How AI can help tackle the cost of touring

When you live in the city, you have a plethora of different cultural experiences to choose from – from dance to opera, from commercial to experimental. You are spoilt by the diversity.

When you live outside of a city, you rely on touring organisations to bring their magic to you. Arts Council England’s Incentivising Touring Scheme has encouraged larger productions out of the city, but it hasn’t helped the smaller scale ones.

How can we help? What can we do to reduce the cost of touring, which has gone up dramatically with cost-of-living, energy and fuel costs rising.

Feasibility study

Last year, StageSwift undertook a feasibility study – funded by Innovate UK – to find out how AI could be used to increase productivity and efficiency. Specifically, we looked into how AI could be used to plan productions that move from one location to another.

The examples we studied varied from traditional touring between venues, to educational tours, community performances, workshops, hospital and prison performances to outdoor festivals and walkabouts. Tour planning is a giant ever-moving jigsaw. Not only do you need to juggle venue/location dates, you also have to schedule crew and performers.

We conducted workshops and interviews with producers of touring organisations across dance, theatre, opera and circus and found so many things to consider. We realised that a lot of information was not being used to their advantage.

For example, the information about the venue/location not included in the spec:

Where’s the nearest laundry?
Is there secure parking?
Are there road closures?
Does the loading bay have a mechanical lift?
How accessible is the backstage area?
What is the accommodation like?

The list goes on.

Developing an AI solution

All these factors matter when deciding on a venue or location. Sometimes they make such a big difference you would never choose to tour there again with that production.

And more than one interviewee commented that information is kept “in my head” or in one of the spreadsheets somewhere. The data isn’t just being used in a strategic manner.

From our feasibility study, we learned which questions producers want answers to when selecting suitable venues or locations. We took that information and began work on an AI tool that would answer those questions.

We have 10 years of touring data in StageSwift – our scheduling app – which we anonymised to use for work on an AI solution. Working with AI is not like regular software development. It’s hard to control, and it has a habit of hallucinating and we had to find ways round it.

We ensured our AI was only taking data from our database, and then only data related to the organisation it was serving. We tried RAG (retrieval augmented generation) to answer questions but it was too flaky – it’s no use informing us that we toured to the Bear Pit Theatre five times, with dates, but them omitting the two other times we were there.

So, we restructured the database and used a technique called ‘function calling’ to ensure the AI got all the data it needed – nothing added, nothing removed. As as prompt writing is very important when using generative AI and not everyone is an expert. we used system prompts to ensure results are as expected.

Of course, money is a big thing in the arts. Understanding the breakdown of finances and how they affect a show is key. We played around with creating an AI that could give a producer useful information to decide whether a location would be financially viable. We asked the AI how much we made on the 2025 tour, and how much at any specific venue.

From our interviews, we found out where producers’ time was being spent and where this could be improved. We collated our findings to show both the details and a top-level view. Our results are tabulated below.