The technical track: applied AI across the brewing process, from grain intake to packaged beer. Grounded in brewing fundamentals — malting, mashing, wort production, yeast and fermentation, maturation, filtration, packaging, hygiene, and quality control — and honest about where the models earn their keep and where they don't. The full track, in order:
-
The Bar Chart in Brewing: Three Beer Use Cases
The bar chart is the workhorse of brewery analytics and the right default more often than any other chart. Three concrete beer use cases (style volumes, account ranking, brewhouse loss by stage) and the rules that keep it honest.
-
The Grouped Bar Chart in Brewing: Three Beer Use Cases
When you need to compare several series side by side (this year vs last, plan vs actual, style across regions), the grouped (clustered) bar chart is the tool. Three beer use cases and when to switch to a different chart.
-
The Stacked Bar Chart in Brewing: Three Beer Use Cases
When the question is composition (what makes up the total), the stacked bar chart shows the parts and the whole at once. Three beer use cases (channel mix, cost-per-hectolitre build-up, volume by style over time) and the 100% variant.
-
The Pie and Donut Chart in Brewing: Use It Sparingly (Three Cases)
The pie chart is overused and often the wrong choice, but it has a few honest beer use cases. Where a pie or donut genuinely works, where a bar chart beats it, and the one rule that keeps it readable.
-
The Treemap in Brewing: Three Beer Use Cases
When composition has many parts and a hierarchy, a treemap shows the whole portfolio in one rectangle sized by value. Three beer use cases (SKU portfolio, sales by region then account, ingredient spend) and where it beats a pie or a long bar chart.
-
The Line Chart in Brewing: Three Beer Use Cases
For anything that changes over time, the line chart is the default and usually the right one. Three beer use cases (fermentation curves, sales trends and seasonality, KPI vs target over time) and the rules that keep it readable.
-
The Combo (Line + Bar) Chart in Brewing: Three Beer Use Cases
Two metrics on two axes (volume as bars, margin as a line), the combo chart shows a quantity and a rate together. Three beer use cases and the dual-axis trap that makes this chart easy to misuse.
-
The Area and Stacked Area Chart in Brewing: Three Beer Use Cases
An area chart is a line chart with the space below filled, good for cumulative totals and, stacked, for how a composition shifts over time. Three beer use cases and when a line or stacked bar is the better call.
-
The Scatter Plot in Brewing: Three Beer Use Cases
When the question is 'are these two things related,' the scatter plot is the answer: each point a batch, account or beer. Three beer use cases (process correlations, account value vs effort, recipe trade-offs) and the correlation-isn't-causation rule.
-
The Bubble Chart in Brewing: Three Beer Use Cases
A scatter plot with a third variable encoded as bubble size: show volume, value and growth together. Three beer use cases (portfolio maps, account prioritisation, market opportunity) and why size can mislead.
-
The Histogram in Brewing: Three Beer Use Cases
Averages hide the story; a histogram shows the whole distribution: the spread, the shape, the second peak. Three beer use cases (fill-volume distribution, ABV consistency, sensory score spread) and the bin-width trap.
-
The Box Plot in Brewing: Three Beer Use Cases
When you need to compare the spread and outliers of many groups at once, the box plot is unbeatable. Three beer use cases (consistency by brand, yield by shift, sensory spread by taster) and how to read the box.
-
The Heat Map in Brewing: Three Beer Use Cases
When two categories meet (hour by day, style by month, account by SKU), a heat map shows intensity as colour, so patterns jump out of a grid. Three beer use cases and the colour-scale rules that keep it honest.
-
The Waterfall Chart in Brewing: Three Beer Use Cases
A waterfall chart shows how you got from one number to another: the steps that build or erode a total. Three beer use cases (margin bridge, volume bridge, budget variance) and the rule that keeps the bridge honest.
-
The Funnel Chart in Brewing: Three Beer Use Cases
A funnel chart shows how many survive each stage of a process and where the biggest drop-off is. Three beer use cases (sales pipeline, NPD stage-gate, trade-promotion conversion) and when a funnel flatters more than it informs.
-
The Gauge and KPI Card in Brewing: Three Beer Use Cases
The gauge and the KPI card answer one question fast: are we hitting target right now? Three beer use cases (production vs plan, OEE, on-time-in-full) and why a single number needs context to mean anything.
-
The Gantt Chart in Brewing: Three Beer Use Cases
A Gantt chart lays tasks against time as bars: perfect for schedules and overlaps. Three beer use cases (the brew schedule, tank and vessel occupancy, NPD project timeline) and where a calendar or a different view serves better.
-
The Geographic Map in Brewing: Three Beer Use Cases
When location is the variable, a map beats any table: choropleths, dot maps and catchment views put beer sales on the ground. Three beer use cases (territory performance, distribution coverage, taproom catchment) and the map traps to avoid.
-
Why Beer Is a Surprisingly Good Fit for Machine Learning
Brewing is a repeatable batch process with measurable inputs and clear targets: exactly the structure machine learning exploits. Here's why, and where it breaks.
-
Predicting Beer Colour (EBC/SRM) From a Recipe
How AI predicts finished beer colour in EBC/SRM from malt colours and grist proportions, where it beats Morey's equation, and where it breaks.
-
Classifying Beer Styles With Machine Learning
How machine learning classifies beer styles from OG, FG, ABV, IBU and colour, where it helps QC and competition sorting, and where styles blur.
-
Predicting Beer Foam and Head Retention
How AI models beer foam and head retention from foam-positive polypeptides, iso-alpha-acids and lipids, where it helps QC, and where it breaks.
-
AI for Draught Line Cleaning and Pour Quality
Use flow meters, pour counts and line temperature to predict optimal cleaning intervals and flag pour-quality issues, without compromising hygiene.
-
Anomaly Detection on Brewery Sensor Data
How anomaly detection learns the normal envelope of brewery sensor data to flag drift, leaks, and faults earlier than fixed thresholds.
-
AI for Brewery Production Scheduling
How optimisation and forecasting plan the brew, ferment, and package schedule across finite tanks, maturation times, and line slots to hit due dates.
-
Electronic Nose and Tongue: AI for Off-Flavour Detection
How electronic nose and tongue sensor arrays plus machine learning screen beer off-flavours fast, where they help QC, and why the panel still rules.
-
Planning for CO2 Shortages With Predictive Models
Forecast CO2 demand against fermentation recovery and supply to size buffers and recovery investment, a lesson sharpened by the 2022 supply crunch.
-
Beer Authentication and Counterfeit Detection With ML
ML on spectral and chemical fingerprints or packaging images can flag counterfeit premium beer, backed by reference libraries and chain-of-custody data.
-
Scaling a Homebrew Recipe to Production With AI
Predict how hop utilisation, evaporation, mash efficiency and fermentation shift when a recipe jumps from homebrew to a bigger system, so the beer survives.
-
Cellar Inventory and FIFO Optimisation With AI
How shelf-life-aware FIFO and allocation across brewery tanks and finished stock cut waste while holding fill rate, and where the approach breaks down.
-
Knowledge Search Over Brewery SOPs With Gen AI
How retrieval-augmented LLMs let brewery staff ask plain-language questions over SOPs, batch records, and manuals with cited answers and what to watch.
-
AI for Brewery Workforce and Shift Scheduling
How optimisation matches brewery labour to the production plan under skills, availability, and labour rules and why people aren't widgets.
-
Predicting Carbonation and Dispense Consistency
How AI models CO2 volumes and dispense quality from inline CO2, line temperature and pressure to cut fobbing, and where the model loses sight of the bar.
-
Building a Brewery Production and Yield Dashboard in Tableau
How to build a brewery production and yield dashboard in Tableau using LOD expressions, brewhouse yield trends and Tableau Pulse digests.
-
Can AI Predict Malt Quality From Barley Analysis?
Machine learning can predict malt quality from barley analysis with useful accuracy, but process variation sets a hard ceiling. Here is what works.
-
Quality-Control Charts for Brewing in Tableau
Build statistical process control charts for brewing QC in Tableau using reference bands, control limits and Explain Data for out-of-control points.
-
AI for Barley Variety Selection: Spotting Malting Winners Earlier
AI can prioritise which barley lines deserve costly malting trials, shortening a decade-long programme, if you respect genotype-by-environment limits.
-
A Live Fermentation-Monitoring Dashboard in Tableau
Build a live fermentation-monitoring dashboard in Tableau with gravity, temperature and CO2 curves per tank, target bands and a TabPy ML forecast.
-
Optimising Brewing Water Chemistry With AI
AI as a salt and acid dosing adviser can hit target mash pH and sulphate-to-chloride balance across drifting source water and many recipes.
-
Visualising Packaging-Line OEE in Tableau
Build a packaging-line OEE dashboard in Tableau with Availability, Performance and Quality as calculated fields, a downtime Pareto and line filters.
-
Predicting Hop Bitterness (IBU) With Machine Learning
Machine learning predicts IBU and hop utilisation from lot alpha-acid assay and boil conditions, compensating for hop lot drift across crop years.
-
AI for Hop Aroma Profiling and Smart Substitution
AI clusters hop lots by GC oil chemistry to rank substitutes by aroma, de-risk inventory, and warn when no near neighbour exists for a short variety.
-
Using AI to Optimise Adjunct and Cereal Costs
How AI optimises the brewing grist bill across malt, adjuncts and syrups against cost, target extract and FAN constraints as commodity prices move.
-
Computer Vision for Grain Intake and Malt QC
How computer vision grades kernel size, spots damaged or foreign grains and reads colour at intake, faster and more consistently than manual sieving.
-
Predicting Malt Extract and Diastatic Power Before You Brew
Predict hot-water extract and diastatic power from routine malt analysis so brewers adjust grist and mash before a brew, not after a yield miss.
-
Can AI Predict Mash Efficiency and Extract Yield?
A model can predict mash efficiency and kettle extract from crush, mash thickness, temperature profile, sparge and malt analysis, flagging low-yield brews early.
-
AI-Optimised Mash Temperature Profiles
How AI optimises saccharification rest temperatures and times to hit a target fermentability and body consistently across varying malt lots.
-
Predicting Lauter and Wort Separation Performance
Predict run-off time and stuck-mash risk, then tune crush, rake depth, and flow from grist and mash data to protect brewhouse turns and extract.
-
Cutting Wort Boiling Energy With AI
Use AI and data to optimise boil vigour, time, and evaporation, cutting the brewhouse's biggest energy load while still stripping DMS and hitting hop utilisation.
-
Predicting Wort Fermentability From the Sugar Spectrum
Predict apparent attenuation limit from mash schedule and malt, then dial the sugar spectrum to target dryness versus body before fermentation begins.
-
AI for Hot Break, Cold Break, and Wort Clarity
Model how boil, whirlpool, and cooling conditions drive hot and cold break, wort clarity, colloidal stability, and fermentation health using turbidity data.
-
Brewhouse Yield: Using Analytics to Find Where Extract Disappears
Use mass-balance accounting and analytics to attribute the gap between theoretical and actual extract across milling, lauter, trub, transfer, and evaporation.
-
Predicting Yeast Viability, Vitality, and Pitching Rate
How to model yeast viability and vitality from lab and process data to recommend the right pitching rate and decide when to re-propagate.
-
Using AI to Time the Diacetyl Rest
Predict when to start and end the diacetyl rest to drive VDKs below 0.10 mg/L without wasting tank time, using gravity, temperature and yeast state.
-
Forecasting Ester and Higher-Alcohol Formation in Fermentation
Model how temperature, pitching rate, wort oxygen, gravity and FAN shape esters and fusel alcohols, so you tune fermentation to a flavour target.
-
Data-Driven Yeast Generation Management and Cropping
Track yeast performance across serial generations to decide when to re-propagate and how to crop, balancing cost against drift in attenuation and flavour.
-
A Sensory-Panel Results Dashboard in Tableau
Build a sensory-panel dashboard in Tableau: attribute scores, panellist agreement and variance, trueness-to-type radar charts and Pulse drift digests.
-
Detecting Wild Yeast and Bacterial Contamination Early With ML
Combine fermentation-signal ML with rapid micro methods like PCR and ATP-bioluminescence to flag Lacto, Pedio and wild-yeast contamination before it spoils a batch.
-
Predicting Stuck Fermentation Before It Happens
How machine learning predicts stuck fermentation risk from pitching rate, viability, wort oxygen and the early gravity curve, in time to act.
-
AI for Fermentation Temperature and Cooling Control
Predictive control tracks a target fermentation temperature curve and pre-empts the exothermic peak, protecting flavour and saving cooling energy.
-
CO2 Evolution as a Real-Time Fermentation Signal
Inline CO2 and density sensors turn fermentation into a continuous signal, replacing twice-daily gravity samples and feeding models that forecast the curve.
-
Predicting Beer Flavour Stability and Staling
Model beer shelf life and staling from oxygen pickup, storage temperature and forced-ageing data to set realistic best-before dates and oxygen targets.
-
AI for Filtration and Centrifugation Optimisation
Optimise centrifuge flow and filter dosing to balance clarity against yield loss, oxygen pickup and filter-aid cost and predict filter run length.
-
Controlling Dissolved Oxygen in Beer With Machine Learning
How machine learning pinpoints dissolved oxygen pickup and predicts package TPO, so you control the biggest driver of beer staling at the source.
-
Predicting Chill Haze and Colloidal Stability
How models predict chill haze risk and the right silica gel or PVPP dose, balancing clarity, colloidal stability, foam, and cost in beer.
-
AI-Assisted Sensory Panels and Taster Calibration
How data science screens and calibrates beer tasting panellists for sensitivity, drift, and bias, and where AI supports but never replaces the panel.
-
Digitising Beer Tasting Panels: Power Apps, Power BI, and AI
How breweries can digitise tasting panels with a Power Apps capture app, Power BI analytics, and AI fault and drift detection, without losing the palate.
-
Predicting Packaging Line Downtime and Lifting OEE
How machine learning predicts micro-stops on the filler and seamer bottleneck to pre-empt the downtime that dominates packaging line OEE losses.
-
AI Vision for Fill-Level and Label Inspection
How machine vision inspects fill height, label skew, and date-code legibility at line speed, rejecting beer defects more consistently than human spot checks.
-
Minimising Total Package Oxygen (TPO) With Data
Use data and ML to model total package oxygen from filler settings, hit tens-of-ppb TPO targets, and protect beer flavour and shelf life.
-
Predictive Maintenance for Fillers and Seamers
Predict filler-valve and seamer-roll wear from vibration, current, and double-seam data to stop leakers and downtime on the packaging-line bottleneck.
-
Reducing Packaging Material Waste With AI
Use AI and checkweigher data to cut overfill, giveaway, rejects, and changeover scrap, trimming packaging COGS and carbon without breaking the strength floor.
-
AI-Optimised CIP: Cleaning With Less Water, Chemical, and Time
Optimise CIP using the TACT levers and conductivity and turbidity feedback to stop over-cleaning, saving water, caustic, and time without risking spoilage.
-
Predicting Microbiological Risk and Hygiene Hotspots
Model where spoilage risk concentrates using ATP swabs, micro results, CIP records, and HACCP control points to target cleaning and sampling where it matters.
-
AI for Brewery Energy and Utilities (Steam, Refrigeration, CO2)
How AI optimises brewery energy (steam, refrigeration, compressed air and CO2 recovery) by forecasting demand, shifting load and spotting drift.
-
Cutting Process Water and Effluent Load With AI
How AI models brewery water use and effluent strength (BOD/COD) to cut water-to-beer ratio and trade-effluent charges through reuse and CIP tuning.
-
Predictive Maintenance for Brewery Process Equipment
How AI predicts failures in brewery pumps, motors, compressors, refrigeration and heat exchangers from vibration and current data before a brew is disrupted.
-
AI + NIR Spectroscopy for Rapid Brewery QC
How NIR/FTIR spectroscopy plus chemometrics predicts alcohol, extract, pH, bitterness and colour in seconds for fast in-process brewery quality control.
-
Building a Brewery Data Foundation Before You Touch AI
Why a process historian, consistent tags, a data model and data quality are the prerequisites for brewery AI and why most projects fail on plumbing, not algorithms.
-
Building a Digital Tasting Program: Power Platform, ERP, and Gen AI
A practical sequence for beer, wine, and spirits: data model first, capture in Power Apps, anchor in ERP, analyse in Power BI, then layer AI and generative AI.
-
How a Brewery Should Start with AI and Gen AI: The Phases
A phased roadmap for a brewery adopting AI and generative AI (from collecting data to dashboards, predictive models, GenAI copilots and agents) with what to do, need and watch at each phase.
-
Databricks for Breweries: 20 Use Cases
How a brewery uses Databricks end to end (ingestion, real-time monitoring, the Delta Lakehouse & Spark, BI and AI) across 20 concrete use cases grouped by capability.
-
Databricks Across the Brewery Business, Vertical by Vertical
A department-by-department tour of where Databricks helps a brewery (from the floor through quality, supply chain, sales, marketing, finance and compliance) on one governed platform.
-
Claude AI and Claude Code for Breweries: Where the Anthropic Ecosystem Helps
A practical tour of where Claude, Claude Code, the API, agents and MCP help a brewery (recipe, production, QC, supply chain, sales, marketing, compliance and knowledge) and where to keep a human in the loop.
-
Snowflake for Breweries: 20 Use Cases
How a brewery uses Snowflake end to end (ingestion, real-time monitoring, Dynamic Tables & Snowpark, BI and AI) across 20 concrete use cases grouped by capability.
-
Snowflake Across the Brewery Business, Vertical by Vertical
A department-by-department tour of where Snowflake helps a brewery (from the floor through quality, supply chain, sales, marketing, finance and compliance) on one governed platform.
-
Microsoft Fabric for Breweries: 20 Use Cases (and 3 Case Studies)
How a brewery uses Microsoft Fabric end to end (OneLake, Data Factory, Real-Time Intelligence, Lakehouse, Direct Lake and Copilot) across 20 concrete use cases plus three worked case studies.
-
Co-Packing and Contract Brewing Analytics: Making Capacity Pay
Contract brewing and co-packing live or die on capacity utilisation and true cost per contract. The analytics that decide whether each customer brand actually makes money and where AI helps.
-
Yeast Pitching Rates and Starters in Excel
Size your yeast properly in a spreadsheet: cells required by gravity and volume, viability decay by pack age, packs needed, and a simple starter-growth model with worked examples.
-
Predicting When Malt Is Modified Enough to Kiln
Machine learning can help call the kiln-off point by predicting endosperm modification and homogeneity from germination data, but the confirming lab tests lag and homogeneity hides trouble.
-
Predicting Slow Runoff From the Malt Before It Reaches Your Mill
Under-modified malt hides the high beta-glucan that slows lautering and filtration. How a model reads the malt certificate to flag runoff risk and why a lot average can still fool you.
-
Mash Water and Temperature Calculator in Excel
Work out strike temperature, mash and sparge water volumes, grain absorption and total water needed in one Excel sheet, with a water balance that shows where every litre goes.
-
IoT in the Brewery: Sensors at Every Step, from Mash to Package
An IBD-grounded guide to IoT in brewing, which sensors belong at mashing, lautering, the boil, fermentation and packaging, the edge-to-cloud stack, and the AI that reads the streams.
-
Build an IBU Recipe Builder in Excel (Tinseth, Multi-Addition)
A multi-addition IBU calculator in Excel: Tinseth utilisation per hop addition, total bitterness, the BU:GU balance ratio, hop substitution and scaling to a target IBU.
-
A Generative-AI Copilot for Reading Malt Certificates of Analysis
A grounded LLM copilot can read a malt certificate of analysis, explain every parameter, flag out-of-spec values against your recipe and draft the brew sheet, if you stop it inventing numbers.
-
Forecasting Mycotoxin and Gushing Risk in the Malthouse
Fusarium-infected barley brings DON mycotoxin and gushing risk. How machine learning scores incoming lots from harvest weather and grain data and why a food-safety call keeps a human in the loop.
-
A Carbonation Calculator in Excel: Priming Sugar, PSI and Line Balancing
Carbonate beer with confidence in Excel: priming sugar from residual CO2, the keg pressure for any target volume and temperature, and a simple draught line-length balance.
-
Build a Brewing Water Chemistry Calculator in Excel
A step-by-step Excel water calculator for brewers: enter your source profile, add salts, and read out calcium, sulfate, chloride, residual alkalinity and the sulfate-to-chloride ratio.
-
Brewhouse Efficiency and Yield Reconciliation in Excel
Separate conversion, lauter and brewhouse efficiency in one Excel sheet, reconcile predicted versus actual gravity, and diagnose exactly where your extract is being lost.
-
Can AI Dial In the Perfect Steep for Malting Barley?
How machine learning predicts steep-out moisture and germination vigour from barley lot traits and where a steep model still needs a maltster's judgement.
-
AI-Optimised Kilning: Hitting Colour and Diastatic Power While Cutting Gas
Kilning is malting's biggest energy cost and a hard trade-off between enzyme survival, colour and gas use. How machine learning optimises the kiln curve and the physics it cannot cheat.
-
20 Advanced Brewing Calculations You Can Do in Excel (Formulas Included)
Strike temperature, IBU, ABV, pitching rate, carbonation, colour, efficiency and more: 20 advanced brewing calculations as ready-to-paste Excel formulas with worked examples.
-
Crunching Recipe and Sensory Data: How I Sped Up Beer Development
Developing a new beer means closing the gap between a target flavour and what comes out of the pilot kettle. Here's how AI and sensory analytics cut the number of trial brews it took me to get there.
-
Predicting Beer Flavour by Stacking Malt Flavour Wheels
Give each malt its flavour wheel, scale it by its share of the grist, sum the wheels and ground the result in the COA. A weighted-sum fingerprint of beer flavour and the perception physics it cannot fake.
-
From Pilot to Full Scale: The NPD Data Problem Nobody Warns You About
A beer that's perfect on the pilot kettle can drift off-spec at full scale. Here's how I used batch data, control charts and AI to close the scale-up gap when developing beers for AB InBev, SABMiller and United Breweries.
-
Can Claude Opus 4.8 Help You Brew a Better IPA? A Hop-Forward Workflow
A practical, honest use case: how I use Claude Opus 4.8 as a hop and IPA development copilot (from hop selection and dry-hop timing to biotransformation and hop creep) and exactly where it must hand back to the lab and the palate.
-
How to Tell a Good AI Answer from a Confidently Wrong One
AI states hop specs, salt additions and yeast ranges with equal confidence whether right or wrong. A brewer's verification toolkit: what to always check, what to trust provisionally, and the tells that an answer is fabricated.
-
Ten Prompts from the Brewery Floor: Bad Ask → Good Ask
Ten real brewery situations (stuck fermentation, recipe scaling, costing, off-flavours), each shown as the vague prompt beginners type and the specific rewrite that actually gets a useful AI answer.
-
A Brewer's Map of Their Own Data — and What Each Source Can Actually Answer
Brew logs, fermentation sensors, lab results, POS and inventory: a plain map of the data a small brewery already has, and the specific questions each source can (and cannot) answer.
-
The Beginner Brewer's Cheat Code: Ask AI What You Should Be Asking
Flip the use of AI: instead of asking for answers you can't yet judge, ask it to generate the questions your brewing experience hasn't taught you, a safe, powerful move for new brewers.
-
You Don't Have an AI Problem, You Have a Question Problem
Why new brewers get generic mush from AI: the missing skill isn't prompt engineering, it's question literacy, knowing what is knowable about your beer and your process.
-
A Brewer's Step-by-Step Roadmap into AI (with Free Resources)
A practical, five-stage learning path for a working brewer who wants to get into AI (data literacy, spreadsheets to Power BI, Python, machine learning and generative AI) built entirely on free, open resources like Microsoft Learn, Kaggle and Google's PAIR guidebook.
-
Foundations First: Data Literacy and Python for Brewers (Free Resources)
Stages 1 and 2 of the brewer's AI roadmap (data literacy with spreadsheets and Power BI, then Python) using only free, open resources: Microsoft Learn, Kaggle Learn and freeCodeCamp, with a real brewing project to learn against.
-
Your First Machine-Learning Model on Brewing Data (Free Courses)
Stage 3 of the brewer's AI roadmap covers the machine-learning concepts that matter, then building a first real model on your own brewery data, using free resources: Google's Machine Learning Crash Course, Kaggle Learn, Elements of AI and scikit-learn.
-
Generative and Human-Centred AI for Brewers (with Google's PAIR Guidebook)
The final stage of the brewer's AI roadmap: using generative AI well and building with it responsibly, grounded in Google's PAIR People + AI Guidebook and Microsoft Learn's responsible-AI resources, so what you make is something a brewer would actually trust.
-
Vibe-Coding a Brewing Tool with Claude Fable 5
Claude's latest model, Fable 5, makes it realistic for a brewer with no software background to build a small, genuinely useful brewery tool by describing what they want. Here's the honest loop, what Fable 5 changes, and where it still needs you.
-
What Claude Fable 5 Changes for Brewers Who Build
Anthropic's latest model, Claude Fable 5, doesn't replace the brewer's judgement: it lowers the skill floor for building. What actually changes for a brewer making their own tools, what doesn't, and how to use it without fooling yourself.
-
Vibe-Coding a Fermentation Tracker: A Brewer's Build Log
A concrete, step-by-step build log of vibe-coding a small fermentation tracker with Claude Fable 5: the actual prompts, the verify steps, the bug it introduced and how it got caught, so you can copy the loop for your own brewery tool.
-
Beyond the Control Chart: Data Visualizations Brewers Underuse
Most brewery dashboards stop at the line chart and the control chart. This opens a series on the visualizations brewers rarely reach for (radar brand-profiles, CUSUM drift charts, shelf-life trajectories and Sankey balances) and how to match the chart to the brewing question.
-
True-to-Target: Visualizing Whether a Batch Matches Its Brand Profile
True-to-target (TTT) tasting asks a different question than 'is it in spec': does this batch still taste like the brand? How to visualize brand conformance with radar profiles, target bands and deviation tracking, so a panel's verdict becomes a picture the whole brewery reads.
-
SPC That Actually Catches Drift: Capability, CUSUM and EWMA for Brewers
The Shewhart control chart catches sudden spikes but is blind to slow drift: exactly how brewing quality usually fails. How process capability (Cp/Cpk), CUSUM and EWMA charts visualize the slow creep a basic control chart misses, in brewing terms.
-
Shelf-Life Trajectories: Visualizing Forced-Ageing and Flavour Stability
A beer's flavour stability is a story over time, not a single number, yet most breweries record it as a pass/fail. How to visualize forced-ageing and shelf-life trials as trajectories and small multiples, so staling becomes a path you can see, compare and predict.
-
Where Your Extract, Water and Energy Go: Sankey and the Brewhouse Balance
A brewery's biggest losses hide in plain sight across a dozen separate reports. A Sankey diagram puts the whole mass and energy balance in one picture (extract from grain to glass, water in to effluent out, energy in to heat lost) so the leaks become impossible to miss.
-
What AI in Beer Actually Looks Like in 2026
Beyond the headlines about AI-designed beer: what artificial intelligence is genuinely doing in breweries in 2026 (recipe ideation, batch consistency, a brewers' chatbot, and the supply chain) and where the hype quietly breaks.
-
AI in the Brewery Lab: What's Real and What's Hype
A map of where AI genuinely fits in the brewery QC lab in 2026 (soft sensors, machine vision, contamination-risk triage, and predictive calibration) and where a confident model quietly breaks. The opening of a 6-part series that walks the lab bench by bench.
-
On the Malt Bench: Vision, NIR and Incoming QC
How machine vision reads acrospire growth, malt colour and foreign material on the raw-materials bench, and how NIR soft sensors predict extract and moisture before the mash, with the honest caveats.
-
Soft Sensors in the Beer-Chemistry Lab
Part 3 of the brewery-lab AI series: diacetyl and VDK forecasting from the fermentation curve, dissolved-oxygen anomaly detection at the filler, and shelf-life survival models, with the reference method still the arbiter.
-
From Plate Counts to Risk Triage in Microbiology
How AI earns its keep in brewery micro: contamination-risk triage that ranks sample points across propagation, packaging, water and CIP, plus vision colony counting. Honest about where it breaks.
-
Machine Vision on the Packaging Line
Inline machine vision on the brewery packaging line: fill-level checks with SPC to catch drift, crown-seal and label-defect inspection, washed-bottle grading, and pasteurisation-unit modelling from tunnel temperatures.
-
AI in Sensory and the QA System
The final part of the brewery-lab AI series: taster notes as data via NLP, a predictive freshness index, predictive calibration from instrument drift, and deviation and out-of-spec triage. Honest about where a confident model quietly misleads.
-
RAG Over the Brewing Literature Without Hallucinated Maths
Part 1 of The Brewer's Agent. Retrieval-augmented generation is good at finding what the brewing literature says and bad at doing what it says. How to chunk formulas and tables, why brewing needs hybrid search, why every answer needs a citation, and why the arithmetic goes to a tool.
-
An Agent That Calls Calculation Tools Instead of Doing Sums
Part 2 of The Brewer's Agent. Give a brewing assistant tools for gravity, attenuation, ABV and IBU, and make it name the method every time. Worked examples show why: the same beer is 64.9% attenuated on one scale and 78.6% on another, and three IBU models give 34, 45 and 46 for one hop addition.
-
Malt Aroma Wheels as Vectors: Similarity Search and the Blend That Averaging Hides
Part 3 of The Brewer's Agent. Turn published malt aroma wheels into 22-number vectors and you can search for substitutes and predict grist character. Real numbers from 51 digitised Weyermann wheels show the traps: raw cosine says every malt is alike, variants collapse onto one vector, and a plain average hides 10% of roasted malt.
-
Scraping and Trend-Mining 250 Public Homebrew Recipes: Regex Where You Can, LLM Where You Must
Part 4 of The Brewer's Agent. A small data engineering project on 250 recently shared homebrew recipes: polite scraping, the parsing traps, normalising to base units, when regex beats an LLM and when it does not, a consistency check the data failed 12 times, and what the recipes actually say about 2025 and 2026 brewing.
-
Evals for a Brewing Assistant: Golden Questions, Numeric Tolerances and the Limits of LLM-as-Judge
Part 5 of The Brewer's Agent. How to test a GenAI brewing assistant before a brewer relies on it: a golden question set written by brewers, numbers scored against tolerances with units checked, citations and refusals tested on purpose, LLM-as-judge only for prose, and a regression run on every change.
-
What 1,000 Beverage-AI Companies Are Actually Building
Part 6 of The Brewer's Agent. A dataset of just over 1,000 companies and people applying AI to beer, wine and whiskey, counted properly. Almost as many make no AI claim as ship AI, the shipping work is operational rather than showy, sensory and recipe AI stays in pilot, equipment makers barely use it, and genuine GenAI is a small slice.
-
What Is AI, Really? The Ladder From Rules to AGI, Told on a Beverage Plant Floor
Part 1 of AI for Operational Excellence in Beverage Plants. Rules, machine learning, deep learning, generative AI, agentic AI and AGI, one rung at a time, each with a real example from a brewery, winery, distillery or packaging line. Honest about AGI: it does not exist yet, and Monday's shift does not need it.
-
What GenAI and LLMs Actually Are: Tokens, Context, Hallucination and RAG, Explained Through a CIP Log
Part 2 of AI for Operational Excellence in Beverage Plants. How a large language model really works, from tokens and the context window to next-token prediction, why it hallucinates, when to use retrieval (RAG) versus fine-tuning, and what multimodal models add, all worked through one CIP log and one SOP.
-
What Agentic AI Is: Tools, the Plan-Act-Observe Loop, MCP and Five Levels of Autonomy
Part 3 of AI for Operational Excellence in Beverage Plants. An agent is a language model that can use tools and take several steps towards a goal. How the loop works, what tools and MCP are, what memory means, and a five-level autonomy ladder (L0 to L4) for deciding how much a plant should let an agent do on its own.
-
Operational Excellence Basics for Data People: OEE, the Six Big Losses, SPC and Where Plant Data Lives
Part 4 of AI for Operational Excellence in Beverage Plants. Before any AI, the language of the plant: OEE with a worked filler-line example, the six big losses, TPM, the lean wastes, statistical process control, and the ISA-95 map of where the data actually sits (historian, MES, CMMS, LIMS, ERP).
-
Classic AI for OpEx: Predictive Maintenance, Soft Sensors and Anomaly Detection on Top of SPC
Part 5 of AI for Operational Excellence in Beverage Plants. The machine learning that already pays in beverage plants: predictive maintenance on fillers and pasteurisers, soft sensors for what you cannot measure in line, multivariate anomaly detection layered on SPC, and energy and water per hectolitre against a fair baseline.
-
GenAI for OpEx: Shift Handovers, SOP Search, Root-Cause Assistants and a Multilingual Floor
Part 6 of AI for Operational Excellence in Beverage Plants. Where generative AI earns its place on the plant floor: shift handover summaries built from the MES and operator notes, SOP answers with citations, a five-whys and Ishikawa assistant that proposes causes but never concludes, one-point lessons, and translation for a crew that speaks four languages.
-
Agentic AI for OpEx: An OEE Loss Agent That Drafts Work Orders, and the Guardrails That Keep It Off the PLC
Part 7 of AI for Operational Excellence in Beverage Plants. An agent that watches OEE losses every shift, finds recurring faults, checks the CMMS and drafts work orders and changeover plans for people to approve. Plus the guardrails that matter: the Purdue model, read-only access to OT data, no writes to PLC or SCADA, approvals, an audit log and replay evals.
-
The AI for OpEx Roadmap, and Where It Breaks: Data First, Pilots by Loss Value, Honest ROI and the AGI Hype
Part 8 of AI for Operational Excellence in Beverage Plants. How to sequence AI in a beverage plant: data foundations first, pilots chosen by the value of the loss, ROI measured against a baseline, change management on the floor, and a clear view of what AGI talk should and should not change. Closes the series.
-
What Responsible AI Actually Means for a Drinks Business: Six Principles and a Use-Case Register in SQL
Part 1 of Responsible AI for Beer, Whisky and Wine. The six Responsible AI principles, mapped to breweries, distilleries and wineries, the frameworks behind them (NIST AI RMF, ISO/IEC 42001, the EU AI Act, India's AI Governance Guidelines and the DPDP Act), and the one artefact every drinks business should build first: an AI use-case register with a risk-tiering query.
-
When the Model Touches Food Safety: Flag-Only Release Gates, Abstain Thresholds and Model Cards for Beer, Whisky and Wine
Part 4 of Responsible AI for Beer, Whisky and Wine. Reliability and safety in practice: AI may rank risk and raise flags, but a person releases the batch and validated methods stay on the HACCP critical control points. Worked through soft sensors, mycotoxin and gushing risk in malt, methanol and the spirit cut, and SO2 and Brett in wine, with a model card, an abstain threshold, drift monitoring and a release gate where the model can block but never release.
-
A Responsible AI Playbook for a Small Beverage Business, and Where It Breaks
Part 8 of Responsible AI for Beer, Whisky and Wine. A lightweight version of ISO/IEC 42001 for a craft brewery, small distillery or winery with no AI team: a use-case register, a one-page impact assessment, an eval harness, monitoring, an incident log and a quarterly review, mapped to the NIST AI RMF. Plus AI's own energy and water footprint against your ESG claims, the regulatory calendar as of September 2026, and a 90-day plan.