Read time: 5 minutes
Highlights
AI sensory mapping will shift product development from late-stage tasting to earlier, data-guided design.
- Wine is a signal: AI-powered sensors are being used to identify aroma patterns in grapes and juice before fermentation.
- Taste remains the test: Alternative proteins still need to satisfy consumers, not just meet technical or sustainability goals.
- Mapping creates options: Connecting product fingerprints with consumer preferences can help teams test and refine concepts before costly production runs.
- The advantage is learning speed: Companies that build useful sensory data now could make better product development decisions sooner.
The wine signal
Machines may soon notice a wine problem before a human nose can. That’s the intriguing signal from a collaboration announced on 7 October 2026 by Moët Hennessy, Analog Devices and the University of California, Davis.
The group is applying AI-powered chemical sensing to Fresh Mushroom Aroma, or FMA, a wine fault. Rather than looking for one chemical at a time, miniaturised sensors capture a broader aroma fingerprint from volatile chemicals. Machine-learning models then look for patterns linked to FMA risk, using historical data from spoiled crops at Moët Hennessy’s research centre.
The system currently targets grapes and juice before fermentation. Early testing has reportedly reached about 99 percent accuracy, while processing each sample takes roughly two hours. The announcement is a research-stage signal, but the shift is fascinating. Instead of asking only whether a particular defect is present, the technology points towards reading a wider chemical fingerprint. That’s where I see a much larger possibility for food.
Why taste tests arrive too late
I’ve long argued that taste has been the make-or-break issue for alternative proteins. Consumers might be interested in a product’s nutrition, convenience or environmental story, but they still expect it to taste good.
That creates a painful development challenge. Human sensory panels and consumer taste tests are valuable, but they can be slow, costly and limited in scale. They often happen after a team has already invested in ingredients, formulation work and pilot production. If a product misses the mark, the most useful feedback may arrive when much of the budget has already been spent.
Alternative proteins make this especially difficult. Different protein sources and processing methods can produce beany, bitter or metallic notes. Texture, aroma and flavour release can also change as a product is cooked or combined with other ingredients. A prototype may pass one test and disappoint in a meal, a different format or a different market.
A sensory instrument can’t tell us what every consumer will like. Taste is personal, cultural and shaped by context. But better measurement can help teams ask more useful questions earlier, then take those questions to people for real-world feedback.
From detection to product design
AI sensory mapping is the opportunity to connect a product’s measurable chemical and sensory characteristics with what consumers say they prefer. It’s not simply a machine replacing a tasting panel. It’s a tool for building a clearer map between ingredients, product attributes and human responses.
Some of the building blocks are already here. Electronic noses and tongues use sensor arrays and machine learning for food quality work. In alternative proteins, Symrise’s ProtiScan already combines analytical information with sensory and consumer data to investigate off-notes and flavour performance. It’s a practical signal that industry is already making flavour measurement more useful in product development.
Imagine a team developing a plant-based burger. It could compare the chemical and sensory fingerprints of several protein blends, then connect those profiles to consumer feedback on attributes like savoury depth, bitterness or resemblance to a familiar product. The team could narrow the options before committing to larger production trials. I explored a similar acceleration in how AI is moving from food formulation to enzyme design, and sensory mapping feels like the next step in that journey.
The same approach can help a company set a target profile for a particular market or eating occasion. A product designed for a spicy takeaway meal might need a different flavour balance from one intended for a family dinner. AI sensory mapping could help teams understand those differences and test more targeted formulations.
What will change over time
Through my FutureCUBED™ process, I look for signals, explore the futures they could create and help organisations decide what to do next. This wine collaboration is a signal. The larger question is what happens when sensory fingerprinting spreads into food product development.
In the coming years, we’ll see sensing platforms mature in wine and neighbouring categories, while food companies learn which measurements are useful for their own products. AI sensory mapping will be the tool that turns those measurements into product development decisions.
First movers will begin building proprietary sensory maps for specific categories. Their advantage wouldn’t simply be owning more data. It’d be having well-structured data that links ingredients and formulations to sensory measurements and consumer responses.
Sensory mapping will become a more familiar part of new product development for alternative proteins. Product teams might use it to screen options sooner, direct consumer testing more effectively and reduce avoidable formulation cycles.
Eventually, I can see development becoming a much tighter feedback loop. Consumer insights could help define a target. Models could suggest formulations. Sensors and people could then evaluate the results, with each round improving the next. Machines might propose products against predicted preferences, while human judgement remains essential to deciding what deserves to reach the market.
The strategic question for food leaders
The companies that build strong sensory maps can learn faster than competitors still relying on trial and error alone. But the goal shouldn’t be to collect data for its own sake. It should be to create a better connection between what a business makes and what people actually want to eat.
I’d encourage food companies to start with three questions.
- What are we measuring today? Identify where sensory information is gathered and how consistently it’s recorded.
- Can we connect the data? Link formulation details and analytical results with sensory-panel feedback and consumer responses.
- Where could we test first? Choose one product challenge, such as bitterness or a persistent off-note, and explore whether better measurement can improve early decisions.
The nose may not literally know everything first. Human taste will remain complex, and today’s wine technology has a specific job to do. But the broader idea is powerful. If we can learn to read a product’s sensory fingerprint sooner, we can start designing towards consumer needs rather than discovering too late that we’ve missed them.
AI chemical sensing + sensory mapping + consumer insight = alternative proteins that finally taste right.
To continue the conversation email me at tony@futuristforfood.com
