The Automation Dream: total repeatability and implementing of AI in roasting coffee
The dream that does not leave us—and at some point will be implemented fully—is still yet to come.
I’m talking about having a fully automated roasting machine that does not require a roast master, just an operator. Someone to press the button while doing other things around the shop. Total cost optimization.
It is the same temptation we see with super-automatic espresso machines: how tempting is it to save money on labor while gaining extra stability? Business owners will understand exactly what I am talking about. Roasters and baristas? They will probably hate it.
The truth is, sooner or later, it will happen. But we are not there yet. For now.
"Controlled Surprises"
So… Why aren't we there yet? I would like to start with an example I usually give about flavor stability and why it is crucial to invest in it.
In my town, there are quite a few coffee shops. When I’m not visiting specialty shops and decide to pass by a regular one, I always choose the same place. They have a fully automatic espresso machine that is usually dialed in.
Why do I do it? Because as much as I like surprises—the possibility of choosing between an Ethiopia or a Colombia on V60, or having a new batch brew —I like controlled surprises. I want to be surprised only when I decide to be surprised.
The rest of the time, I prefer stability. I go to that shop because the coffee will most probably taste the same today as it did a week ago. It will not vary; it will not depend on the skill of the barista or their mood. This can be good or bad, actually. I am missing out on "amazing" coffees, but I am also missing out on "very bad" ones. To me, that sounds like a fair trade.
Automation in Coffee Roasting: Volume vs. Value
What does this case of me having an espresso have to do with the automation of the roasting process? Everything.
Imagine you can adjust a roast profile to get a cup that is not "the best," but certainly not "the worst." A coffee that is sellable and "ok"—not a mind-blowing experience, but a consistent one. How attractive is that? Very attractive.
I know many business owners who love this idea. It feels like having the best of both worlds: being able to say you roast your own coffee without having to invest heavily in labor and specialized knowledge.
It is the same with super-automatic espresso machines. They serve a purpose in specific conditions:
High labor rotation.
When the focus is not purely on the coffee.
When the customer wants volume and stability over an exceptional experience.
There is a detail though: fully automatic espresso machines only work in tandem with very stable roasts and coffee that is properly rested. If the roast changes, the system breaks.
Does Automatic Mean Stable?
So, do automatic roasts equal stable roasts? Yes and no.
Yes: It can very sucessfully substitute and even be much better than someone with little to no experience.
No: It does not substitute for a skilled Roast Master or a good taster who understands when a profile needs tweaking—even when the graph looks identical to the previous one.
To understand why automatic roasts aren't always stable, we have to look at the metrics of repeatability. What exactly are we repeating?
Bean temperature curves?
Exhaust temperature?
Inlet temperature?
Heat application?
Total roast time?
Drop temperature?
Development ratio?
What is our metric? What is our goal?
The "Probe" Problem: Path vs. Destination
This alone showcases why a roasting machine operator needs a deep understanding of the system and the capacity for sensory analysis.
The technical part can be handled by AI or applications that suggest how to repeat a curve. But the second part is purely human: describing what you are tasting, why you are tasting it, and how to fix it today so you can repeat the success tomorrow.
We must not forget that a roasting graph is not the coffee itself. It is a record of changes in the probe readings. Probes can be poorly located, thin or thick, old or new, different models... Repeatability based on probes is just the destination. What actually creates the flavor is the path taken to reach that destination.
The Variables of Failure
If you rely on a profile created by someone else on the same machine model, it only works if everything is identical:
Altitude: Atmospheric pressure changes the roast.
Coffee Condition: The same lot changes as it the bag stays open or ages.
Hardware: Machine position, heat source, and probe calibration must be identical.
Coffee is a living thing. Profiles require adjustments as the coffee gets drier, more moist, or simply older. You must mitigate the problems that appear with aging to maintain the same flavor.
Finally, there is the issue of tasting. How do you taste if a profile was truly repeated well? How do you identify what is "off" when the graph looks perfect?
There is a high demand for automatic profiling in coffee roasting, and much work is still to be done.
Is it better to have a top-tier professional serving you the best espresso in town? For sure—that is an amazing experience.
But depends on what you are looking for.
Repeating a profile is a solution for achieving a certain degree of stability. It may not be spectacular, and it may not be perfectly accurate in flavor, but it serves a purpose.
Is it better to have an experienced roast master at the helm? Always. But again, it all depends on the goal of your company.
