Essay AI & Automation
The Last League Anyone Will Build by Hand
AI won't fix sports software by adding a chat box. It will fix it by doing the work operators were never supposed to do.
Nobody starts a sports league because they love configuring software.
They start because their town needs a place to play. Because their daughter's team needs better competition. Because they believe they can build something people will want to come back to.
Then they open their laptop.
Create the season. Add the divisions. Set the age groups. Enter the fees. Build the registration form. Attach the waiver. Set up payments. Figure out the schedule.
Somewhere in that process, the person who wanted to build a sports community becomes a software administrator.
We have accepted this for a long time. We shouldn't have.
A chat box is not the answer
In my last essay, I wrote that complexity is a failure of design. When an operator has to carry the context, remember the exceptions, and tell the system what to do at every step, the software is leaving too much work unfinished.
AI gives us an opportunity to finish more of that work.
But most of the industry is aiming too low. The easy move is to put a chat box next to the same old forms and call it AI. The operator still does the translating. They just type it into a different box.
That isn't the shift. The shift is software that takes on the work itself.
What has changed is that software can now understand intent expressed in plain language. And a platform that already holds the registrations, payments, schedules, and rules has the context to act on that intent, not just record it.
Every setting is a translation problem
For years, sports software has followed a familiar pattern. An operator needs something. We add a setting. Another operator has a slightly different need. We add another setting.
Each addition makes sense on its own.
Eventually, building a league means understanding how dozens of settings interact. The operator knows what they want to run. Now they have to learn how the software wants it described.
That translation is work.
A registration fee might affect payment options. An age group might affect eligibility. A venue limitation might change the entire schedule. The operator becomes the person connecting all of it.
More features can help. They can also give that person more things to connect.
Describe the league, review the result
The shift I care about is what happens when software can take on that translation.
An operator should be able to say:
"I'm running a six-week youth basketball league. Three age groups. Games on Saturdays. Registration is $150, and families need an option to split the payment."
That should be enough to get the work started.
The system should use the information it has, ask for what is missing, and produce something the operator can review.
There will still be decisions. There will still be exceptions. But the operator should spend their time on those decisions instead of entering the same information in five places.
Where we are today
That is the direction we are building toward at Natty Hatty.
We are building Arya into the checkout builder so a customer can describe the checkout they need in a conversation and have the system build it.
The scope matters here. A checkout is a specific piece of work. Creating and running an entire league is a much bigger ambition.
The checkout work is a step toward that ambition. It is not evidence that we have already arrived.
Trust is earned in the details
Even within a checkout, the standard has to be higher than generating something that looks right. Fees need to be correct. Payment terms need to match what the operator intended. Missing information needs to trigger a question, not a confident guess.
An agent earns trust by getting the details right and making its work easy to check.
That is how this expands: useful work, a clear review, and more responsibility as the system proves itself.
The longer-term goal is for customers to create and run leagues by talking with Arya. To describe what they want to accomplish, work through the decisions, and let the system handle more of the setup and follow-through.
An agent earns trust by getting the details right and making its work easy to check.
What stays human
The human responsibility does not disappear.
Someone still has to decide whether a league is affordable for its community. Whether an exception is fair. Whether a parent needs a conversation. Whether a competitive division is actually serving the kids in it.
Those decisions require judgment and relationships. Automating the administration should give operators more room for both.
A prediction
By the end of 2028, describing a league and reviewing a system-built setup will be a normal expectation in sports software.
Operators will still make the important calls. But manually configuring every division, fee, form, and rule will increasingly feel like doing work the system should already know how to do.
The title of this essay is a prediction, too. Somewhere, an operator is building their last league by hand.
They probably don't know it yet.
They're just trying to get registration open before dinner. We should build software that lets them.
Share this essay