Case Study: MAMCO
Three prototypes demonstrated in half a day, where the quote read CHF 8,000 per language
Situation
MAMCO, the museum of modern and contemporary art in Geneva, works like most cultural institutions with tight budgets and a team that wears several hats. A multilingual digital set-up, voice agent, podcast and information chatbot, had been quoted at three months of production and CHF 8,000 per language. At that price the project never left the concept note.
Objective
Test in real conditions whether prototypes built on the museum's own content, designed by MAIjin then demonstrated and operated by the team itself, can bring a multilingual digital project back within a museum's order of magnitude rather than an agency's, and give the team the appetite and the means to carry on alone.
Previous cost, per language
CHF 8,000
Previous lead time
3 months
Engagement
4 half-days
Prototypes prepared in advance
3
Process
3 July 2025: the half-day that clears the budget wall
A half-day with the museum team, built around three prototypes MAIjin prepared in advance on MAMCO's real content: a multilingual voice agent, a multilingual podcast and a tool for handling visitors' recurring questions. The session doesn't stop at a demo: the team operates all three tools itself. These are exactly the tools quoted at three months and CHF 8,000 per language. The session ends in applause.
7 July 2025: the debrief that triggers the follow-up
Four days later, the museum team gathers for a debrief. The energy from the half-day hasn't faded: everyone comes out of it delighted, and the team asks for more on its own.
December 2025: three half-days, one per department
Three sessions, one per department: communications, conservation, and exhibition design and administration. A shared foundation runs through all three: critical thinking about AI, hands-on prompt engineering, building infographics and visuals. The outcome includes a shared toolkit: ChatGPT Business with a DPA (two shared accounts), Claude Teams for spreadsheets, ElevenLabs for the podcast. On the conservation side, the decisive edge is multilingual search: finding exhibitions on a given theme, across sites worldwide, between two dates, helped by agentic browsers that summarise paywalled articles the museum already has access to. On the exhibition design and administration side, a 3D room with objects laid out becomes an immersive, navigable 3D world, and an email automation project tackles the two to three hours the director spends on email every day.
The pivot worth telling: from writer to editorial coach
In the communications session, AI fails to write from scratch in MAMCO's voice, even with a style guide built together during the session. The pivot that works: AI stops being a writer and becomes an editorial coach, offering ideas, options and structure rather than a finished text. The museum's FAQ is turned into a voice and text conversational agent. On the podcast, the team weighs two tools with opposite strengths: one spontaneous but hard to edit, the other controllable but less natural.
Results
Three prototypes demonstrated and operated in a single half-day: a multilingual voice agent, a multilingual podcast and a tool for handling recurring questions, for a project quoted at CHF 8,000 per language and three months of production
The half-day ended in applause, and the team's own request extended the engagement into three half-days by department in December 2025
The most useful outcome of the engagement isn't a flawless win: it's the honest failure of AI as a writer in the museum's voice, and its successful pivot into an editorial coach
A three-step research process that spread spontaneously among participants in the conservation session
A 3D room turned into an immersive, navigable world, a moment participants summed up in one line: “no more need for SketchUp”
The evening of the first half-day, the thank-you email carried a handwritten postscript: “It was really good, can't wait for what's next!”
Six months later, the closing email ended with a signature that sums up our training's whole approach: “written with the heart and without AI ;-)”
The proposed follow-up: a refresh in 6 to 12 months, or a 4-hour workshop on a specific case preceded by a mini-audit
Testimonials
- A huge thank you for this morning's AI training — it was clear, stimulating and accessible all at once. You made a complex subject feel alive and concrete, and it really sparked some interesting exchanges. We're leaving with plenty of ideas in mind (and a few more questions, inevitably!), but above all with the drive to explore further. Thanks again for your availability and your energy. (Our contact at MAMCO, email from 3 July 2025)
- A thousand thanks for this great recap and for your presence throughout these three sessions. Everyone really appreciated your various contributions and your teaching approach. We won't hesitate to call on your services again for anything further we might need down the line. (The MAMCO team, email from 17 December 2025)
Contact us
- Want to upskill your teams?
Let's schedule a meeting to assess your training needs
- Want to identify high-ROI use cases for generative AI?
Let's schedule a strategic call