Illustrative examples. These are composite scenarios built to show the guidebook's ideas in action, not reports on specific named businesses. The numbers are realistic but representative.
Craft brewery · Beer
Holding a flagship lager in spec
A regional craft brewery, ~15,000 hL/year, one flagship pilsner that is 40% of volume.
The challenge
The flagship occasionally drifted: a buttery note in some batches, the odd one finishing a couple of points high. Customers noticed. The head brewer suspected fermentation temperature control but could not prove which batches were drifting or why.
What they tried
They started by writing a one-page quality definition for the pilsner, target FG 1.008 to 1.010, clean, no diacetyl, with a clear pass/fail line. Then they pulled their last 40 batches into a single clean sheet and used an AI assistant to talk through which batches missed spec and what those batches had in common.
Keeping the brewer in control
The assistant never touched the tanks. It surfaced a pattern (the out-of-spec batches clustered around warmer ambient weeks with a slower diacetyl rest) as a hypothesis. The brewer verified it against the actual fermentation logs and a small controlled trial before changing the rest schedule.
What changed
Out-of-spec batches on the flagship dropped sharply over the next two months. More importantly, the brewery now has a written spec and a habit of reviewing drift, the tool just made the pattern visible.
Distillery · Spirits
Making years of cask notes searchable
A single-site malt distillery with two decades of handwritten and spreadsheet cask records.
The challenge
Almost all the knowledge about which casks and warehouses produced which character lived in one stillman's head and a stack of ledgers. When he was away, the team was stuck. Finding "which ex-sherry casks from the 2015 fill we should pull for the next batch" took hours.
What they tried
They digitised the records into a consistent format (one row per cask, consistent units and names) then used an AI assistant to ask questions of that data in plain language: maturation, fill dates, warehouse position, past sensory scores.
Keeping the brewer in control
The tool proposed shortlists; the master distiller still nosed every cask before any blending decision. The assistant sped up the search, it did not make the call. They also kept the original ledgers as the source of truth.
What changed
Selection that took half a day now takes minutes, and the knowledge no longer depends on one person being in the building. The nose still decides.
Taproom brewery · Beer
Triaging an off-flavor under time pressure
A small taproom brewery, 7 bbl kit, one brewer and one assistant.
The challenge
A pale ale came back from the tank with a buttery note the day before it was due on tap. The brewer needed to decide fast: rest it longer, dump it, or serve it. No second brewer to sanity-check against.
What they tried
The brewer described the symptom to an AI assistant with real detail (style, FG 1.014, fermented warm at 22°C, fast primary) and asked for likely causes and options, then asked it to show its reasoning.
Keeping the brewer in control
The assistant pointed at incomplete diacetyl reduction and suggested a warm rest. The brewer treated that as a hypothesis, confirmed the FG and tasted again after a 48-hour rest, and only then decided. When the tool offered a confident-sounding additive dose, the brewer ignored the unverified number and trusted measurement instead.
What changed
The rest cleared the note and the batch went out. Just as valuable: the brewer used the tool as a thinking partner for triage without handing it the decision.
Production brewery · Beer
Planning production without tying up tanks
A growing production brewery juggling a core range and frequent seasonals across limited tank space.
The challenge
Tank scheduling was guesswork. They were either short of a core beer or sitting on seasonal stock that aged past its best, with cash tied up either way.
What they tried
Using their past sales and seasonality, they had an AI assistant draft demand forecasts and a candidate brew schedule that respected tank availability and turnaround times.
Keeping the brewer in control
The forecast was a starting point, not an order. The production manager reviewed and adjusted every cycle, overriding the model for known events (a festival, a tap takeover) that the data could not see. The schedule was a draft the team corrected, never an autopilot.
What changed
Fewer stockouts on core beers and less aged seasonal stock. The planning conversation got faster because there was a sensible draft to argue with.