AutomatonAgent

Case Studies

Case studies

Five deployments across four industries. Customer names are withheld at their request; references are available under NDA.

Textiles · Fabric production

Fabric Hand-Feel & Finishing Copilot

Customer: A woven and knitted fabric manufacturer

Captured senior experts’ hand-feel judgment in an explainable model that recommends finishing recipes for a target feel.

Challenge

  • Fabric hand-feel was judged by touch by a small group of senior experts.
  • Finishing recipes (stenter temperature, speed, softener dosage, overfeed) were tuned by repeated sampling.
  • Lot-to-lot color deviation was often discovered only at final inspection.
  • Process know-how was at risk as experienced staff retired.

What the agent does

  • Predicts tactile and visual properties from finishing parameters.
  • Explains which parameters drive each property, in language engineers trust.
  • Reverse-optimizes a recipe for a target hand-feel within process limits.
  • Warns of ΔE color deviation before shipment and answers questions from SOPs and past lots.

Data connected

  • Finishing parameters
  • Expert sensory ratings
  • Lab / ΔE measurements
  • SOPs & lot records

Outcomes

  • Expert hand-feel judgment became a reusable, explainable digital asset.
  • Fewer sampling trials before reaching the target hand-feel.
  • New engineers receive recipe guidance with reasons, not just numbers.
Food & Beverage · Bottled milk tea

Bottled Milk Tea Batch Consistency Agent

Customer: A producer of ready-to-drink bottled milk tea

Kept taste and texture consistent across batches and lines, and traced every deviation back to raw materials and process steps.

Challenge

  • Taste and texture varied across batches and lines as tea extract and milk ingredients changed.
  • Stratification and sediment could appear during shelf life.
  • Deviation investigations meant manually joining paper records, LIMS data and line data.
  • QA documentation for each deviation took significant effort.

What the agent does

  • Monitors critical-to-quality indicators (Brix, pH, viscosity, fat, color) from blending to filling.
  • Flags drift early with SPC rules across blending, homogenization, sterilization and filling.
  • Links each deviation to raw-material lots and process steps with an evidence chain.
  • Drafts deviation and CAPA reports aligned with HACCP for QA review.

Data connected

  • Raw-material COAs
  • LIMS results
  • Blending & sterilization records
  • Filling line data
  • Shelf-life tests

Outcomes

  • More consistent taste and texture batch to batch.
  • Deviations traced faster with a complete, auditable evidence chain.
  • QA reports prepared automatically, ready for human approval.
Automotive Parts · Injection molding

Injection Molding Process Tuning Agent

Customer: A Tier-2 supplier of automotive interior and functional parts

Turned shot-level machine data into ranked defect causes and guard-railed setpoint recommendations for technicians.

Challenge

  • Short shots, flash, sink marks and warpage recurred after changeovers and resin lot changes.
  • Parameter tuning depended on a few senior technicians.
  • Mold trials and restarts consumed material and machine hours.

What the agent does

  • Reads shot-level parameters: melt and mold temperature, injection speed, holding pressure and time, cooling.
  • Ranks the likely causes of a defect and explains the evidence behind each.
  • Recommends setpoint changes inside the validated process window, for technician approval.
  • Builds a mold-specific knowledge base from every trial and incident.

Data connected

  • Injection machine data
  • Mold temperature controllers
  • MES
  • Inspection results
  • Mold trial records

Outcomes

  • Faster stabilization after changeovers and mold trials.
  • Tuning know-how retained per mold instead of per person.
  • Less scrap from undetected parameter drift.
Automotive Parts · Casting

Die Casting Porosity & Defect Agent

Customer: An aluminum die-casting supplier of automotive housings and structural parts

Linked every shot’s process curve to X-ray and machining results, so defect risk is flagged at the casting cell instead of downstream.

Challenge

  • Porosity, shrinkage and cold shuts were found late, at X-ray or after machining.
  • Many coupled variables: melt temperature, die temperature, shot profile, vacuum, die spray.
  • Root-cause meetings relied on experience rather than shot-level evidence.

What the agent does

  • Correlates each shot curve with X-ray / CT and machining results.
  • Detects abnormal shot profiles and die thermal drift in near real time.
  • Explains the top contributing factors for each defect type.
  • Generates evidence-backed 8D reports.

Data connected

  • Shot curves
  • Die thermal data
  • Melt & degassing records
  • X-ray / CT results
  • Machining scrap

Outcomes

  • Defect risk flagged at the casting cell, before value is added downstream.
  • Root-cause analysis grounded in evidence from every shot.
  • Less time spent assembling 8D documentation.
Telecom Equipment · R&D process

R&D Process Agent

Customer: A telecommunications equipment manufacturer

Automated requirement analysis, design review, test planning and traceability across a complex hardware and software R&D flow.

Challenge

  • Large requirement and specification sets from standards and customers.
  • Manual reviews, traceability and test planning consumed senior engineering time.
  • Lessons from past defects were rarely reused in new projects.
  • Changes rippled across modules with unclear impact.

What the agent does

  • Parses requirements and specifications into structured, traceable items.
  • Checks design documents against review checklists and applicable standards.
  • Generates test cases and maintains requirement–design–test traceability.
  • Analyzes change impact and retrieves similar historical defects and fixes.

Data connected

  • Requirement documents
  • Specs & standards
  • Design documents
  • Test management
  • Defect tracking

Outcomes

  • Shorter review cycles with consistent checklists.
  • Requirement gaps detected earlier in the development flow.
  • Engineering knowledge reused across product generations.

Start with one problem on one line.

Tell us about a quality, process or engineering challenge. We will show you what an agent can do with your data.