Why AI Can’t Bottle Joy: How We Taste 800 Wines a Year So You Don’t Have To

Why AI Can’t Bottle Joy: How We Taste 800 Wines a Year So You Don’t Have To

From our Head Winemaker's desk: Proof that AI still can't make great wine.

It’s almost 2027 (or, depending on when you’re reading my ramblings, it already is), and we see AI everywhere: our news, our podcasts… even a local tire spot in Napa airs an AI-generated commercial (and it’s baaaaad).

Yes, our super-intelligent computer overlords are talented; however, they have yet to discover how to synthesize wine from a series of ones and zeroes. So you, human-friend, can take comfort and solace in the fact that there are other friends (like me!) using real human senses and tastes and emotions to create wines for our own human enjoyment.

And how do we go about doing that?

The first step in bringing you great wines is finding you great wines from our partners all over the world. We’re fortunate that we have an extensive stable of trusted winemakers and wineries to choose from, so whittling down the choices to create an exciting menu is a challenge–a fun challenge, but still difficult nonetheless.

To simplify, we taste. A lot. Somewhere to the tune of 800 wines a year.

Our winemaking team evaluates each and every one against a rubric of base qualities we feel like all our wines should be able to deliver to our customers:

  • Wines should have great “bones”: For example, wines should have present and discernible primary fruit aromas that match their grapes (we on this side of the business call it “varietal character”). California Sauvignon Blanc? We should be getting a heavy hint of stonefruit in there. Sangiovese from Italy? Expect cherry and lots of it. You get the idea.

  • Natural retained acidity and freshness: Wine with little or so acid can come across as out-of-balance, and it will have difficulty stacking up against foods with richness or acidity of their own. Freshness is a big deal to us–we’re not in the business of selling aged Nebbiolo or Sauternes, so high-energy, aromatic wines are valuable to us.

  • Tannin—and to be specific, tannin quality: Wines are composed of so many elements (acid, alcohol, sweetness, etc.) and need structure to hold all these pieces together in a linear way. And to that point, more tannin is not necessarily better. High-tannin Cabernet Sauvignon can be delightful, but high-tannin Sauvignon Blanc will probably be the most bitter thing you’ve ever put in your mouth. Appropriate tannin levels are wine-dependent. High-quality tannins support the wine from front to back, on the sides and the mid-palate, and provide what may feel like length and weight without being disjointed or chalky. Low-quality tannins are aggressive, may hit you in places that feel odd (like your teeth), and could result in a bitter, short finish. Nobody has time for that.

I hope you’ve been keeping up with your wine ratings, because we look at those, too. Historical data is used by us to guide our palates and break ties, if needed. If our members have traditionally preferred a saline, Granny Smith apple-driven Chardonnay compared to one with heaps of butter, we’re more inclined to select one over the other. Or, if membership has shown to consistently love something like California Zinfandel, we may feel comfortable offering something like Teroldego in the future.

Once we’ve identified the wines that pass our initial triage, we score each on characteristics like aromatic intensity, balance, and how close it hits regional benchmarks for flavors.

Finally, possibly most importantly – we ask the question ‘Is it delicious?’

These wines need to have a happiness-factor (perhaps something close to ‘undefinable charm’?) which is something that I don’t believe machine learning has perfected yet.

Contributed by Garrett C.
Head Winemaker
Firstleaf Wine Club

Home > Resources > Why AI Can’t Bottle Joy: How We Taste 800 Wines a Year So You Don’t Have To