Ticketing and AI Part II

Last December, I said that I didn't have a use for LLM based generative AI in ticketing. 

Well, I found one. 

At the Ticketing Professionals Conference 2024, in the wake of his keynote presentation that I didn't actually go and see, I was chatting with Rob Williams, late of AudienceView, about my scepticism and his enthusiasm for our coming AI overlords. And one of the things that he said really stuck: that LLMs "work" by (in some way) translating an English sentence into a set of statistical weights in some internal private language that represents the meaning of the sentence, and then it can do stuff with this, and then translate the answer back into English... but it can translate the answer back into French or German or computer code equally well. 

And this made me realise that if I gave an LLM my database schema - but none of the actual data - then it could translate questions into SQL, which I could then run on my database, and get useful, accurate answers. The answers would be SQL - auditable and checkable and repeatable, not just some numbers that it had made up and had no way of showing the working. If the SQL was valid, it would run on the database, if it was invalid, it wouldn't. There would be much less danger of the LLM "hallucinating" an answer and making stuff up and getting it wrong. And there would be no danger of private data being sent to an AI and being used as training data and leaking out in some way.

So that's what I built:

Survivorship Bias

Standing the foyer of The Maltings waiting to pick Polly up from her drama group, I just saw someone ask whether there were any children's films on today, and then walk away disappointed.

My first thought was "the box office staff should make a note of that, to inform programming decisions".

My second more useful thought was "hey, couldn't the ticketing system record searches that DON'T return any results, to tell us what people want that we aren't offering".

 

 

 

We record tonnes of data about what we DO sell and who we DO sell to. We ought to spend even more effort interrogating what we don't sell because we haven't made it available.

Ticketing and AI

Fundamentally, after the robots have taken all the jobs, there will be nothing left for humans to do but create and consume culture. Live people will still buy tickets to see live people perform on stage, so this is a good business to be in. 

It doesn't matter how good the machines get at predicting the next word in a sentence or the next note in a tune, there will still be an audience for live performances. 

So if everyone can just calm down a bit...

--------------------

The first potential use we investigated for AI in ticketing was integration with personal voice assistants - Sira, Cortana, Alexa, etc. We imagined saying "Hey Siri, book me two tickets for the panto next week". 

I was quite excited about this in 2019:

 

We eventually decided this was not worth, because the information about which performances are available and which seats are available is not easily conveyed by voice, and a web page is actually more convenient. From a customer's point of view, phoning the box office to book tickets is less good than being able to see all the available performances and the seating plan on the website. Automated phone booking lines for cinema tickets are a thing I remember using and hating; having better speech recognition and natural language processing wouldn't help much there. Also, you would still need box office staff answering the phone to deal with exchanges and customer services and refunds, so we couldn't envisage either extra tickets being sold or money being saved by the venue as a result of this facility being available. And that at the end of the day means there's no reason to do it.
 

Another area we quickly dismissed was using LLMs to automate the text generation part of event setup. We don't think this is a good idea at all. Ticketing requires accuracy, and the risk that an LLM would hallucinate blurb containing incorrect prices or dates is unacceptable.

Dynamic pricing and predictive analysis tools are a good fit - but then they already exist, already use AI, at least they already use machine learning to crunch vast piles of data. And whilst we are happy to integrate with them on request, we have no plans to develop our own.