Our approach to industrial AI

Understand the answer. See what it rests on.

We combine engineering knowledge, data analysis and AI to support a real operating decision. The evidence, assumptions and controls belong in the user’s view.

Explore an example See the work

The Propel glass box

Trust comes from
being able to check.

An engineer needs to inspect the measurements behind an alert. An environmental adviser needs the method behind a figure. A knowledge user needs the passage and version supporting an answer.

Our glass-box approach makes those links visible. Users can challenge assumptions, review uncertainty and decide when a specialist needs to take over.

Visual information that leads to action

Start with what
needs attention.

Explore an item, then inspect its source, checks and next action. Dashboards and visual views help people see the situation before they need to ask a question.

From overview to evidence

Items for attention

Example / Pump P-204

A wear trend needs review

Engineering review

Sample readings are shown with units, test methods and the equipment’s history. The analyst can open the source record.

Read the relevant study

A broader view of AI

Choose the method
that fits the decision.

Language models are one part of the toolkit. We use them alongside established engineering methods, learning from data and clear visual interfaces.

Engineering models & optimisation

Calculate energy balances, compare schedules and respect equipment limits. Numerical methods own the calculation.

See an example

Machine learning & forecasting

Learn useful patterns in consumption and condition, test on unseen periods and make uncertainty visible.

See an example

Generative AI

Interpret variable documents and explain supported findings, with references and review where the task needs them.

See an example

Agentic AI & controlled workflows

Coordinate retrieval, checks and follow-up steps within agreed permissions and review points. Keep the actions and their owners visible.

See an example

Decision support & visualisation

Turn connected data into timelines, comparisons, exception views and work queues that help people act.

See an example

Your data and your operation

Agree the boundaries.
Build within them.

Your data remains yours. Hosting, residency, access, use and retention are agreed as part of the design.

Connect the systems you already use.

Work with approved interfaces and controlled imports. Retain source identity and make data quality and missing context visible.

Fit the deployment to the requirement.

Agree which services process the data and where. Private environments, local deployment and offline operation can be considered against the application’s needs.