Matthew Sabourin built his own production planning system before he thought about buying one.
He drew it with his intuition, charts and years spent waiting for the beer to ferment in Nonsuch Brewing Co.a majority Indigenous-owned craft brewery in Winnipeg’s Exchange District. She had an intelligence about her, but the whole system lived in Sabourin’s head, and when the brewery needed her elsewhere, he was caught.
“It became so difficult for me that we had to shelve the tool,” says Sabourin, president and co-founder of Nonsuch. “We just stopped using it because I had to be other places.”
Most companies have one. The table that started as a quick fix that became a system, which then turned into a minor hostage situation.
An industry friend saw what Sabourin had built and told him he needed to talk to Cameron Bergen, founder and CEO of Fashion40an enterprise data intelligence firm based in Steinbach, Manitoba.
The two became one Canadian Food Innovation Network webinar on May 28, moderated by Digital Magazineto walk through how they are solving it.
Nonsuch is not alone in this. In production. When operational knowledge lives in one person, the cost shows up on the production floor as capacity you didn’t know you were losing.
Capital spending across Canada’s food and beverage sector to fall 5.3% in 2025according to Farm Credit Canada (FCC), and early indicators suggest investment may weaken further in 2026. The sector is also 50,000 workers shortaccording to the Canadian Agricultural Human Resources Council.
For most of the sector’s 11,000-plus businesses, this starts with the operations they already have. Unfortunately, some of them are hidden inside a spreadsheet called “Final_FINAL_revised_v7”.
If you are luckyit is marked with a date or with someone’s initials.
Nonsuch and Mode40’s partnership is a working example of what it takes to bring AI to a small manufacturer’s floor, messy data, acquisitions, slow build.
Small operations carry billions of dollars of complexity
Beer sounds simple when you’re the one ordering it at the bar. But behind the scenes, the customer wants it tomorrow, the tank needs three more days, and the production board starts to feel more like it’s painting blindfolded.
Bergen has solved this type of problem in larger facilities. Nonsuch packs it into a piece of space.
“It’s probably one of the most difficult production planning calculations we’ve seen, even across multibillion-dollar enterprises,” he says.
You can’t tell a batch of beer when it should be done fermenting. An additional day in the cascades of the reservoir is spent throughout the plan.
“Why not just make big batches of everything and shelve this forever, right?” Bergen says. “Because quality is very important to what they do.”
The underlying problem is familiar to any operation with unpredictable results and customer requests arriving faster than production can respond.
In Nonsuch, the variable is biology. We’re talking about yeast and the environment it requires to do its job. It’s done when it’s done.
Elsewhere it could be supply chain delays, machine variability or changing customer specifications. The details vary, but the headache is familiar.
Mode40’s system creates a digital model of the brewery’s production process, then uses Nonsuch’s own rules about quality, capacity and customer commitments to help plan what should be brewed, packaged and prioritized.
Before Mode40, Sabourin managed it all across multiple manually updated spreadsheets. A beer may need to be packaged in different keg sizes and formats, while multiple batches are placed in different stages of fermentation and orders arrive on the same day they are due to ship.
“I built a spreadsheet that’s manually filled in, updated every now and again by hand, (and) prone to human error,” says Sabourin.
Unfortunately, every production window lost compounds.
“Once that opportunity is gone, you can’t get it back,” he adds. “Gone forever.”
One of the less glamorous problems was getting the data out in the first place. Sabourin says Nonsuch’s enterprise resource planning (ERP) didn’t have an export button, so the Mode40 team pulled the information from the website’s code.
It doesn’t get any less manufacturer-coded than “the data exists, technically, but good luck recovering it.”
HE cannot fix what he cannot see
Bergen starts every consultation on-site, because what a manufacturer describes on a phone call and what he finds on the floor are almost always different problems.
“If you haven’t seen it, don’t try to figure it out,” he says.
Mode40 first builds a digital model of the object, then puts operator decision rules on top. When two orders need the same tank on the same day, what wins? The bigger the customer, the tighter the margin, the beer you can’t wait for?
“You’re really going to respond to them,” Bergen says. “Those sets of rules become the fundamental barriers to how the system works.”
Without this foundation, even advanced AI will produce results that look right on the screen and fall apart on the floor.
This goes beyond food and drink. Applying AI before the operation map is how companies end up with a sure answer to the wrong problem.
The Nonsuch-Mode40 partnership is ongoing, with sales forecast work still in progress. Bergen says one of the early opportunities is to make capacity issues visible before lost orders are made.
He says the goal is to detect conflicts early enough that manufacturers can adjust production or call customers before a missed order becomes a relationship problem.
“You’ve mitigated the risk ahead of time without letting a customer know,” says Bergen.
No one wants another babysitting screen
Sabourin entered the implementation process more technically confident than most, having built his platform in a previous business. However, his employees had questions about the six different systems they had put together.
“Without full buy-in, you’re going to be fighting uphill all the time,” he says. “So you need everyone to believe that this is the right solution, so they can be partners in implementation and not resistance.”
Fair enough. Most workers have met at least one system that was supposed to save time and somehow create a new part-time job.
Mode40 conducts facilitation sessions before implementation begins to show early concerns, everything from data privacy to scars from legacy systems that promised more than they delivered. Bergen says most of these fears stem from the way the technology was sold.
“The more complicated it is that way, the more likely they are to sell you on the idea,” he says.
Modern tools should be harder to explain than to use. Bergen’s reading is that when a platform overwhelms you with screens and features, you’re paying for complexity, not results.
For Sabour, profit is the physical plant. Nonsuch has a fixed number of tanks and a fixed amount of floor space, and the goal is to get the most out of both.
“The more we can get out of it, the better,” he says.
A new FCC report found that 3% annual GDP growth in Canadian food and beverage production over the next decade could add $40 billion to the national economy and create 217,000 jobs.
For the SMEs that make up the bulk of the sector, this kind of growth starts as it has at Nonsuch, making the most of what’s already on the floor.
The last shots
- Many manufacturers are working with institutional knowledge locked in one person and forecasts spread across disjointed systems.
- If an AI platform can’t map how your operation works before it starts to automate, the results are likely to be wrong, even if they look safe.
- When evaluating new tools, the Bergen test is whether the architecture allows you to start with a problem and grow from there.
Watch the full webinar, moderated by Digital Journal, below:
