How to evaluate an AI solution for your institution: the questions to ask

01First question: is the solution specialised for the cultural sector?

The most decisive question. Generic voice agents do not understand cultural specifics: guided tours, temporary versus permanent exhibitions, collector calls as commercial opportunities. A specialised solution has dedicated tools: visit booking with confirmation numbers, guided tour scheduling in multiple languages, exhibition information, tiered pricing, accessibility, events and memberships. Ask directly: what cultural-sector-specific tools are integrated?

02Second question: how does the agent learn my institution's information?

The best solutions offer a knowledge base where you import documents by drag and drop. The agent integrates them within minutes. Less mature solutions require manual form entry or rigid response scripts that quickly become obsolete. Ask: can I import my existing documents as-is, and how quickly does the agent integrate them?

03Third question: how many languages are available?

The most advanced solutions offer fifteen or more languages with mid-conversation switching. Ask: how many languages, can the agent switch during a call? Test voice quality in several languages.

04Fourth question: what happens after each call?

The best solutions generate automatic summaries, satisfaction scores, action recommendations and structured visitor cards. Ask: what data is generated after each call?

05Fifth question: how does pricing work?

Three models exist: fixed subscription, pay-per-minute and hybrid. Ask: what is the exact cost per minute, and are there hidden fees? Request a simulation based on your actual call volume.

06Sixth question: is my data protected?

Verify EU hosting, institution data isolation, no third-party model training and preparation for the August 2026 European AI regulation. Ask: where is my data hosted, and can you provide compliance documentation?

07Seventh question: what happens when the agent cannot answer?

The best solutions handle this gracefully: acknowledge limits, inform the visitor, transfer to your team. Ask: how does the agent react, and how does transfer work? Test this scenario.

08Eighth question: can I test before committing?

Any serious solution must be testable. Ask: do you offer a trial or demonstration?

09Ninth question: what support is available?

The first days after deployment are critical. Ask: what is the support response time, and is support included?

10Tenth question: does the solution preserve my identity?

The agent must present itself in your institution's name, not the provider's. This is white-label operation. Ask: does the agent introduce itself with my institution's name?

11Making an informed choice

By asking these ten questions and comparing answers in a structured way, you will have all elements for an informed decision based on objective criteria rather than marketing promises.

AI ARTEDUSA answers each of these questions. Find out how at ai.artedusa.com.