During my years singing chorus at Opera Australia, I spent more time than I’d like to admit standing in the wings listening to front-of-house staff patiently explain, for the hundredth time, that no, supertitles are not available in Italian for an Italian opera. Yes, there’s an interval. No, you can’t bring your glass of champagne into the auditorium.

Same questions. Every performance. Every season.

So when I heard that several Australian performing arts companies were experimenting with AI agents to handle routine audience enquiries, my first thought was: about bloody time.

The Always-On Box Office

State Theatre Company of South Australia was one of the early adopters. Their marketing director, Emma, told me they were drowning in messages across seven different platforms. Email, Facebook Messenger, Instagram DMs, WhatsApp (international tourists), SMS, phone calls, and their website chat.

“We’d have a single person asking the same question across three channels because they didn’t know we’d already responded via email,” she explained. “Meanwhile, actual booking enquiries were getting lost in the noise.”

They implemented an AI agent system using OpenClaw, an open-source platform that’s become surprisingly popular in the tech world (192,000+ GitHub stars). The platform connects AI agents across multiple messaging channels, so when someone asks about wheelchair access on Facebook, the system knows not to send them the same answer via email ten minutes later.

The results were immediate. Response time for basic queries dropped from an average of four hours to under two minutes. But more importantly, their small box office team could focus on complex bookings, group sales, and the kind of personalised service that actually builds subscriber loyalty.

The Skills Marketplace Problem

Here’s where it gets interesting, and slightly concerning. OpenClaw works through a marketplace of nearly 4,000 pre-built “skills” that handle specific tasks. Book an appointment, send a reminder, process a refund, that sort of thing.

Sounds efficient. Except recent security research found that 36.82% of these skills have security vulnerabilities, and 341 were confirmed malicious. For arts companies handling subscriber data and credit card information, that’s not exactly reassuring.

Melbourne Theatre Company learned this the hard way when they discovered their OpenClaw instance was publicly exposed due to a misconfigured skill they’d installed for email triage. No data was breached, thankfully, but it was close enough to prompt a complete security review.

Several companies I spoke with have since moved to managed AI agent platforms where someone else handles the security audits, hosting, and compliance requirements. It’s more expensive than the free open-source version, but arts funding barely covers the basics as it is—nobody needs a data breach eating what’s left.

Beyond the Transactional

What surprised me most wasn’t the box office automation. Every industry’s doing that. What caught my attention was how some companies are using AI agents for actual audience engagement.

Sydney Chamber Opera has an AI agent that monitors their subscriber Discord server. Yes, opera companies have Discord servers now. Yes, I’m as shocked as you are.

The agent answers questions about upcoming productions, shares relevant articles, and facilitates discussions about opera history. When someone asks a question beyond its scope—say, a detailed question about baroque vocal technique—it immediately tags the appropriate staff member.

“It’s like having a really enthusiastic volunteer who never sleeps and has perfect recall of our entire production archive,” their digital engagement manager told me. The platform can access their historical program notes, reviews, and production photos going back decades. Ask it about their 2019 production of Orpheus, and it’ll share details, images, and even connect you with other subscribers who attended.

According to research from Stanford’s Institute for Human-Centered AI, this kind of scalable personalisation is becoming crucial for arts organisations competing for audience attention in an oversaturated cultural landscape.

The Human Touch Still Matters

I asked every company the same question: what do humans still do better?

The answers were consistent. Anything involving artistic judgment, pastoral care, or nuanced negotiation stays firmly in human hands.

When someone’s distressed about a cancelled performance, a real person responds. When a long-time subscriber asks for special seating considerations due to a new health condition, a human handles it. When an artist liaison needs coordinating, no AI’s touching that.

The AI consultants in Sydney I’ve spoken with say this is actually the sign of a well-designed implementation. The technology handles the repetitive work so humans can focus on the relationship-building that actually matters in the arts.

What This Means for Regional Companies

The most promising application might be for smaller, regional companies with tiny budgets and even tinier staff.

Canberra Youth Theatre told me they’re basically a two-person operation running a full season plus youth programs. Before implementing an AI agent for routine communications, they were working 60-hour weeks just keeping up with enquiries.

“We’re artists and educators, not customer service specialists,” their artistic director said. “But if we don’t respond quickly, families book their kids into something else.”

Now their AI agent handles class enquiries, registration questions, and performance information across WhatsApp, email, and Facebook. The two actual humans can focus on, you know, making theatre.

The cost is about what they’d spend on coffee for opening night, but the time saved is measured in hundreds of hours per season.

Still in Beta

Look, this technology isn’t perfect. I’ve seen AI agents confidently give wrong information about running times. I’ve watched them struggle with nuanced questions about content warnings. And I absolutely witnessed one try to sell a subscriber a ticket to a production that closed in 2023.

But neither are human box office staff perfect. We’ve all been given incorrect information by a well-meaning volunteer who’s been trained for exactly fifteen minutes.

The difference is that AI agents get better with feedback, they don’t forget their training, and they don’t call in sick during tech week.

For an industry built on transformation and storytelling, it’s interesting to watch performing arts companies write their own narrative about what role technology should play. The ones doing it well aren’t replacing the human experience—they’re protecting it by automating the noise away.

That feels rather operatic in itself, doesn’t it? All that technology in service of something fundamentally, irreducibly human.