Applied intelligence systems: what is happening, what to correct, what to prevent, and what to watch next.
I design applied intelligence systems that help people understand what is happening, determine the corrective action, identify the preventive action, and recognise what needs attention before a problem becomes a larger one.
They work from what an organisation already holds: structured data, evidence, events and rules. The system reads the current position from that record rather than from memory, shows what needs attention now, and keeps the unresolved in view. That last part is not prediction. It is noticing what is already visible and has not yet been dealt with.
The work behind that is unglamorous: architecture, logic and data integrity, so a system stays truthful and maintainable as it grows, and what it reports can still be traced back to what actually happened. I am most interested in AI-assisted and governance-aware systems, where evidence, traceability and long-term consequence carry more weight than throughput.
The objective is not intelligence for its own sake. It is better judgement, better action, and less avoidable damage.
A quality and service excellence platform supporting 14 frameworks, AI-powered analysis, audit readiness, and team collaboration for organisations serious about governance.
A multi-client project and site tracking system for telecom infrastructure, with workflow automation, document management, and real-time analytics.
An independent academic writing platform built on the principle that AI should train cognitive discipline, not replace it. Designed for postgraduate researchers.
A multi-brand retail command centre with real-time sales analytics, demand forecasting, inventory risk detection, and promotion impact simulation. Piloted with Marrybrown.
Two layers sit under it and have no interface of their own: a labour intelligence engine that computes the workforce figures the command centre reads, and a point of sale event layer that emits what happens in a store as it happens.
Issue centred operational communication rather than a chat tool: a problem is raised, an action is taken, a result is recorded, and what was learned stays attached to it. Running with one operations pilot.
A navigator for people building a distribution business: goal, gap, activities, follow ups and reflection. It deliberately does not calculate commission. The comp plan provides the destination; this provides the navigation. In use on one programme.
Care operations for a practice: rostering, practitioner views, care plans and reporting, built around the principle that the software should not interrupt a clinician mid task. Still in development, and not yet running in a practice.
A governance layer and an executable ontology for renewable energy projects, with a verification engine that refuses to let anything into the codebase it has not checked. There is no application and no user interface yet, by design: the domain model is being proved before a line of product is written.
A unified quality platform holding APQP and DMAIC as two first class methodologies rather than variants of one another. The public pages state plainly what is not built, and the phase level tool mapping is deliberately absent until a practitioner defines it.
Controlled documents with an append only history: what the current version is, what it replaced, and who approved the change. Built as a demonstration of the control, not yet as a product.
The system behind the Human and AI Workforce position: AI employees with a role, standing responsibilities, deliverables, delegated work, an approval gate they cannot cross, and the standing to refuse. The platform runs. No employee is hired anywhere yet.
The Personal Quality System put into a form a young person can actually use. Finished and closed to the public, waiting on a first school or youth organisation.
An oil palm plantation operation, organised as estates, divisions and fields, where the record of what happened on the ground begins with a supervisor standing in it.
Three things move through the system: the field supervisor's daily report, carrying water level, piezometer reading, pump run time and pump running status; the fertilizer application record, entered against a programme rather than in isolation; and dated site observation photographs. Those entries become a report history held centrally instead of on one supervisor's phone, and can be read back over a chosen period as pump efficiency and piezometer trend.
It stops at capture and reporting, deliberately. The deeper analysis stays unbuilt until the recording discipline holds and the data earns trust, because an analysis layer laid over thin data does not reveal an operation. It only makes weak data look authoritative. The system carries the operator's own branding rather than mine.
This shapes how systems are designed.
Several of these are not waiting on more building. They are waiting on a domain partner who brings the industry, a practitioner who can settle what I deliberately left open, one organisation willing to run the first real use, or in a single case the field work funded. Each is listed below with what it is actually stuck on.
Needs a practitioner, and the field validation funded
The architecture is deliberately frozen until the model is validated in the field on a solar farm in Kedah. That validation needs somebody who has actually delivered renewable energy projects. Until then it does not proceed, and that is written into the project rather than hoped for. It is the one item here that a person alone does not unblock: being on site in Kedah with a practitioner is a funded piece of field work, not more building.
Needs a quality practitioner
Which of the eighteen tools belongs to which APQP or DMAIC phase is not established, and inventing it would be worse than leaving it open. It needs a working quality practitioner to define it alongside me.
Needs a first practice
Built around care operations and ready to be tested against a real roster and real care plans. It needs one practice willing to run it and tell me where it gets in the way.
Needs a first school or youth group
Finished and waiting. It needs one school, church group or youth organisation prepared to put it in front of young people and see whether it holds their attention.
Needs a first employer
An AI employee with a role and an approval gate is a proposition that can only be tested by employing one. It needs an organisation willing to give an AI employee real standing responsibilities and hold it to account.
Needs recording discipline, not features
Capture and reporting work. The analysis layer stays unbuilt on purpose until the daily recording holds and the data earns trust. What would move it forward is another estate willing to record consistently, not another module.
Most of what is above is blocked on a person or a first user rather than on funding, and it is listed that way on purpose. Estate Intelligence needs recording discipline, and money would not improve it at all.
None of this is offered as finished. It is offered as work in progress that is far enough along to be useful to somebody, and honest enough to say where it stops.
Get in touch →