One trusted picture of your operation, right now.

The core problem we solve

What is it, in one sentence, and who feels it?

Stated plainly, this is what we are hired to fix:

Do we understand the customer

Who buys, who champions, what triggers the conversation?

The conversation usually starts with an integrated operations centre or control-room build-out. It may then lead to control-system modernisation, a maintenance-system upgrade, or a new plant.

The Unified Namespace

What is it, and why is it what we build first?

We already have one. Our live demonstration joins plant control, manufacturing-execution and maintenance data into one live operating picture. It is a unified namespace.

The knowledge graph

What the namespace is the foundation for.

• Sequence is settled. Start with the namespace, then extend it into the graph so that each grows with the other.
• It is the scaffold agents traverse. A language model has limited context. The graph guides it to the relevant node, where it reads the current state. Without a graph, there is no real agentic reasoning across a plant.
• It makes tribal knowledge queryable. What an experienced operator tells an agent is applied to the graph and can answer the next operator's question. Root-cause traces cross silos; asset types can be compared across sites.

How we serve them

What does working with us feel like?

• The senior engineer is the architect and the project manager. There is no hand-off from a sales engineer to a delivery team working from stale knowledge.
• Full-stack fluency in one head. Control, MES, CMMS, ERP, edge and cloud are all running today in our demonstration environment.
• Agile, short iterations. Waterfall assumes we already know what we need to know. We do not, and neither does the client. The second question always changes the first answer.
• Educate, do not stay long. With each phase, we do less and teach more, so the client becomes capable.
• Engineering discipline the client can read. Verify the artefact, not just the absence of an error. A missing reading must never look healthy. Alert, do not log. Every rule was paid for.

Where we play: Industry 4 and Industry 5

What the next two years look like, and where White Horse Clips sits in them.

The line we will not cross

A controller is deterministic: the same input gives the same output every time, within a fixed time window. That is the basis of industrial control, and it is not suited to probabilistic reasoning. A language model is probabilistic by design: the same prompt and context can still take a different path. You can constrain it, and a human in the loop provides a constraint, but you cannot make it deterministic.

So agents live in the decision layer, not the control layer. Our rule of thumb: if a wrong or late answer can hurt someone or break something, it stays deterministic. If a wrong answer that gets caught costs a review cycle, that is where agents earn their place.

What you buy, and what gets built

This is the split we will advise you on. Buy the infrastructure: brokers, historians, modelling engines, data transformation and pipelines. That market is mature, and building your own is waste. Build what sits above it: how your people and agents analyse, visualise, find patterns, report and solve. That is where your operation differs from the one next door.

The problems have not changed: do more with less, do it safer, reduce impact, and get more out of what you already own. These are data-to-decision problems, and they always were. What is changing quickly is how they are solved.

The operator

Who is building this, and what has he already delivered?

J.C du Preez most recently led an OT and automation department of more than forty-eight people across mining and processing operations. Its work covered optimisation, data analytics, monitor and control, and OT governance, including the cybersecurity framework and the enterprise data platform. Before that, he spent nearly nine years at a global mining group, ending as Maintenance Superintendent of a platinum concentrator. He led a thirty-four-person mechanical, electrical and instrumentation team and implemented ISO 55000 asset management across crushers, mills, flotation and 33kV substations. He serves as Deputy Chair of the RTIMS Technical Steering Committee at the Mandela Mining Precinct — a public-private partnership between the Department of Science and Innovation and the Minerals Council of South Africa — leading the IT-OT convergence work package and contributing to the AI framework for mining.

Two results are worth stating because they were measured on a plant rather than modelled on a slide. Volumetric scanning introduced across a 150-truck fleet delivered a 17% load improvement in BCM and more than US$60M a year in revenue. A neural-network haulage digital twin he directed was used to test fleet sizing and dispatch. The automated load-curtailment response he directed — the plant answering a grid curtailment event on its own and protecting production through constrained supply — won the Schneider Electric 2024 Sustainability Impact Award. He led this work as the department head who carried the budget, the team and the accountability for it. These are the Director's own record, not a promise for any other site.

Start with a conversation. Where we begin is an assessment on one part of your operation: whclips.co.za/contact.html

The plant shown in any diagram is an illustrative example, not a client.