As we move through this decade, the conversation has shifted. We are no longer asking if AI in museums is happening—it’s already here. The real question is how we move past the shiny new toy phase and into a period of safe AI adoption for museums that actually moves the needle on impact and efficiency.

The Opportunity: Efficiency Meets Impact

For an “outsider” like me—someone who skipped the traditional academic route to dive straight into the gears of digital engagement—the most exciting thing about artificial intelligence in cultural institutions isn’t the tech itself. It’s the capacity to democratise the past.

1. Driving AI-Driven Museum Efficiency

The back-office “drudge work” is where AI is currently winning. From automating image alt-text for accessibility to categorising massive digital archives, generative AI for museum curators is acting as a force multiplier. By offloading the mechanical tasks of data entry and initial research, we allow our human experts to do what they do best: synthesising complex histories into compelling narratives.

2. Transforming the Museum Visitor Experience

Museum visitor engagement tech has moved far beyond the static audio guide. Through AI storytelling in museums, we can now offer hyper-personalised journeys. Imagine a visitor who loves heritage railways being guided through a gallery with a narrative lens that focuses specifically on 16mm narrow-gauge engineering—not because a curator wrote ten different scripts, but because the AI understood the visitor’s intent and mapped the collection to it in real-time.

The Challenges: Navigating the “Hallucination” Minefield

We cannot talk about digital transformation in heritage without addressing the elephant in the gallery: trust. Museums are among the few remaining “bastions of truth” in a post-truth world. If an AI “hallucinates” a historical fact or reinforces a colonial bias, it doesn’t just provide a bad user experience—it damages the institution’s core brand.

• Algorithmic Bias: AI models are trained on data that often reflects historical inequities. Without careful curation, AI risks shouting over the very voices we are trying to elevate.

• Data Sovereignty: How do we protect our intellectual property when feeding data into Large Language Models (LLMs)?

• The “Human-in-the-Loop” Necessity: We must resist the urge to fully automate. AI should be a co-pilot, not the captain.

A Framework for Safe AI Adoption

How do we move forward without losing our soul? Ethical AI for cultural institutions requires a framework that prioritises transparency over speed.

Start with “Vibe Coding” and Low-Stakes Pilots

Don’t overhaul your entire digital infrastructure on day one. Start with internal-facing tools—efficiency boosters that help your team manage projects or draft social copy. This builds “AI literacy” within the organisation before you expose the technology to the public.

Prioritise Radical Transparency

If a visitor is interacting with an AI, tell them. If an image in an exhibition was enhanced or generated using AI, label it. The trust we’ve built over centuries is our most valuable asset; we shouldn’t trade it for a sleek interface.

Focus on Democratisation

The ultimate goal of my project, the Museum of AI, and my upcoming book, AI + Culture, is to show that this technology isn’t just for the “Nationals” with massive budgets. It’s for the independent museum, the local archive, and the community trust.

The Path Ahead

The future of AI in museums isn’t about replacing curators with algorithms; it’s about using technology to make our collections more accessible, our operations more sustainable, and our stories more resonant.

We have a choice: we can let the tech happen to us, or we can take the wheel and drive it toward a brighter, more inclusive future. I know which side of the “outsider” fence I’m sitting on. Let’s get to work.