A Human AI Could Never Replace: Kevin Huvane, Agent to Stars
You’ve probably never heard of Kevin Huvane, but you do know Julia Roberts, Jennifer Aniston, Sarah Jessica Parker, Meryl Streep, Brad Pitt, and Jennifer Lopez, all of whom he has represented as an agent at CAA. When I first wrote about AI, here, after considering diplomat and Freudian psychoanalyst, I settled on CAA agent as the job safest from replacement by non-human intelligence.
I don’t know Huvane, but I am acquainted with other agents and managers and even had one 30 years ago in connection with a potential movie sale of my novel (never happened). So I can imagine what skills Huvane has had to employ to represent these stars successfully, which I separate into two general categories: Pattern Recognition and Emotional Intelligence (EI).
Experts in many narrow specialties rely on pattern recognition to make decisions. Chess masters evaluate board positions quickly not by looking many moves ahead, as you might think, or as a computer would, but by rifling through the thousands of games they have played or watched and are stored in their brains.
As someone who has worked for decades in a specialized area of investing, I assess every new idea first by seeking an analog in situations I’ve seen before. A lot of these prior cases are memorialized in writing that AI could read, e.g., Bloomberg stories, but the key factors that the past events turned on may reside only in my head. Outcomes in the world of investment, and business generally, often involve irrational choices that would be hard for even the most advanced model to infer. Look no farther than Trump, Musk, and Tesla.
AI is better at detecting patterns in situations where cause and effect is clear and feedback is quick, called “kind environments” by the late behavioral psychologist Daniel Kahneman. In “wicked environments,” where new scenarios are constantly presented, pattern recognition starts to break down and human judgment is needed. Computer science is a kind environment, making basic coders replaceable by AI, but equity investing is not, especially in emerging markets.
If I’m approached to invest in a junior mining company in Africa — actually, I just was — my mind calls up similar investments I’ve made over the years. A couple were home runs, more were zeroes, and the only way to find the pattern is by identifying nuanced elements in each situation. In one big loser, the CEO showed up to our first meeting in two-toned shoes, with his mistress in tow. I’ve never written about this before and now that I have, in a public post, ChatGPT can improve its model by adding the risk in trusting a Kazakh promoter with two-toned shoes and a mistress on his arm.
EI is also part of my job, e.g., assessing company managements and dealing with co-workers and clients, and I’m pretty good at it. However, relationships in finance are businesslike and when investors have something to say, it tends to be straightforward. (Brits require the most reading between the lines and Australians the least.) So it is possible that AI could manage much client interaction in finance, but in some fields the interrelationship between pattern recognition and EI is so complex it seems impossible for a computer to handle.
CAA’s Kevin Huvane
Picture Kevin Huvane making a movie deal. He has a script in hand, his client W is into it, and now he has to consult his mental file of all the studio executives, producers, directors, etc., to figure out where to take it. He knows that executive X had a hit with a similar movie; he knows that director Y wants to work with W and can set it up with producer Z; but he also knows that Z has hated X ever since they were assistants together at William Morris. And then there’s the screenwriter, V — but who cares what he thinks? Could Anthropic figure out what to do next by reading every copy of Hollywood Reporter and Variety since the Silent Movie Era? I doubt it.1
Having played 4D chess and assembled the deal, Huvane must now extract the best fees for his clients (and CAA of course). ChatGPT could suggest an ask for client W after scanning every fee reported in the trades, but to AI these are just statistics, whereas Huvane and his associates know the stories behind the numbers. For example, ChatGPT reads that Bette Crawford, an actress with similar clout to W, just got $2 million for a picture; but Huvane also knows that Crawford accepted a reduced fee because the studio also greenlit her pet project where she plays Dian Fossey (“that woman who liked gorillas”), which will flop.
One aspect of an agent’s work where AI might be helpful is deciding whom to sign. By analyzing massive amounts of data, a model can detect trends and try to predict future stars. On the other hand, there are already today newcomers who seem to have been generated by AI and aren’t resonating, no matter how much the studio PR machines shove them down the public’s throat. What creates a screen star is ineffable — and which AI model would’ve identified Adam Driver as a romantic lead?2
Then there’s the client side, where it’s not about AI at all, but EI. Huvane has represented some of the most physically beautiful movie stars and seems to specialize in divas. While his client Streep is reputed to be a mensch, what about Roberts, Parker, and Lopez? I’ve had some bad days as a fund manager, but at least I’ve never had to deliver bad news to Jennifer Lopez. Huvane must have, yet in photos he looks none the worse for wear. ChatGPT may be able to deal emotionally with stars, but only when they themselves are artificially generated (which may be coming soon).
One personal experience I recall from 1993 of my agent giving bad news was after he’d read my spec screenplay. He began by saying, “I’m thrilled! You know how to write a script, and you can definitely do it!” Pause. “But I can’t sell this one.” I left the call with various emotions, none of which was anger at him or desire to find another agent, proof that he deployed EI well. After a period of reflection, I concluded that instead of trying to write for Hollywood, I would start a hedge fund. It just seemed easier.
In his new memoir, Barry Diller writes that when he was working in the William Morris talent agency mailroom, he spent nights reading every document in the files. He was like a human AI simulating years of personal experience, so it’s no surprise that he was already seasoned and successful by his late twenties.
I have an investment through a fund in a company called beatBread that lends money to musicians, collateralized by their future streaming revenues. The company’s algorithm operates in a relatively kind environment because it doesn’t have to predict whether the artist is going to become a star, but only whether the specific track or group of tracks will generate enough to pay back the loan.
The picture of "Huvane with Pitt and Aniston" is not Kevin Huvane.
Are you trying to replace him with an AI image?