A Toronto brewer created a beer using AI. But what does that mean?

At this stage, artificial intelligence in craft beer brewing is more about marketing than taste


Artificial intelligence infiltrated the beer world a few years ago, but the trend has only just trickled into the local craft brewery lexicon.

In September, the Kensington Brewing Company put the AI in IPA with A Little Robotic, Toronto’s first beer designed using artificial intelligence. As in many other sectors, brewers can harness data analysis software to automate production processes. For example, a big commercial brewery with a global presence might use AI to optimize a recipe for consistency in mass-produced batches or to incorporate customer feedback.

But that’s not why Kensington’s brewers spent a year working with Toronto-based machine learning company Dessa to concoct an smoothie-style, double dry-hopped IPA made with mango puree and lactose.

For the brewery’s president Michael Gurr, AI is a way stand out in the city’s increasingly crowded craft beer market.

“The sheer volume of high-quality product available now is overwhelming,” he says, adding the last time he visited the LCBO’s Queens Quay location he saw a craft brew rack stocked with around 100 different varieties. “It’s increasingly difficult for small-to-medium-sized brewers to stay competitive.”

As far as marketing initiatives go, Gurr calls A Little Robotic “not inexpensive as a one-off.” The beer’s initial run is 2,000 litres – a “cautious” batch size by Kensington’s standards.

“Never before has it taken us a year to produce a single beer. It was a fascinating exercise for us internally,” he explains. “But there’s still a ways to go before [AI] can be viable for a brewery on our scale.”

The process basically works liked this: Dessa analyzed a huge amount of publicly available data – beer recipes, beer styles, food pairing information – and the brewery “tweaked knobs” to refine the recipe. The computer might spit out 20 recipes, which the company’s (human) brewmaster looked over to determine what works, what doesn’t work, and reprogram the algorithm. This took six months compared with a typical three-week production cycle.

Beyond a marketing hook, Gurr says the value-add of using AI for a craft brewer is the potential to innovate in the research and development phase. Kensington drinkers enjoy trying new things, he explains, but coming up with experimental yet commercially viable recipes on a regular basis is challenging.

“If we had some way to spark that creativity, I think that could improve the quality of our product and speed the process up,” he says.

Multinational brewery Carlsberg began doing this in 2017. The Copenhagen-based company is spending millions to analyze yeasts and develop “sensors” that detect beer flavours and aromas – what it calls “beer fingerprints.” The goal is to cut development time by 30 per cent, improve quality control and get product to market faster.

This requires a huge amount of time and data – not to mention actual humans who must taste things and do research.

“By the end of the process of creating an AI program in order to produce recipes, you’ve essentially ended up with a brewer,” says Toronto-based beer writer, podcaster, historian and sensory expert Jordan St. John. “A really qualified human taster is probably always going to be better than an AI program just because it’s very difficult to give an AI program a tongue.”

However, there is another step in the beer lifecycle in which AI could sideline human beings. He gives the example of a small-batch Creemore Springs beer called Mad & Noisy from 2012 – the first IPL brewed in Ontario.

St. John says the beer had a “brilliant citrus and pine character to it, with maybe a little bit of tropical fruit.” Then the brewer gave it to a focus group.

“A soon as they focus grouped it, it lost all the rough edges,” he says. “It wasn’t interesting anymore and nobody bought the thing. I think that AI standardization probably replaces focus grouping. You don’t have to test individual batches against the public. You do that once and then continue based on the idea that you’ve got an artificial intelligence quality control.”

Which brings us back to marketing. Kensington has gone all out with packaging, adding a sleeve to A Little Robotic that animates the can design when you slide it up and down. It reinforces the brand’s goal to appear playful, adventurous and collectible.

So, should you try it? If you like your IPAs on the sweeter side and high in alcohol content – and can afford premium pricing.

“Regardless of how good the beer is,” says St. John, “it certainly draws the attention.”

@KevinRitchie

Brand Voices

Leave a Reply

Your email address will not be published. Required fields are marked *

NOW Magazine