Interview: National Gallery of Art (DC)'s Chief Information Officer, Rob Stein
We chat leadership in the sector around tech and AI, the principles of good leadership and change-making around tech and culture
Rob Stein is the Chief Information Officer at the National Gallery of Art in Washington, DC, where he leads on IT, data analytics, artificial intelligence, and cybersecurity, generally helping the institution use technology effectively. “I’ve really been enjoying thinking broadly about what it means to be a national museum and how we might connect with audiences who cannot visit the Gallery in person”.
Rob started working in academia around technology and high-performance computing, and later transitioned into the museum world. He’s held deputy director roles at the Indianapolis Museum of Art, the Dallas Museum of Art and the Milwaukee Art Museum. And was an executive vice president at the American Alliance of Museums.
Ahead of his talk at Communicating the Arts in Amsterdam this June, he shares insights into the National Gallery’s AI initiatives, leadership in times of change, and the role of cultural organisations in an AI-driven future.
You’re presenting on AI at the Communicating the Arts conference at the Rijksmuseum in June. Can you give us an overview of what your talk will cover?
Absolutely. Everyone attending will have thought about AI—what it means for their work and their lives. I’ll be exploring AI from a leadership perspective.
AI is a disruptive technology, but not the first we’ve encountered in the cultural sector. We’ve seen transitions before—digital cataloguing, websites, social media—and adapted. My talk will look at how leaders can navigate the uncertainty AI brings, how to read the horizon, and how to equip teams with the strategies they’ll need for what’s coming next.
Can you share how the National Gallery of Art is using AI?
We’ve got three main projects.
1. Visitor Feedback Analysis
We get thousands of visitor comments—handwritten, emails, online reviews—and it’s impossible to read them all. We have all of this information where our visitors have taken the time to tell us something, and most organizations don't have the bandwidth to necessarily read, process or learn from all of that feedback.
So we've built some systems that leverage AI to digitize and analyze that feedback. The result is a chatbot trained on real visitor responses. Staff can ask it targeted questions, like what people loved or complained about in our café, or how feedback changed over time. It's a way to help our staff think from a visitor's perspective. That AI can almost embody an icon or a symbol of our actual visitors' words.
2. Alt-Text for Accessibility
We're also using AI to tackle a major accessibility challenge. We're a visual arts organization, and there's something like 20 million Americans who have some visual impairment that causes it to be hard for them to see images on the internet. Those people with visual impairments rely on screen readers, yet most of our collection images lack alt-text descriptions.
We've used AI to generate detailed alt-text for artworks, trained on our curators’ tone and language. It’s fast and cost-effective—it takes about 10 seconds and costs 15 cents per image.
The big challenge is accuracy: it’s wrong about 15% of the time. And right now, we can't know a priori when it's made a mistake. We're working blind. That's an area that we're thinking about right now. I think many, many institutions are going to wrestle with this same thing.
My advice is to find your users and try to work out a way of publishing this content that brings your users value, but is up front about the tools you used to get there.
3. Reimagining Collections Access
Art collection websites, by and large, all look and interact the same. I don't think that's a really compelling experience as a way to spark curiosity among folks.
There's this critical gap – museums catalog all of their objects in the language of registrars and curators. And so if you are a registrar or a curator from another institution, you'll find exactly what you want, because we use the same language you do.
But if you're somebody that's just getting started in their career, or you're a hobbyist, or you're just a normal person who got interested in Mark Rothko, you might not know how to use the right words. So I think something that AI is really good at right now is translating the semantics between different vocabularies.
We're thinking and working and planning all of the ways that AI can transform access to those collections regardless of media type, and can adapt its language to whatever the visitor is coming in with. Our visitors have a lot of knowledge, but it might not be academic art historical knowledge, so AI, I think, can help us leverage that different expertise and hopefully give our digital online visitors a much better experience exploring what is really a fascinating set of information, it's just hard to get to right now.
How much of this is off-the-shelf AI versus built in-house?
It’s a mix.
Our visitor services team uses commercial AI tools—for example, chatbots that can communicate in multiple languages to help our international guests.
At the same time, we’re building our own tools, like the audience feedback chatbot. We use APIs from commercial LLMs, but the development is done in-house. And that's becoming more accessible—AI can now write code, so the barrier to entry is coming down.
AI is evolving so fast. Does that affect how you plan your projects?
Definitely. We're constantly navigating the tension between innovation and infrastructure.
Sometimes you need to build one-off tools just to learn. But don’t promise they’ll last forever. Use them to explore, gain experience, and test hypotheses.
At the same time, you have to discipline yourself to invest in lasting infrastructure too. We call it avoiding "shiny object syndrome." You need both. Leadership here is about balance—knowing when to take risks and when to build foundations.
Looking further ahead, how do you see digital change unfolding in the next 10–20 years?
I think we’re heading into a transformation on the scale of the Industrial Revolution—with both upsides and downsides.
One upside could be a shift in leisure time, which historically boosted the arts and culture sector. Another will be changes in education. Self-paced learning is highly effective, and AI supports it well. People will need to learn continuously as AI takes over routine digital tasks.
But we also face risks. AI may amplify misinformation. It’s going to be increasingly important to teach critical thinking: how to distinguish fact from fiction, how to evaluate sources. Humanities and cultural organizations have a critical role to play there.
Has the National Gallery of Art created an AI policy?
Yes, we began in early 2023 and finalized a policy after a convening of peer institutions later that year.
It includes a quick-start guide for staff, with clear dos and don’ts:
✅ We say that it is okay to use commercial AIs – that means off the shelf – but only with public data. That means public is something you can find on the internet easily. That means you can use those off-the-shelf AIs for lots of things, and we encourage our staff to do it, because actually playing around with these tools is imperative to learn how it works.
❌ We also say that non-public information shouldn't be used in off-the-shelf AIs. That goes for business information that we have. But also personal information; don't put your salary in there, don't put your kids’ names in there, I'd leave your street address out. Think about privacy in the context of these tools.
❌ We talk a lot about bias and the fact that the AIs are trained on the internet, and so every bias that exists on the internet exists in the AI training set. And also that AI's get it wrong. So are you good to take that chance? So we say that there should be human review of everything that an AI generates, and that's where we're standing right now from a policy basis.
✅ But another do is that we are committing to always disclosing when an AI was used to generate content for the public. That's in our policy as well. We feel like one of the best tools for our end user in knowing how to trust or not trust some of this content is to disclose how the content was made. And there's a real difference when a PhD art historian has written this text, or, you know, my intern with an LLM license, chatbot license wrote this text. And it's not that the PhD art historian doesn't make mistakes. They do. We've all made those mistakes. But I think being transparent about content, authenticity is one of the best policy standards that I see around right now.
Are there any sessions you’re particularly looking forward to at the Communicating the Arts conference?
Yes—definitely the session by Morris Hargreaves McIntyre. Their data insights on audiences are always really useful.
And being at the Rijksmuseum itself is exciting. That team has long been at the forefront of digital innovation in museums. I’ve learned a lot from my colleagues there and look forward to reconnecting in person.
The museum tech community is global and incredibly generous. That spirit of sharing—ideas, experiences, challenges—is unique. The pandemic set us back in terms of in-person connections, so it’ll be great to be back together again.
What are you looking forward to outside the conference programme itself, but about being in Amsterdam?
I always love visiting Amsterdam’s museums. The Stedelijk and Van Gogh are favourites. I’m also keen to check out the new STRAAT Museum for street art.
Studio Drift, one of my favourite artist collectives, is opening a space in Amsterdam—I’m hoping it’ll be open when I’m there.
Bonus: Leadership and Change in the Sector
What makes big change projects succeed in cultural institutions?
I think success and change are almost entirely about leadership.
Successful organizations have leaders who understand how change works and guide their teams with compassion, vision, and a coaching mindset.
Technology always brings change. My work is about helping institutions understand how tech fits their mission and connects them to audiences. That’s been my passion for 25 years—and I hope for 25 more.
What defines good cultural leadership?
It’s the ability to hold both a big vision and grounded pragmatism.
You need a compelling vision, but also the realism to navigate systems and complexity. Organizations are like bodies—if one part isn’t working, the whole system suffers. Strong leaders know how to identify those pain points and keep everything aligned and healthy.
Has a cultural experience ever shifted how you see your role in the sector?
Yes, when I transitioned from academia into museums.
In academia, I wrote papers that maybe 20 people would read. Then I worked on museum tech projects in Indianapolis that reached hundreds of thousands. Seeing the impact on real people—neighbours, kids, visitors—was transformative. That sense of connection and impact has kept me in the field ever since.