Can AI handle rehearsal schedules?
In this 60-sec interview, Ky Lloyd, Head of Creative Services, gives us her first-hand experience on using AI to work more efficiently.
by Ky Lloyd
AI is all anybody can talk about these days and for creative studios, it serves a different purpose. How did AI help you write rehearsal schedules?
As my colleagues well know, scheduling rehearsals is no walk in the park—it’s like trying to play 3D Tetris with a bunch of odd-shaped pieces. There are so many moving parts: different availability, sensitivities, and a ton of parameters to juggle. It’s a real puzzle that needs strategic thinking, constant tweaks, and a lot of teamwork to get a sensible proposal together.
I started with the standard ChatGPT-4 model. It was quick, but for a more complex rehearsal schedule, it completely failed. It looked great on paper, but lots of facts and parameters just didn’t make any sense. So, I tried the new ChatGPT-4 o1-preview model to see if it could do better.
How did the o1 compare?
As it has a more thorough approach, I prepped my prompts differently this time. I fished all my prompts from the 4.0 model conversation and put them into a comprehensive two-page briefing. The o1-preview model response demonstrated a more in-depth approach. It adjusted based on my feedback, handled complex scenarios better, and made decent use of proposing parallel spaces and backstage area use—things the 4.0 model just couldn’t work with.
Sounds like a step in the right direction! Were there any issues?
Definitely. The o1-preview model crashed after just six commands and kept crashing after this. It would be useful to have an “I give up” feedback to know concretely what the challenge was before crashing. Compared to the 4.0 model, the o1-preview was slow—I had to wait a full minute just to get a response after my two-page prompt. So, while it was more thorough, the slower speed takes getting used to and should be considered to make efficient use of work time.
What are some other pros and cons?
As it is a preview model, o1-preview can’t import any attachments or photos, so I had to write out my scene-by-scene breakdown by hand, which was time-consuming. Plus, the chat takes things very literally—it didn’t really get banter or casual language, which made it feel a bit stiff as a sparring partner.
But I will say, it was a solid tool to frame and gather thoughts and information, and just to get the ball rolling. I felt it was like having an untrained but logical assistant, sparking new ideas and helping me refine my approach. For my next trial, I’m planning to draft the schedule personally and then see if the AI can spot any new efficiencies to improve upon. It’s not perfect, but it’s definitely worth experimenting with.
Sounds like we’re ways away from having AI create our schedules but maybe not too far away. Where do you see the future of AI in the studio?
In the end, writing a good rehearsal schedule is more art than science. The o1-preview model shows promise, but there’s still a big gap between what AI can do and the nuanced decisions that come from hard-earned experience. This whole process has made me appreciate even more the complex work that goes into rehearsal scheduling. Kudos to all my colleagues who do this every day— I appreciate (even more) what we all do to pull a good rehearsal schedule together!
