Proof of life. Blair & El, keeping the human touch at the heart of Addicted.
We use AI. Quite a lot, actually. We use it for research, to interrogate ideas, analyse information, improve processes, translate material and occasionally tell us that something we’ve written makes absolutely no sense.
What we don’t want is AI pretending to be us.
That distinction has become increasingly important as the art world works out exactly what it wants to do with artificial intelligence.
A recent article in Artnet News looked at how galleries, auction houses and other art businesses are beginning to use AI. Bonhams is exploring AI-powered market analytics. New platforms are promising better collector intelligence. Galleries are being encouraged to make their content more discoverable to AI search. And technology that was once financially viable only for the biggest players is increasingly available to small businesses.
That last part interests us, because we’re one of them.
The little guys just got some serious technology
One of the genuinely exciting things about AI is that it can narrow the technology gap between a small independent gallery and mega galleries with vastly greater resources.
We don’t have a technology department, a research department or a data science team sitting somewhere down the corridor. Actually, we don’t have a corridor.
What we do have is access to technology that can help us research faster, interrogate large amounts of information, spot connections, improve how we work and spend less time doing things that don’t particularly need Blair or El to do them.
That matters. Former Art Basel director Marc Spiegler recently made a similar point, arguing that AI is putting high-performance technology within reach of small businesses and individuals for whom it previously wasn’t financially viable.
That’s potentially transformative, but access to better technology and handing over the gallery are two very different things.
We didn’t just switch on AI
One of the slightly misleading ideas surrounding generative AI is that you simply open a chatbot, tell it what you want and marvel as your business becomes 37 per cent more efficient before lunch.
Our experience has been rather different.
Addicted already had a voice. We’ve spent more than a decade developing one through articles across our editorial series, including Word On The Street, Artist Of The Week, Art Education and Art History, alongside our daily social media posts and virtual exhibitions.
When we started integrating AI into the gallery, we spent several weeks getting it to review that body of work. We wanted AI to understand how we communicate, how we talk about artists and their work, the language we use and avoid, how our different editorial series work and, just as importantly, what doesn’t sound like us.
That became the foundation of what we call our Addicted Editorial Framework.
The distinction is important. AI didn’t create our voice. The voice already existed. We used AI to help us understand and document something that had developed organically over more than a decade, then apply those lessons more consistently.
And we’re still developing the framework, because AI can produce a perfectly grammatical sentence that we absolutely hate. It can write something technically accurate that sounds like it escaped from a corporate marketing department, describe an artwork in language no human being standing in front of the work would ever use, or occasionally get something completely wrong.
So we question the results, correct the output, reject things, rewrite things, challenge interpretations and check facts. Sometimes we spend an absurd amount of time arguing over one bloody word. AI assists, we edit and decide. The voice remains ours.
The purpose of the framework was never to teach AI how to impersonate Addicted while we disappeared for lunch. It was to make the technology more useful on our terms.
So what’s the point?
If we’re still doing all that, you might reasonably ask what AI is actually saving us.
Quite a lot, as it turns out.
The value isn’t simply getting AI to produce something faster. Some of its greatest value comes from using the technology as a tool for thinking.
AI can interrogate an idea from several directions, find weaknesses in an argument, compare information, organise research, remember context, suggest something we hadn’t considered and take repetitive work off the desk.
Increasingly, AI can also help galleries understand their collector base. In the Artnet article, collector and technology investor Alan Lau suggested thinking of AI as your own “Emily”, the frighteningly knowledgeable assistant from The Devil Wears Prada who remembers the context of every client relationship.
For a gallery, imagine a system that remembers that a collector bought a particular artist three years ago, prefers smaller works, expressed an interest in photography, doesn’t particularly like editions and once mentioned that they were looking for something for a house they were renovating.
That could be enormously useful.
Now imagine the same system autonomously contacting the collector, pretending to be you and manufacturing a personal conversation designed to make a sale.
We’re considerably less enthusiastic.
There’s a difference between technology helping us remember a relationship and technology pretending to have one.
Emily can help Miranda. Emily doesn’t need to become Miranda.
Some inefficiency is worth keeping
Businesses have spent decades trying to eliminate friction: make everything faster, automate the process and reduce the number of interactions between wanting something and buying it. For many industries, that makes perfect sense.
But art is strange.
A collector spending 40 minutes talking to a gallery about an artist isn’t necessarily an inefficiency. An artist explaining why a work exists isn’t a process waiting to be automated. Remembering that somebody saw an artwork two years ago and was still thinking about the piece long afterwards isn’t simply a useful data point. Those things are part of the experience.
The relationship between artist, gallery and collector isn’t an obstacle sitting between the customer and the checkout button. Done properly, it’s part of why galleries exist. Technology should remove the friction around those moments, not remove the moments themselves.
And that’s where we think the line starts to matter.
Where do we draw the line?
We’re perfectly comfortable using AI to help research an artist, but we’re not comfortable allowing AI to decide whether we should represent one. AI can help us analyse market information, but we’re not handing an algorithm the job of deciding what an artwork is worth and accepting whatever number comes back.
The same applies to collectors. If AI can help us identify an artwork somebody might genuinely respond to by connecting information and context we already have, that could be extremely valuable. But the conversation about why we think that person might connect with the work should still be ours.
And yes, we use AI to help develop articles like this one. AI can challenge the argument, identify repetition, suggest alternative ways of expressing an idea and tell us when something isn’t working.
What AI can’t do is decide what Addicted thinks.
There have been plenty of occasions when AI has suggested something perfectly reasonable and we’ve rejected the suggestion for a very simple reason: it doesn’t sound like us.
That might seem like a small distinction. We think it’s a rather important one.
And then there’s the art
There is another reason galleries need to be careful about where they draw the line. We’re dealing with something fundamentally human.
Artists make work from experience, imagination, obsession, memory, culture, anger, humour and occasionally things even they can’t entirely explain. Collectors respond to those works for equally complicated reasons. Reducing that encounter to an optimisation problem would be a peculiar form of progress.
That isn’t an argument against algorithms helping people discover art. In fact, discovery is one of the areas where AI could be genuinely exciting. The traditional art world hasn’t exactly perfected the process of helping people discover artists outside the usual circles.
If AI introduces somebody in Singapore to an extraordinary artist working in Mexico whom they might otherwise never encounter, brilliant. If AI can help an independent artist become visible without waiting for the usual gatekeepers to notice them, even better.
But discovery should start a human experience, not replace one.
AI is already in the building
The debate about whether the art market will adopt AI already feels slightly outdated. We’re using AI. So are plenty of others. The more interesting questions now are how we use AI, what we allow AI to do and what we deliberately keep for ourselves.
There will undoubtedly be galleries that automate far more than we would. There will be collectors perfectly happy dealing with an AI adviser. There will probably be AI-generated sales conversations indistinguishable from human conversations, if there aren’t already.
Maybe that works, but it isn’t what we’re trying to build.
We want AI doing the things it does extraordinarily well: processing, researching, remembering, comparing, questioning and helping us see things we might otherwise miss.
We keep the conversations, opinions, relationships, judgement and occasional arguments for ourselves.
Because the point of AI, at least from where we’re sitting, isn’t to remove the human from the art business.
It’s to give the human more time to be human.
Until the next one...
Blair & El xoxo