Structuring AI Use Cases in Cultural Institutions Through Reproducibility

Introduction

Artificial Intelligence is entering cultural institutions at a rapid pace. Museums, galleries, and heritage organisations are experimenting with everything from automated cataloguing to chat-based guides, from predictive analytics to AI-generated exhibition content. Yet as enthusiasm rises, so does a fundamental strategic question:

Where does AI genuinely create long-term value in cultural institutions—and where does it merely produce short-lived novelty?

After working with multiple cultural institutions and observing AI-driven projects across Europe, we’ve noticed a recurring pattern:

the success or failure of an AI initiative depends less on the technology itself and more on where in the institutional process it is applied.

To navigate this landscape, we propose a simple but powerful lens: reproducibility.

Reproducibility refers to whether an AI improvement can be applied repeatedly, across exhibitions, across seasons, and ideally across institutions. When AI addresses stable, recurrent processes, value compounds over years. When it addresses temporary exhibition elements, the value often evaporates once the exhibition closes.

This article outlines a practical three-layer framework for structuring AI use cases in cultural institutions, helping decision-makers prioritise solutions that deliver lasting, scalable benefits.

Why Reproducibility Matters

Cultural institutions operate under tight budgets, limited staff capacity, and high expectations for public value. This makes it essential to invest in AI solutions that:

• remain useful over many years
• scale across exhibitions and projects
• reduce workload consistently
• improve visitor experience beyond a single context
• become part of the institution’s core infrastructure
  

AI solutions that fail these criteria tend to fall into the category of “one-off experiments”—interesting, engaging, even groundbreaking, but seldom reused.

Reproducibility provides institutions with a method to distinguish between strategic AI and theatrical AI.

The Three-Layer Framework

To analyse reproducibility, we divide cultural institution processes into three layers:

1. Back-Office Processes
2. Core Visitor-Facing Museum Operations
3. Exhibition-Specific Elements
   

Each layer responds differently to AI, and each offers different levels of long-term impact.

1. Back-Office Processes (Operational & Administrative)

High stability, high reproducibility, high institutional impact

These are the internal processes that ensure the institution operates smoothly. They do not change significantly from one exhibition to another, making them ideal candidates for repeatable AI enhancement.

Key examples include:

• Collection management – automated tagging, metadata enrichment, deduplication, digitisation support
• Archival operations – document classification, multilingual transcription, OCR
• Exhibition planning – scheduling algorithms, resource allocation, spatial optimisation
• Financial operations – forecasting visitor numbers, modelling shop revenue
• HR & staffing – shift planning, competency mapping, training guidance
• Internal communications – AI-assisted knowledge search, summarisation, or workflow triggers
  

Why AI fits here:

• Processes are consistent over years
• Large amounts of structured or semi-structured data exist
• Improvements have cross-departmental effects
• Increased efficiency frees staff for higher-value work
• Solutions can be replicated across institutions with minor adjustments
  

Example impact:

If an AI system reduces cataloguing time by 40%, that benefit repeats every week, every month, every year. Over time, such improvements reshape how an institution operates.

This layer represents the highest return on AI investment.

2. Core Visitor-Facing Museum Operations

Moderate stability, high relevance, strong reproducibility

These are the services visitors interact with across all exhibitions. Although the content may change, the underlying systems remain the same.

Key examples include:

• Visitor information systems – multilingual text, labels, signage
• Guidance systems – audio guides, mobile apps, AR overlays
• Ticketing & access control – recommendation systems, personalised itineraries
• Digital content – video guides, interactive stations, chat-based interpretation
• Accessibility tools – text-to-speech, sign-language avatars, simplified explanations
• Navigation solutions – indoor wayfinding, crowd-flow optimisation
  
• Shops and membership platforms – personalised recommendations, smart merchandising
  

Why AI fits here:

• These services persist regardless of exhibition theme
• AI improvements directly enhance visitor experience
• Tools can be reused and scaled institution-wide
• Many solutions can be fed with continuously improving data
• Standardisation across the sector is possible
  

Example impact:

An AI-powered audio guide system that adapts content based on visitor preferences can be used for every exhibition—only the underlying content changes, not the system itself.

This layer is crucial because it directly affects public perception and the inclusivity of cultural experiences.

3. Exhibition-Specific Elements (Temporary Installations)

Low stability, low reproducibility, short-term cultural value

These elements are tied to one exhibition and usually vanish afterwards. While they often generate the most attention and media coverage, they deliver the least long-term infrastructural value.

Key examples include:

• Unique, exhibition-specific AI installations
• Thematic interactive displays
• Temporary AI-generated artworks
• Special AR/VR scenes created for one exhibition
• Sensor-based immersive rooms
• Chatbots tied to a single exhibition narrative
  
  

Why AI fits here—but with limits:

• High creative potential
• High visitor engagement
• Excellent for experimentation and public imagination
• But: limited lifespan
• High maintenance costs for short-term deployments
  

Example impact:

An interactive AI installation that reacts to visitor emotions might delight audiences for three months — and then becomes irrelevant once the exhibition closes.

This layer is valuable for innovation and cultural expression, but not for building institutional capacity.

Comparing the Three Layers

The takeaway:
Layers 1 and 2 are where long-term transformation happens.
Layer 3 is where creativity happens.

A balanced AI strategy should address both—but not confuse them.

How Institutions Can Use This Framework

1. Prioritise AI investments based on stability

Before commissioning AI, ask:
“Will this still be useful in two years?”

If not, consider downsizing the investment.

2. Require reproducibility from vendors

Ask technology partners:

• Can this system adapt to new exhibitions?
• Can we reuse it with minimal cost?
• Is the underlying technology content-agnostic?
  

3. Build internal capacity for Layers 1 and 2

Develop staff skills in:

• data literacy
• AI-assisted workflows
• evaluation and moderation of AI outputs
• digital responsibility and ethics
  

4. Treat Layer 3 as a creative sandbox

Instead of seeing temporary installations as infrastructure, view them as:

• experimental pilots
• opportunities to test new ideas
• environments for audience research

Successful concepts from Layer 3 can eventually be migrated into Layer 2.

Conclusion

AI offers enormous potential for cultural institutions—but only when applied with strategic clarity. By distinguishing between back-office systems, core visitor services, and exhibition-specific experiences, institutions can make informed decisions that favour long-term impact over short-term novelty.

Understanding where AI can create reproducible value ensures that investments strengthen cultural infrastructure, support staff, and expand public access to culture—not just for a single exhibition, but for years to come.

‍