Skip to content
Discrete and process manufacturers, and the planners behind them

Manufacturing & Supply Chain

Plan reconciliation, deviation triage, supplier document intake, maintenance scheduling, shop-floor capture and traceability — built so the system assembles and a planner or engineer decides.

The data already exists. It is in the ERP, the MES, the quality system and three spreadsheets, and none of them agree. The work is reconciling them fast enough to matter, because most planning decisions are made on figures nobody can trace.

Workflows

What we automate, and where it stops

Every one of these ends one step short of the decision. In a regulated industry that is not caution, it is the architecture — and it is what makes the work approvable.

Demand and supply plan reconciliation

  1. ERP + MES + spreadsheets
  2. Reconcile
  3. Variance queue with sources

A planner commits the plan. The system reconciles; it never commits.

Continuous matching across the systems that each believe they hold the truth, with every disagreement surfaced as a ranked variance carrying both figures and where each came from.

Built with Platform & Data

Quality deviation and non-conformance triage

  1. Deviation
  2. Gather batch evidence
  3. Ranked queue with rationale

A quality engineer dispositions every deviation. None is auto-closed.

Each deviation arrives with its batch record, process data and related history already assembled, so the engineer spends the time on judgement rather than on gathering.

Built with Agentic Systems & Automation

Supplier document and certificate intake

  1. PDFs and portals
  2. Extract
  3. Typed records, gaps flagged

Refused fields go to a buyer. A certificate is never assumed valid.

Certificates of analysis, conformance documents and supplier declarations turned into typed records carrying the page each value came from. Fields the model is not confident about are refused rather than guessed.

Built with AI & Applied ML

Condition monitoring and maintenance scheduling

  1. Sensor streams
  2. Detect drift
  3. Ranked work orders

A maintenance planner schedules. Nothing takes a line down on its own.

Drift detected against each asset’s own baseline rather than a generic threshold, with the evidence for every alert attached to the work order it generates.

Built with Delivery Telemetry & Observability

Shop-floor data capture

  1. Machine signals
  2. Normalise
  3. Typed events with provenance

An unreadable signal reports as unknown. It is never interpolated into a reading.

Signals from mixed-vintage equipment normalised into typed events, with the gaps left visible. A sensor that stopped reporting is recorded as having stopped reporting.

Built with Robotics & Embedded

Genealogy and recall traceability

  1. Batch records
  2. Link
  3. Trace with an audit trail

A quality lead declares the scope of any recall. The trail is the deliverable.

Forward and backward genealogy assembled on demand across batches, components and shipments, with the derivation of each link recorded.

Built with Platform & Data

Constraints

What is actually hard here

There are no client names, logos or percentages on this page. Until there is a real figure with permission to publish it, the space is better spent on the difficulties we can defend.

Nothing stops the line

Every workflow stops short of the operational decision, with a named role at the stop. A system that can halt production is a system that will, at the worst possible moment, halt production.

Read the long version

An unknown reading is not a zero

A sensor that went quiet and a sensor reading zero render identically once both are the number 0, and only one of them is true. Carrying the difference through the pipeline is unglamorous and it is most of the value.

A pilot on clean data proves nothing

The demonstration runs on the tidy extract. Production runs on fifteen years of accumulated exceptions, renamed parts and plants that do it differently. The gap between those is the project.

Read the long version

Batch size is set by the changeover, not the software

Shortening a release cycle in a plant whose economics are governed by setup cost changes nothing. The constraint has to be found before it can be moved.

Read the long version

Verification is the work

Generating a draft is cheap and checking it is not. Capacity planning that assumes otherwise produces a pile of unreviewed output and calls it throughput.

Read the long version

Bring us the workflow, not the brief.

Describe what your team does by hand today and where it backs up. We will tell you which part is assembly, which part is judgement, and which of the two we would touch.