Cask & Maturation: Angel share and aging
Forecast warehouse aging, cask extraction, angel share loss, and release readiness with auditable recommendations tied to each rack, cask, and target profile.
Predict how every cask moves toward spec.
The agent tracks barrel history, warehouse zone, fill strength, climate exposure, wood program, lab results, and sensory notes. It recommends cask moves, sampling plans, and blend candidates before years of capital drift off target.
- Select the right cask for each new make profile and age statement.
- Model evaporation and proof drift by rack position and warehouse climate.
- Forecast extraction, oxidation, and readiness windows.
Warehouse signals become a maturity forecast.
Caskent fuses climate telemetry, historical fills, strength, color, oak markers, congeners, and tasting notes. The forecast separates expected angel share from avoidable loss, then ranks actions for review.
- Loss curves are compared against cohort baselines.
- Extraction pace is predicted before a sample is pulled.
- Every recommendation keeps recipe and age constraints intact.
Capabilities
Focused controls for the work that determines yield, quality, loss, and release confidence.
Cask selection
Match incoming spirit to wood type, char, fill count, warehouse zone, and target release profile.
Angel share modeling
Predict evaporation and proof drift across humid, dry, hot, and cool rack locations.
Extraction forecast
Estimate oak, color, lactone, tannin, ester, and oxidation trajectory from signals and samples.
Sampling plans
Prioritize casks that carry the highest uncertainty or highest blend value.
Release readiness
Flag casks that are ready, early, or at risk of over-oaking before value is lost.
Grounded in plant evidence.
The agent reads operational signals, lab context, quality data, and business constraints before making a recommendation.
| Signal group | Examples | Status |
|---|---|---|
| Warehouse climate | Temperature, humidity, airflow, seasonal exposure, rack and pallet location. | Used |
| Cask record | Wood source, toast or char, fill count, prior contents, entry proof, fill date. | Used |
| Quality signals | ABV, color, congeners, oak markers, sensory notes, lab samples. | Used |
| Commercial constraints | Age statement, batch size, target profile, inventory plan, release date. | Used |
Governed warehouse actions
Recommendations are not a blind sequence. They are presented as bounded actions with confidence, value at risk, and sampling evidence.
- Step 1CONTROLLED
Move or rotate
Suggest rack movements when climate exposure is pulling a cask away from target.
- Step 2CONTROLLED
Sample or wait
Queue sampling where uncertainty affects blend value, and defer low-signal pulls.
- Step 3CONTROLLED
Reserve or release
Mark casks for single-cask reserve, blend candidate pool, or extended maturation.
Measured with run-level evidence.
Illustrative design-partner targets. Caskent ties each number to source signals and operator decisions.
Bounded autonomy, human approval, complete trace.
Every recommendation is grounded in data, constrained by recipe and safety limits, and written to the audit log with the source signals that produced it. SSO, RBAC, encryption, tenant isolation, and SOC 2 Type I work in progress support enterprise deployment.
Works with the rest of the stack
Caskent agents share evidence through the digital twin and audit layer, so each workflow improves the next one.
- 01Open →
Blend & Proof
Blend plans, dilution, filtration, and target profile fit.
- 02Open →
Robotic Bottling
Fill, cork, label, inspect, pack, and trace the line.
- 03Open →
Energy & Yield
Steam, recovery, spirit yield, feints, and loss optimization.
- 04Open →
Distillation and Maturation Digital Twin
Simulate cut, aging, blend, and release outcomes before action.
We stopped treating the warehouse as a black box. The agent shows which casks need attention and which ones should be left alone.
Put Cask & Maturation under bounded autonomy.
Start with one line, warehouse zone, or production workflow. Prove the metric, review the log, then expand site by site.