The AI conversation in the museum sector is dominated by large institutions — the Staatliche Museen, the Kunsthalle Hamburg, the Städel. They have digital departments, IT infrastructure, and staff whose job title includes the word “digital.” They can afford to experiment.

But 80% of Germany’s 6,800+ cultural institutions have no in-house tech staff at all. A regional Heimatmuseum in Sachsen-Anhalt with two full-time staff and a volunteer curatorial team is not going to run a machine learning pipeline. The question for these institutions isn’t “how do we implement AI” — it’s “can AI do anything useful for us at all, given our actual constraints.”

The answer is yes. But the path looks very different from what the conference speakers describe.

Start With the Problem, Not the Technology

The most common mistake smaller institutions make when approaching AI is technology-first thinking: “we should do something with ChatGPT.” This leads to unfocused pilots that produce nothing usable.

The right starting question is: what takes the most time and produces the least differentiated value?

For most small museums, the honest answer is one of three things:

  • Writing object descriptions for catalogue updates, exhibition labels, or grant applications
  • Responding to standard visitor enquiries (opening times, accessibility, where to park)
  • Translating existing content for international visitors or grant applications to EU bodies

These are exactly the tasks where off-the-shelf AI tools deliver real value without requiring any technical infrastructure.

What You Can Do Without a Tech Team

Content drafting with general-purpose AI

Tools like ChatGPT, Claude, or Gemini are genuinely useful for drafting object descriptions, exhibition text, and grant narrative — with appropriate human review and editing. A curator who spends four hours writing exhibition labels can use AI to produce rough drafts in 20 minutes, then spend the remaining time on the interpretation and voice that actually matters.

This requires no setup, no vendor relationships, and no technical knowledge. The cost is a monthly subscription of €20–30.

What it doesn’t replace: curatorial judgement, institutional voice, and factual accuracy about your specific collection. Every AI-drafted text needs a curatorial review pass. But “reviewing and editing” is faster than “writing from scratch,” and that time saving accumulates.

Translation of existing content

Most small German museums have rich German-language content — exhibition catalogues, object descriptions, educational materials — that has never been translated because translation costs are prohibitive. AI translation tools (DeepL Pro, or LLM-based translation with post-editing) now make translation economically viable for institutions that previously couldn’t afford it.

A 5,000-word exhibition catalogue that would have cost €500–1,000 to translate professionally can now be machine-translated for €5 and post-edited by a bilingual volunteer or staff member in a few hours. The quality gap between professional and AI-assisted translation has narrowed enough for most museum use cases — visitor-facing content, educational materials, web copy.

Exception: legal documents, loan agreements, and official correspondence still warrant professional translation.

FAQ automation for visitor enquiries

A significant portion of museum email and phone enquiries are about logistics: opening times, ticket prices, accessibility, parking, group bookings. An AI-powered FAQ page or simple chatbot grounded in this information can handle the routine enquiries, freeing staff time for the questions that actually need human attention.

Tools like Tidio, Intercom, or even a well-structured FAQ page built with AI assistance can be deployed without technical staff. The key is scoping clearly: the tool handles logistics, humans handle everything else.

What Requires at Least Some Technical Help

Two levels of AI application genuinely require external technical support:

1. Collection cataloguing at scale. If you want to use AI to process and enrich hundreds or thousands of catalogue records — as described in my previous piece on collection cataloguing — you need someone who can build and run the pipeline. This is a one-time project cost, not ongoing infrastructure. A well-scoped engagement of 4–8 weeks can produce an enriched, machine-readable catalogue that serves the institution for years.

2. Visitor-facing chatbots grounded in collection data. A chatbot that answers questions about specific objects in your collection — rather than just logistics — requires RAG infrastructure connecting to your catalogue. Again, this is a build-once, maintain-lightly project for a small institution, not an ongoing operational commitment.

For both, the right model is a fixed-scope engagement with an external consultant or agency, not an ongoing retainer. Small institutions don’t need permanent AI infrastructure support; they need help building something that then runs with minimal maintenance.

How to Evaluate Whether You’re Ready

Before investing in any external technical help, three questions:

1. Is your catalogue digitised? If your collection records are in paper ledgers or an unsupported legacy system, that needs to be addressed before AI can help. Digitisation is a prerequisite, not a parallel track.

2. Do you have a staff member who will own this? AI tools require a human owner — someone who reviews AI outputs, maintains the tool, and escalates problems. In a two-person institution, this is probably the director or the curator. That person needs to be genuinely willing to engage with the technology, not just approving it in principle.

3. Is there a clear problem the tool solves? If the answer to “what specific task will this replace or accelerate?” is vague, the project will be vague. Start with one clear use case.

The Honest Realistic Outcome

For a small German museum with no tech staff, AI realistically delivers:

  • 30–50% reduction in time spent on routine content drafting
  • Cost-effective translation of existing materials into 2–3 additional languages
  • Reduction in low-value email and phone enquiries handled by staff
  • A more complete, machine-readable catalogue (if a cataloguing project is commissioned)

What it does not deliver: a transformation of the visitor experience overnight, autonomous systems that run without human oversight, or savings large enough to justify staff reductions in an already lean organisation.

The value is real. It’s just quieter than the conference presentations suggest.

I work with cultural institutions of all sizes across the DACH region — including those without in-house technical capacity. If you’re trying to figure out where to start, a 30-minute conversation is usually enough to identify the one or two things worth doing.