ABOUT US
AI ARTEDUSA IS A TECHNOLOGY COMPANY SPECIALIZING IN AGENTIC AI FOR ART AND CULTURE.
WE BUILD AI AGENTS FOR THE INSTITUTIONS THAT WELCOME, MEDIATE, EXHIBIT AND SELL CULTURE.
Agentic AI for museums, galleries, auction houses and theaters
OUR STORY
AI ARTEDUSA was born from observation. At the front desk of an accredited museum, the same questions come back ten times a day, in five languages, while a queue builds at the ticket counter ten meters away. The evening before a vernissage, a collector calls a gallery to confirm a private viewing — but the gallerist is hanging the show. At an auction house, a consignor asks for a quick estimate while the auctioneer prepares a flagship sale. At a theater, a season-pass holder wants to swap a date for an audio-described performance — while the box office absorbs a pre-curtain rush. We founded AI ARTEDUSA to give these teams an AI agent that speaks their language: not a generic conversational layer bolted onto a website, but a vertical agent that knows the difference between a consignor and a bidder, between a vernissage and a late-night opening, between a sign-language tour and an audio guide. We are based in Menton, on the French-Italian border — at the crossroads of municipal museums, private foundations, regional auction houses and large coastal institutions, where the same workday often mixes Italian, English, French and German.
OUR VISION
The cultural sector doesn't need yet another chatbot. It needs an AI that acts — that books, qualifies, transcribes, escalates, summarizes — and speaks French, English, Spanish, Italian and German at the same level, with extension to additional languages on demand. All our R&D is calibrated for four verticals and their operational realities: for museums, absorb the multilingual front desk and qualify school visits without interrupting mediation; for galleries, qualify the collector (budget, timeline, artist focus) before the gallerist has to step away from the hang, and orchestrate price-on-request without surrendering placement strategy; for auction houses, pre-qualify the consignor (provenance, condition, estimate, urgency) and prepare KYC ahead of AMLD5 thresholds; for theaters and opera houses, smooth box-office peaks, explain audio description, sign-language tours and accessibility by default, and handle subscription swaps. Verticality is what separates an assistant that frustrates from an agent that serves.
« Verticality is what separates an assistant that frustrates from an agent that serves. »
— AI ARTEDUSA · founding manifesto
OUR TECHNOLOGY
We combine real-time speech recognition (sub-second latency), natural multilingual voice synthesis, tool-calling reasoning models, and per-tenant retrieval-augmented generation (RAG). The agent never answers from the open web: it relies strictly on your own content — collections, exhibitions, archives, artwork records, loan agreements, programming, FAQs, terms and conditions, press kits. Our stack is hosted in the European Union on ISO/IEC 27001-certified infrastructure (OVH, Roubaix). No transcript is ever reused to train third-party models. No data is resold. No cross-tenant leakage is possible (Row-Level Security at the database layer, isolated vector embeddings). The Data Processing Agreement (Article 28 of the GDPR — Regulation (EU) 2016/679) and the Data Protection Impact Assessment (DPIA) are available on request. We actively track the rollout of Regulation (EU) 2024/1689 — the AI Act — whose obligations for high-risk systems become enforceable on 2 August 2026, and we design the platform with that deadline as a design constraint, not a retrofit.
OUR VALUES
LISTENING
Understand before answering. True for a visitor call, true for the relationship we build with your institution. Our product cycles always start from an observed front desk, never from a marketing brief.
AUTHENTICITY
No AI jargon, no promises we can't keep, no screenshots of features that don't exist. We build agents that run in production, and we say it plainly.
VERTICALITY
Not a generalist agent we try to specialize through prompting. A pre-trained agent for your trade, with your trade's vocabulary baked into the system — not into a script. We refuse horizontal deployments where the agent answers everything and serves nothing.
TRANSPARENCY
Per-minute pricing, no subscription traps. You pay for what the agent handles, you know exactly how much, and you keep control of human escalation. No hidden onboarding fee beyond the scope we agree on together.
OUR COMMITMENT
AI ARTEDUSA is not a frozen product. We iterate continuously with the teams who test us at the front desk, the box office, the saleroom, the auditorium. Each vertical — museum, gallery, auction house, theater — has its own domain dictionary, growing as we learn. Over the next twelve months, we are opening native integrations with the most common ticketing and management systems in the sector (Tessitura, SecuTix, Artlogic, accesso and their European equivalents), automated PDF and DOCX archive ingestion with source citations, real-time hot-lead alerts, and additional channels — web chat, SMS, WhatsApp. What we promise: never trade verticality for generality. What we refuse: becoming yet another SaaS chatbot that answers everything and serves nothing.
Trust & compliance
Data Processing Agreement (Article 28) and Data Protection Impact Assessment (DPIA) available on request.
Strictly vertical use, transparency on agent identity from the first sentence, human escalation on demand, decision logging. High-risk obligations enforceable on 2 August 2026.
For auction houses and galleries: pre-qualification flow compatible with the obligations of the 5th Anti-Money Laundering Directive, including the structured collection of information required for KYC above the €10,000 threshold. The final decision remains human.
Knowledge base, vector embeddings and conversation summaries stored exclusively in the European Union, on ISO/IEC 27001-certified infrastructure.
No training on your transcripts. No data resale. No sharing with data brokers. No unsolicited outbound phone calls placed by the agent. Ever.
How we differ from a generic chatbot
Vertical, not generic
Pre-trained for museums, galleries, auction houses and theaters. The agent knows the difference between a consignor of record and a bidder, between a season-pass holder and a single-ticket buyer, between a sign-language tour and an audio guide, between a vernissage and a late-night opening. The domain vocabulary lives inside the system — not in a prompt re-injected on every call.
Acts, doesn’t just answer
Tool-calling LLM: books guided tours, registers bidders over the phone, swaps subscription tickets, triggers email or SMS follow-ups, emails your team a structured record after every conversation (short summary, long summary, satisfaction score, next action, reminder). Every call produces an actionable deliverable — not a raw transcript to re-read.
A human is always one sentence away
Any caller can ask for a human at any moment, and the agent escalates explicitly — never silently. No contractual or financial decision is taken without human validation. Automated decision-making with no opt-out is not what we sell: it is also what the AI Act will require from August 2026 for sensitive uses in the public and cultural sector.
— Get in touch
LET'S TALK ABOUT YOUR INSTITUTION
Audit calibrated to your activity. Demo prototyped on your documents. No commitment.
Tailored to your trade — museum, gallery, auction house, theater
Direct line to the founding team