AI coffee analysis: Will it replace cupping?
As technology continues to enhance the coffee industry, Jenna Gottlieb speaks with Nicolette Yeo of ProfilePrint to find out how the practice of cupping is advancing.
Artificial intelligence (AI) is everywhere. It surrounds us from the moment we wake up to the moment we go to sleep, and can be found in almost every industry, be it healthcare, farming, or urban planning.
According to recent statistics, around 80% of retail executives expect their companies to adopt AI-powered automation by 2027, while PwC predicts that up to a third of jobs could be automated by the mid-2030s.
The coffee industry is no stranger to AI. Over the years, it has benefitted from a range of AI-powered devices, from self-serve kiosks to innovative superautomatic machines, like the Carimali SilverAce.
However, a more recent arrival in the space is AI tools for analysing and grading coffees.
Coffee analysis and grading is an important stage in the supply chain. It involves evaluating the visual, physical, and sensory properties of a coffee to determine its overall quality and price. This includes aroma, moisture content, density, and range of defects.
“In the coffee industry, sensory analysis is typically done through cupping by a sensory panel to ascertain certain profiles of a coffee, such as flavour, balance, body, sweetness, acidity and so on,” says Nicolette Yeo, head of marketing at ProfilePrint, an AI tool that creates a digital fingerprint for food samples.
“Visual examination involves checking for defects, bean size, colour, etc. All these parameters come together to form a coffee ‘rank’ on a 100-point scale.”
To qualify as “specialty”, green coffee beans have to score 80 points or more. However, while the cupping score protocols are relatively robust, there is still room for subjectivity. In other words, what one person considers to be an 82 coffee could be an 84 for someone else.
The emergence of AI in coffee quality analysis has simplified this process and provides exciting new options to farmers, roasters, and buyers. But will it replace traditional coffee analysis altogether?
Which AI tools have emerged in recent years?
For centuries, coffee beans have been graded through a fairly labour-intensive process that culminates in a cupping by a sensory panel of Q graders. It must be done in a controlled environment and often requires a range of equipment, including grinders, scales, and a moisture and density analyser.
However, over the last few years, various AI-powered tools have opened up opportunities for automation, while also improving the objectivity of the coffee grading process. Nicolette explains that this has benefits for people across the entire supply chain.
“Combining sensor technology and AI, we’re able to analyse and predict sensorial parameters based on a molecular analysis of the coffee sample,” she says. “In addition to being able to synthesise information objectively and rapidly, the technology also makes sharing coffee grading reports with partners and buyers more effective.”
With AI, additional data about the bean can also be collected to identify key trends, such as how origin, humidity, and altitude can affect the quality of a bean. It combines data with algorithms to discover patterns and generate insights or predictions, helping free up time for more value-added work.
“Using the example of how ProfilePrint analyses and predicts cupping scores for green coffee beans, a model was first trained using a number of green coffee beans and corresponding scores based on SCA parameters by professional Q graders,” Nicolette explains.
“Thereafter, it requires just 50g of an unknown coffee bean sample to generate a ‘fingerprint’ and digital report of its cupping score and other sensorial parameters, non-destructively and within seconds. The AI is also continuously retrained with new samples to improve the accuracy and robustness of predictions.”
And ProfilePrint isn’t the only one making headway in this space. Colombian-Israeli startup Demetria recently closed a $3 million seed funding round to develop its near infra-red sensors which analyse green coffee beans for biomedical markers.
This AI-based platform accurately matches each bean profile according to the industry standard coffee flavour wheel. After beans are collected from the farmer, they are scanned and photographed, and a unique ID is created for each coffee bean. The data is then entered into the traceability data cloud. According to Demetria, the analysis can be done at any stage of the production line, from green to roasted coffee.
As such, AI technology can also be of help to coffee farmers. With AI tools, farmers can receive greater insight into the quality and profile of their green coffee beans, allowing them to manage their farming practices and optimise the quality of their coffees. Among other things, this can help them gain a stronger hand at the negotiating table when it comes to pricing their beans.
“Farmers can access potential buyers of their beans at a lower cost to increase profits,” Nicolette explains. “They can also receive faster cupping reports that are affordably priced, and require just 50g of coffee.”
As well as speeding up the process, farmers can also easily share their cupping reports with buyers globally, which reduces the costs involved in sending physical samples to buyers.
Will AI tools replace traditional methods?
Some in the coffee industry say that cupping is expensive and time-consuming.
It is mainly carried out by coffee experts outside of the origin countries, and is prone to human error and subjectivity. AI, on the other hand, can offer more objective, reliable, and consistent results.
It should come as little surprise, then, that investors are showing an increasing interest in this technology. In addition to Demetria’s recent funding, ProfilePrint has also received investment from some high-level backers, including Olam Food Ingredients, Louis Dreyfus Company, and Sucafina.
Yet despite the excitement around new technology, AI is not without its challenges. One of the most significant concerns is its accessibility to farmers. While traditional methods have their associated costs, AI-powered equipment can be expensive, particularly as the technology is still in its infancy.
It’s doubtful, therefore, that AI will completely replace cupping any time soon. Instead, it is more likely to work alongside it, much in the same way that other AI-powered tools do.
“The intention is never to replace, but rather to help expedite the cupping process,” Nicolette says. “Like any AI, it exists and works thanks to the input of humans.”
Photo credit: ProfilePrint