SteelWorks Intelligence Systems · a Luma Crest Holdings company

We solve operational problems at the root, with systems built like a business.

SteelWorks designs and operates structural AI systems: governance and decision rights, cross-functional teams, specialized agents with a scope they cannot exceed, and the strongest model chosen for each task on measured score. Not a chatbot, not a pile of automations — an operation with a chart, a security function, and an owner for every decision.

Built for energy, oil & gas, distribution, security, financial, healthcare, retail and marketing. Based in Acworth, Georgia; delivered remotely.

Why structure

Capability without accountability is just faster mistakes.

Most AI programmes stall in the same place. A tool gets adopted, it does something useful, and then nobody can say who owns the output, what it was allowed to touch, or why it chose what it chose. The work has no home, so it does not scale and it cannot be audited.

A structural system fixes that before it does anything clever. Units get names and decision rights. Approval gates go where a person should decide. Every agent carries one job, a tool scope, and a report line. Then — and only then — you point the strongest model at the work.

The four layers

The four layers of a structural AI system, governance, teams, agents and models, with security running through all four
Governance, teams, agents, models — and security through all of them.
The SteelWorks org chart with command, control and assurance, delivery, revenue and capability tiers and agent counts per unit
Our own org record. The first deliverable of every engagement is a chart like it.

Hierarchy

The org chart is the deliverable.

Before an agent is hired, the chart exists: which unit owns which operation, what it decides without asking, what it must escalate, and what must never be decided automatically. Agents are then recruited into it through a real pipeline — role card, training, examination, placement, certification — because a specialist with no defined post is just a script with a personality.

Clients keep that chart whether or not they keep us. It is a written account of how their operation actually makes decisions, which most organisations have never had.

Niche markets

Eight markets, one method

We build where operational depth meets regulatory weight — the conditions under which a general assistant stops helping and a structured system starts paying for itself. Each vertical needs a domain corpus, an evaluation set scored in that domain, the regulatory envelope that shapes its gates, and a reference org chart.

Eight niche markets against four requirements, with security marked built and the remaining markets method-ready
Security is the vertical we built on ourselves. Everywhere else we say method-ready, not delivered.

What each market's envelope adds

Run on ourselves first

SteelWorks is the first system SteelWorks built.

The chart above is ours. The controls are ours. The routing table is ours. Every number on this site is read from live system state when the site deploys, which is the same standard we hold client reporting to.

61

agents, each with a written role and a tool scope

552

scheduled operations, no hour of the day empty

5,878

graded turns behind the routing decisions

25

security controls measured weekly and published

How it starts

Assessments

Every engagement opens with a scored report you keep whether or not you go further. The AI Security & Risk Audit covers an AI operation you already run. The Digital Intelligence Assessment covers the whole operation and ends in a prioritised plan with the price of each step.

AI Security & Risk Audit
$1,500–$3,000
Digital Intelligence Assessment
scoped, priced first
output
scored report + ranked fixes

Compare the assessments

What follows

Engagements

A fixed-scope build puts a governed slice of the system on your hardware: three to five operations, approval gates, reporting, and the documentation to run it without us. Ongoing care and an Intelligence Partner engagement exist for the operations that keep growing.

AI Operations Build
$3,500–$8,000
Operations care
$1,500–$3,000/mo
hardware
yours, not ours

See the engagements

Straight answers

What is a structural AI system?

An AI operation built like a business: named units with decision rights, cross-functional teams assembled per operation, specialized agents with tool scopes they cannot exceed, and the strongest model selected per task on measured score — with security running through all four layers rather than bolted to one.

Do you build proprietary LLMs?

We build proprietary systems that select, govern and measure the strongest available model for each task, across thousands of graded turns. Vertical model programs are on the roadmap. We will say so when one exists and is measured, not before.

Which markets do you serve?

Energy, oil and gas, distribution, security, financial, healthcare, retail and marketing. The method works anywhere the operation is complex enough to need an org chart.

What we will not do

Claim a model we did not train or a result we did not measure.

How does an engagement start?

With an assessment that ends in a scored report you keep. No engagement begins with a twelve-month transformation plan before anything runs.

Is the security audit a certification?

No. It scores your AI operation against the NIST AI Risk Management Framework and hands you owned, dated fixes. It is evidence for your own governance, not a certificate, and it does not guarantee that nothing can go wrong.

Who else is under the group?

SteelWorks sits under Luma Crest Holdings alongside SteelClips, SteelTrades and SteelProducts. Keeping media, trading and products in their own companies is what lets this one stay pointed at operational systems.


Start here

Tell us what breaks, and we will tell you what it would take.

Leave your company and the operation that costs you the most time. You get a written reply with the shape of the system that would carry it — and an honest note if we think you do not need one.