Dr Aruna Dayanatha PhD
The case of:
How I Built a Theater Booking System in 8 Hours with ChatGPT and Notion AI
When the request came to design a booking system for a theater facility at a Pradeshiya Sabha, the obvious assumption was that such a system would take weeks to put together. Yet, with the right mix of AI collaboration, I was able to complete it in around eight hours. What made this possible was not a single tool, but the way I orchestrated multiple AI systems and my own critical judgment to bridge their blind spots.
The challenge was clear. The theater needed to manage three daily time slots, with each confirmed booking requiring three payments: a booking fee, a refundable deposit, and a balance. Each booking also came with an associated rehearsal day, meaning that the system had to track both a main event and its rehearsal without creating overlaps. A calendar view was essential to visualize everything, and conflict detection was critical so that no two parties could book the same slot. On top of this, a dashboard needed to give staff a simple view of payments due, deposits held, and upcoming events.
Instead of jumping into menus and building blindly, I began by framing these business rules in natural language. This became the foundation for how I interacted with ChatGPT. I asked ChatGPT to translate these requirements into a database structure and workflow in Notion. Within minutes I had a proposed schema: Bookings, Payments, Events, Parties, Time Slots, Settings, and Refund Rules. The logic behind each property, rollup, and formula was spelled out. This blueprint gave me a clear plan of attack before I even touched the tool.
Of course, designing on paper is not the same as implementing in Notion. This is where Notion AI came in. Notion AI is adept at quickly creating properties, views, and template scaffolds. However, it has blind spots: it cannot configure automations, and it often refuses to create formulas directly. This is where the interplay between the two AI systems became fascinating. ChatGPT provided formulas and automation logic, but sometimes its syntax was not perfectly aligned with Notion’s formula engine. When I pasted ChatGPT’s formulas into Notion, they sometimes failed. At that point, I leaned on Notion AI’s built-in Formula Helper, which refined the syntax into something that executed correctly. In a sense, I let one AI draft, the other AI correct, and myself orchestrate the process.
My thought process always followed a loop. I would ask ChatGPT to design a component — for example, a formula for Outstanding payments, or a method for detecting duplicates. I would take that into Notion, where Notion AI or the Formula Helper refined it, and I would test the outcome. If it failed, I returned to ChatGPT with the exact error or limitation. ChatGPT then gave me alternatives, such as using a Conflict Key grouping instead of a formula, or shifting to Automations. Through this loop, the system became more accurate and more complete.
The Payments database is a good example. ChatGPT explained that each payment record should pull a percentage from a Settings database, roll up the booking total, and calculate the installment amount automatically. My first attempt at the formula did not work as intended. By pasting it into Notion AI, I got back a refined version that aligned with Notion’s syntax. Once corrected, the system automatically split a booking total into 30%, 40%, and 30% installments, including the refundable deposit. From there, rollups in the Bookings database showed the amount paid, outstanding balance, and deposit held.
Events were another area of refinement. Initially, ChatGPT suggested using templates to pre-create Booking Day and Rehearsal Day. With Notion’s new tabbed template layout, that workflow did not fully match reality. The solution was to use Automations: as soon as a booking is confirmed, the system automatically creates two linked events in the Events database, each with the right type and date. This eliminated manual setup and ensured consistency.
Perhaps the most critical problem was avoiding double bookings. ChatGPT proposed formulas for a Conflict Key and grouped views. While workable, I needed something stricter. By combining ChatGPT’s logic with Notion’s new Automations, I created a rule that checked both Booking Date and Rehearsal Date against existing entries. If a conflict was detected, the system automatically flagged the booking with a Duplicate warning.
At every step, there were blind spots. ChatGPT occasionally generated incorrect formula syntax. Notion AI could not handle workflow automation or multi-step logic. The Formula Helper inside Notion sometimes needed several attempts before producing a functional result. My role was to navigate these imperfections. I became the human mediator who translated ChatGPT’s conceptual logic into workable structures, handed them to Notion AI for refinement, tested them in practice, and asked both systems for alternatives when I hit limitations.
By the end of eight hours, the system was fully operational. Bookings created payments automatically, including refund rules. Each booking generated two events. A dashboard showed upcoming activities, deposits held, and payments due. Conflict detection was embedded directly into the workflow. What seemed like a complex multi-day development task had been compressed into a single day’s work.
The real insight from this experience is not just that AI can accelerate system building. It is that humans can now orchestrate multiple AI agents, each with different strengths and weaknesses, to co-create solutions. ChatGPT was my systems designer. Notion AI was my assistant implementer. I was the conductor, ensuring harmony, spotting blind spots, and integrating their outputs into something that worked.
This is what I call orchestrating intelligence. AI is not a monolith. It is a collection of tools, each partial, each imperfect. But when you place yourself in the middle, using judgment to integrate them, you can achieve results at a speed and quality that would have been unimaginable before.