but certainly in the next 5 years event organizations will be able to create good websites and signup mechanisms for themselves with AI. Many events will not have the time or want the liability with this approach. Also, vendors will still have some advantages. My guess is that as many as 20% of events will build their own with AI.
To study this more, I built a little website and registration test for the Scott Coffee Run that I help with. I also write up some additional in depth thoughts below. Here are the sections:
- AI DIY Website and Registration Site today
- AI will get better
- AI Shortfalls
- Traditional Systems Plus AI Chat
- What Does This Mean for Registration and Ticketing Vendors
My AI DIY Race Website and Registration App – 1 Hour, 10 Minutes
Initial Application and Prompt:
This was the first attempt by my v0.dev vibe coding tool from the following prompts. Really I only used one prompt – “I want to build a race registration website for a race” and answering a few questions like name, distance and price method and pricing. AI was smart enough to ask me those questions.
AI was smart enough to:
- Build a website with nice design elements like the pricing tiles.
- Help me signup for a database (Supabase) and payments (Stripe sandbox).
- Come up with wording for what was included and the highlights of the race.
- Signup form which included the common elements in race registration – date of birth, gender, shirt size, waiver checkbox.
I did a total of 37 versions of this application in the 1 Hour and 10 Minutes I spent building it (I was originally shooting for 1 hour, but ran long because of a combination of me adding features and the mistakes it made). About 10 of those prompts were functionality and 27 were asking it to correct a mistake. During that time it did database migrations to fix things and add fields that I wanted. I also had to do a workaround where the checkout page had to be bright up in another tab of my browser.
Most important, within that 1 hour and 10 minutes I was able to process transactions against my test Stripe sandbox account using a dummy card number and see the money in my Stripe Sandbox.
The final result added:
- A nice banner image rather than the red background.
- A Report page that was password protected (just a single password, no login).
- Details button on the report page.
AI Is Getting Better Fast
As I will write about below, this is a long way from something that a race would really use. But there is no way AI could have created this much with so little prompting a year ago. AI understood coding things like how to create HTML/CSS, it also knew how to connect to multiple systems (database and payments), how to create a database and add functionality. AI ALSO understood what a race was and the types of information that are important in a race.
And that was only 1 year since the first time I vibe coded. Imagine what 5 more years will bring.
AI Shortfalls
I think the biggest shortfall can be explained by the discussion Stephen and I had back when we started RunSignup – “We will be done with building features in 2 years”. Obviously, we are a long way from being done after 16 years. This is a very thin application. No coupons, age based pricing, teams, dashboard for flexible reporting, no way to reconcile transactions on Stripe with transactions in the application, etc. And certainly no advanced features like bib assignment, race day checkin and scoring, realtime tracking, etc.
Since AI will improve greatly over the next 5 years, some of this will be able to be addressed. It might approach the collective wisdom of what races need that RunSignup employees have built in over 500 years of service at the company. It might approach the ability to write all of these features. However, it is still likely to have difficulty with putting all of this together, especially in a multi-race environment.
The strategic thing I see missing is the human – AI interface. Chat is nice as the first round, with prompts based development. But it gets unwieldy at large scale. If I wanted to add all of the pages, menus and content on our RunSignup hosted website for the Scott Coffee Run, a chat interface is nearly impossible to manage. Even if AI is able to track everything, humans will pile bad prompts on top of bad prompts. AI will write bad code on top of bad code and produce an unstable website.
The Future – Traditional Systems PLUS Chat with AI
We are betting that this is the future. When we declared 2025 the Year of AI, we came up with a framework to think about it. I am more convinced than ever this is the right future. A central transactional database with both traditional User Interface on desktop and mobile of buttons, checkboxes, text fields, etc. AND an AI driven Chat interface. With both traditional and AI working on the same data and infrastructure.
Our development plan for AI in 2026 calls for exactly that. We are building out our own AI Agent infrastructure (on Amazon AWS Bedrock Agent Core) that will enable our whole development team to begin adding chat interfaces to every dashboard page as well as on the user front end. So event directors and their customers will have the choice of both traditional user interfaces and chat.
While I was building the little test app I was thinking about how complicated it would be to create a dashboard or to create a multi-page website via Chat. I was thinking about how our Website V2 is so much simpler – to create new pages by pointing and clicking and changing the menu order by dragging and dropping as an example.
This vision of the future was confirmed to me when I saw V0 has added a design mode that is a traditional UI that changes specific numbers that are in the code. Similar to how our Website Builder can change attributes in a traditional UI (of course the AI is only doing this directly in code for a single website, we do ours in a database since we support tens of thousands of websites).
And it was not only the User Interface. The AI system is connecting to a traditional database and a traditional credit card processing system.
It occurs to me that AI will start to add more and more traditional methods, while RunSignup is adding more and more AI interfaces to our transactional system.
The other “big thought” that occurs to me (I am always an optimist!) is that since our own development team is using AI to write our code, event directors will have a tough time competing with them to write code with AI. Again, this indicates that everyone winds up heading to the same spot using AI, but RunSignup has several advantages – experience, a highly functional system, lots of features, and a well tuned business operation that makes us very efficient.
What AI DIY Means for Registration and Ticketing Vendors
There are obviously hurdles to events using AI to build their own. However, with AI getting better, and more people wanting to see what it can do it seems obvious that some percentage of events will move this way. Certainly that is what happened when the Internet came around and early adopters did online registration and ticketing. Many of the vendors today came from those early efforts.
Vendors have two things going for them:
- Momentum. To build it yourself takes time and increases risk that will give most event directors pause before building their own.
- Advanced Features. Many events have a need for advanced features that would take far too long to build even with AI.
My job is to focus on RunSignup, and I see us having a number of strategic reasons customers will still choose us over building their own. However, we need to keep pressing forward with our use of AI and offering AI within our products. We have done well so far, and knowing our internal progress in our development team gives me a lot of hope. Especially when I compare what we are doing with the very few events who are actually doing any real building with AI. I am realistic and do expect increased churn of accounts over the next 5 years. Particularly on simple events and especially in simple ticket events. We have seen this for years with people using Google Forms or FromStack connected with Stripe or PayPal (we actually see a number of customers getting tired of DIY). My expectation is that as much as 10-20% of our business moves to AI DIY, so we will need to continue to grow 🙂
When I think about other vendors, especially in the US registration market, there will be significant pressure on their businesses. We are already seeing a couple of registration vendors a year go out of business because they lack the critical mass to compete on a functionality and price basis. And with AI DIY as an additional competitor, things could get even tougher. The same is true for the very large number of small ticket vendors. And it will be compounded with the fact that ticket sales are a lot simpler than race registration (which makes me happy we have spent the past 3 years developing our patent pending calendar based ticketing).
Summary
I am really glad I spent time to do this. I intend to do samples every 6 or 12 months to see the state of the art in AI DIY and to track progress. If anyone does build something, let me know. I would love to hear about your experiences and keep learning how AI DIY will impact events.