Industry
Arts & Culture
Location
London, UK
Model
Rapid R&D pilot
Scope
Large gallery area
The Challenge
Visitors to the National Gallery of course have access via mobile devices to the internet and naturally use a range of ways to access information on their own devices, including generalist AI services to answer their questions. Whilst the Gallery will provide many layers of interpretation already - from labels to audio guides and also our front of house staff - it is constantly looking at the best methods to support the visitor and in the most seamless way. But it also wants to provide the best context and make use of the wealth of curatorial knowledge available. How might we make best use of the tools available, provide the best contextual information, and create a more personalised journey for the visitor?
The Solution
The National Gallery commissioned Dwel to develop a pilot experience - a conversational companion that let visitors speak directly to ‘The Voice of the Gallery’ about the paintings, the building and their surroundings, with every response grounded in the Gallery's own curatorial knowledge.
Run as an R&D pilot encompassing a large area within the Gallery, the project was an exploration into what having a voice of their own might mean, both for them as a centuries-old cultural institution, and for the people who visit them.
“One of the key objectives we have with any interpretation approach in the Gallery is really getting the visitor to slow down. What struck me most about the Voice of the Gallery pilot wasn't the technology itself, but how it did just that. During the research people who told us they'd normally walk through a room in minutes were stopping, asking questions, looking for longer and engaging on their own terms more deeply.”
The Results
Over the course of the pilot, visitors consistently chose to keep talking, explaining how comfortable they were with the experience and how it encouraged longer engagement.
71%
Of testing sessions went past 5 conversation turns
31.6
Average turns per session for those who kept going
120
Turns from one tester in a single 30-minute session
They Didn't Want to Stop
When testers engaged, they really engaged: of the 71% of sessions that went past 5 turns, the average length was 31.6; one tester racked up 120 turns over 30 minutes while moving naturally from one painting to the next across multiple rooms. Visitors followed their own curiosity rather than a set route, with the Voice of the Gallery providing a genuine expert companion that accompanied them wherever they wanted to go rather than just leading the way.
AI You Can Trust
Testers instantly saw the Voice of the Gallery as a trusted source and fundamentally felt different to a generic service. They held their AI conversation to the same high standard they would hold a Gallery curator, expecting responses to be entirely accurate at all times, and were very satisfied when they learned the AI did indeed rely solely on Gallery knowledge for its answers.
Read more about how our AI works →“Trust was the question we cared about most going in, and it was also the most interesting thing we learned. Visitors inherently trusted our experience.”
No Silly Questions
Many testers - especially those with less experience of being in cultural spaces - reported that they are often reluctant to ask questions they think might expose their lack of knowledge or make them sound naive, yet felt very comfortable asking those same questions to the Voice of the Gallery. Freed of that self-consciousness, visitors leaned further in rather than holding back, often going deeper into a painting or a story than perhaps a guided tour would typically allow. As one tester put it, the experience was “really good for people who aren't very confident... and do have questions but want to be able to ask them without feeling embarrassed.”
A Voice for All
Some of the pilot's most striking moments came from the ease and speed with which the majority of testers mastered the “hold-and-speak” interface - including older demographics that are traditionally least likely to engage with new technology. Some were initially wary, but were typically holding fluent, natural conversations within minutes, often becoming the most enthusiastic participants in the room by the time they left. As one older tester said, “the only problem is that you'd never get me to leave!”
“The research told us that there was a different type of engagement happening, a shift to a two-way conversation. The AI supported the visitor's questions and also helped to make them feel more comfortable to ask those questions on their own terms.”
Speaks Your Language
The Voice of the Gallery's reach went well beyond English. Testers held extended conversations in French, German, Spanish and Italian, several remarking on how naturally it handled their own language rather than defaulting back to English - a small detail that matters enormously to visitors more used to being translated at than spoken to directly by a cultural institution.
Looking Ahead
The pilot's reception has already led to a further test, with Dwel supporting the National Gallery's Masterpiece Tour. Starting at the South Shields Museum & Art Gallery, a Dwel-powered conversational companion provides a way for visitors to explore Pierre Mignard's ‘The Marquise de Seignelay’ (1691) on their own terms, asking the painting whatever they want and discovering the stories behind this astonishing masterpiece.
“There's nowhere more humbling to prove what Dwel can do than the National Gallery. Watching visitors lean in closer to a painting, ask another question, then another, because a piece of the national collection has genuinely come alive for them - that's the moment this technology is built for.”
More Case Studies
Want to see what Dwel could do for your place?