AS Enterprise AI Advisory LLC

Positioning

Most AI advisors produce slide decks. This practice produces working systems.

The principal takes a business pain point, designs the solution, builds the working prototype himself using AI coding tools, specifies the stack end to end including hardware and GPU sizing, operating systems, databases, security and data management, and then hands a demonstrated working system to engineers who take it to production. Thirty years of running enterprise infrastructure is what makes those stack decisions credible.

“I set the technical bar by getting to a working thing first, not by describing one.”

An important distinction. He is an architect and a builder of prototypes. He designs solutions, builds prototypes with AI coding tools, specifies the stack end to end, directs the build and leads delivery. He does not write the production code. Engineers do that, working from a specification his prototype has already proven.

Services

The practice works in four areas. No pricing is published; commercial terms are discussed per engagement.

AI strategy and operating model

Value cases, target operating model, governance, risk tiering, adoption approach and roadmaps. These are delivered as interactive multi-scenario models rather than slide decks: configurable scenarios, resourcing, cost and total cost of ownership, use-case specifications with acceptance criteria, and a reference architecture carried through to a procurement-ready bill of quantities. Change an assumption and the resourcing, the cost and the plan move with it.

AI program delivery

Taking use cases from pain point to production, working through teams that do not report to him. The sponsoring department owns the outcome and the engineers own the build. Each use case reaches a demonstrated proof of concept, and the build is then directed to a system that is live, measured and governed, with metering and human-in-the-loop controls in place from the first day.

Technology operations

Infrastructure, availability and continuity, service operations, vendor and contract governance, and the executive metrics that show whether any of it is working. This is the discipline the principal has practised for thirty years, and it is why the stack decisions inside an AI programme come from someone who has been accountable for running systems at scale.

Security and data governance

Cybersecurity operating models, risk registers, data ownership, data quality and lineage frameworks, and audit readiness. An AI system inherits every weakness in the data and controls beneath it, so governance is treated as part of the architecture rather than a review at the end.

Method: five steps from pain point to working proof of concept

1. Start with the pain point, not the technology

   Sit with the people who own the problem, understand what it costs them in time, risk and effort, and agree what a solved version looks like and how that would be known. Every later decision is measured against that.

2. Design the solution

   Architecture, data sources and integration points, the risk tier and the human-in-the-loop controls that tier requires, and the acceptance criteria the system has to meet. The design is written to be built, not to be presented.

3. Build the working prototype

   Using AI coding tools, a working prototype is built against real documents and real data wherever governance allows. The prototype is the argument. The people who own the pain point use it, and the conversation moves from whether this could work to what it needs before it goes live.

4. Specify the stack end to end

   Hardware and GPU sizing, operating systems, databases, security, data management, model routing and cost metering.

5. Hand over a demonstrated system

   Engineers take a working, demonstrated system to production, not a document that describes one. The specification has already been proven by the prototype. The principal stays involved to direct the build, hold the technical bar, and keep the outcome tied to the original pain point.

The effect is that the distance between a decision and a live system gets shorter. The engineers who take it to production are not interpreting a vision. They are hardening something they have already seen work.

Track record: an AI programme in production

Over roughly two years the principal designed and led an AI adoption programme at a large research university that put ten AI workflows into production across nine business functions. Every one of them shipped into a department that did not report to him. The platform has been sustained in production since October 2024. It is not a pilot.

What changed

• Service operations moved from days to minutes on routed requests.
• Document-heavy review dropped from more than thirty minutes to under five.
• Demand forecasting is now used by 25 departments for planning.
• 367 governed documents are served through a single role-aware gateway across five channels.

Representative workflows

• An institutional answer gateway over governed policy documents, with citations.
• AI-native admissions screening using vision, optical character recognition and document parsing.
• An advising and course-planning engine running against live student records.
• Procurement split-request detection using similarity analysis and graph clustering.
• An AI service desk that drafts every response and sends none. A human approves every reply.

Track record: thirty years of mission-critical operations

Telecommunications service operations and information security at national scale. Information and communications technology and trunk infrastructure for a major Gulf development programme. Most recently the full technology estate of a research university.

Platform and stack fluency

The AI platform described above runs in-house, and that is a deliberate choice.

• On-premise data centres.
• Clustered SQL Server, including vector search.
• .NET applications.
• Microsoft 365, Teams and Entra single sign-on.
• Large language models reached through a governed in-house gateway over direct provider interfaces, more than 20 models across 5 providers, with routing, personally identifiable information masking, audit logging and per-query cost metering.
• Retrieval-augmented generation with citation.
• Vision and optical character recognition document intelligence.
• Human-in-the-loop controls set per risk tier.

Running a full production AI platform in-house on direct provider interfaces is what delivers both the data-control story and the unit-cost story. The data never leaves the institution's governance, and every query is metered against a known cost.

Credentials

• CISSPCertified Information Systems Security Professional
• CDMP MasterDAMA International, the highest examined level
• ISO/BS 27001 Lead AuditorInformation security management systems
• ITILIT service management
• MCSEMicrosoft Certified Systems Engineer
• CCNACisco Certified Network Associate
• B.Sc. Electrical EngineeringControl Systems, The University of Arizona, 1990

Career background

1. 2026 – presentPrincipal, AS Enterprise AI Advisory LLCGarland, Texas
2. 2023 – 2026Director of Digital Servicesa leading Gulf research university
3. 2013 – 2023Head of Information Technologya UAE higher-education institution, full P&L
4. 2011 – 2013Senior Manager, Information Technologya UAE technical institute
5. 2008 – 2011Senior Project Manager, ICT and Trunk Infrastructurea Dubai development group
6. 2001 – 2008Information Security and Quality Managera regional telecommunications operator
7. 1995 – 2001Regional Technical Manageran IT integrator in Dubai

Clients and employers are referred to by description rather than by name. This is deliberate.

Frequently asked questions

What does the practice actually deliver?

A working system, plus the specification and stack decisions needed to take it to production. Strategy work is delivered as interactive multi-scenario models rather than slide decks.

Does the principal write production code?

No. He designs the solution, builds the working prototype with AI coding tools, specifies the stack end to end, and directs the build. Engineers write the production code, working from a specification his prototype has already proven.

Is this a firm or an individual?

An independent, principal-led practice. One person. There is no staff, and no office beyond Garland, Texas.

Where does the practice operate?

Remotely from Garland, Texas, in the Dallas–Fort Worth area, serving clients internationally, with deep Gulf experience.

How large can an engagement be?

Engagements have spanned multi-year programmes across many business functions, delivered through teams that did not report to the principal, as well as individual use cases taken from pain point to demonstrated proof of concept.

What is the pricing?

Pricing is not published. Commercial terms are discussed per engagement.

How does someone start?

Email [email protected] or call (469) 922-0170. The most useful first conversation starts with a specific business pain point rather than a technology question.

What makes the stack advice credible?

Thirty years of accountability for running mission-critical infrastructure, including a research university estate of 600 or more servers, 390 laboratories and a US$41M budget, and information security at national scale for a telecommunications operator.

Can AI systems run without sending data to external providers?

The programme described above runs a full production AI platform in-house on direct provider interfaces, with routing, personally identifiable information masking, audit logging and per-query cost metering, which is what delivers both the data-control story and the unit-cost story.