MC Consulting, Inc. · Est. 1984 · Automation & Controls

Give your engineers
back the hours lost
to RFQs, specs,
datasheets and
documentation.

In automation and controls firms of 20–200 people, senior engineering time spent on documentation, quoting and specification work typically runs 25 to 40 percent. Most owners we talk to guess low — often by half. We build verified AI workflows that take that time back, using the same discipline our founder has applied to nuclear commissioning and high-voltage instrumentation for 52 years. The first question is what your number actually is.

Engineering capacity automation.
Verified AI workflows that automate everything around the expertise — never the expertise itself.

DisciplineMeasurement-Governed
Frontier ModelsClaude · GPT · Perplexity
VerificationCross-Sourced
Ship StandardBench-Proven
◆ INPUT · TECHNICAL DOMAIN ◆ MODEL SELECTION ◆ CROSS-SOURCE VERIFICATION ◆ CONSTRAINT CHECK MC EST. 1984
Bench-measured, not brochure-quoted 52 years engineering discipline Verified against primary sources Multi-model, not single-vendor Deployed only where proven Bench-measured, not brochure-quoted 52 years engineering discipline Verified against primary sources Multi-model, not single-vendor Deployed only where proven
[ 01 ] The Pattern

Three things show up in almost every automation and controls firm we look at. How many of them are true in yours is the part only you can answer.

01

Expensive people, inexpensive work

Engineering time in this sector generally costs $150–$250 an hour fully loaded, and a meaningful share of it goes to writing specifications, chasing datasheets, formatting proposals, and hunting down whoever solved this last time. That is precisely the class of work AI handles well — provided the output can be verified.

Ask yourself What share of your senior engineers' week is drafting and searching rather than deciding?

02

Confidence without verification

If you have already watched an AI tool invent a part number, you know exactly why verification is not optional. If you have not yet, you will. Consumer-grade AI has no discipline of verification — it will cite a datasheet that does not exist and hand back a bill of materials that does not build.

Ask yourself If a wrong number reached a customer, what would it cost — and who would have caught it?

03

The build is itself engineering work

Getting from "somebody should try AI here" to a workflow the team actually uses is a real engineering project: model selection, workflow design, verification checks, integration with existing systems, and a discipline your people can defend to an auditor or a customer. Most firms have neither the hours nor the appetite to work that out on their own time.

Ask yourself Who inside the business currently owns this — and what else are they supposed to be doing?

We do not know which of these applies to you, and we would not guess. The two instruments further down this page — the fit check and the workflow diagnostic — exist so that you can find out with your own numbers before speaking to anyone.

[ 02 ] What We'd Ask You

Before we would tell you anything, these are the questions — taken verbatim from the instrument we use on every engagement.

Most consultants keep the diagnostic behind the meeting. We would rather you saw it first — partly because it tells you more about how we work than any claim on this page could, and partly because a few of these will be uncomfortable to answer, and you should find that out on your own time rather than ours.

§ 1.0 · THE FIRST QUESTION
How much of your engineering payroll is currently being spent on work that does not require engineering judgment?
Why we askAlmost nobody has calculated it, and it is the only number that turns this from a technology conversation into a financial one. It leads straight into the sequence that follows: RFQ review, prior-job retrieval, specification comparison, datasheet research, proposal drafting, submittal preparation, documentation — and then engineer approval, which we never touch.
§ 2.2 · TIME COST
Of the time this task consumes, how much is drafting from scratch versus reviewing and correcting?
Why we askThe single most predictive number in the whole audit. AI collapses drafting time; it does not collapse review time. A task that is 80% drafting is a strong candidate. A task that is 80% judgment is not — and we will say so.
§ 2.4 · SOURCE OF TRUTH
When your engineer needs the right number, where do they look — and is that source machine-readable or somebody's memory?
Why we askThis determines whether verification is even possible. No verifiable source of truth means no verified workflow. That is a finding worth reporting, not a reason to proceed anyway.
§ D.2 · REVISION CONTROL
If we asked three people for the current revision of the same specification, would we get the same file?
Why we askThis separates a document storage problem from a document control problem. Storage is a filing exercise. Control is an engineering discipline — and it is the one with warranty and liability consequences.
§ A.4 · QUOTING & RFQ
When you decline an RFQ, is it because it was a bad fit — or because nobody had time to respond?
Why we askThe second answer is usually the largest recoverable number in the business, and almost nobody has calculated it. Declined per quarter × average contract value × historical win rate is a figure most owners have never seen written down.
§ 2.7 · PEOPLE
Who on the team is most likely to object to this — and what would their objection be?
Why we askAdoption failure kills more workflow deployments than technical failure does. If the person who owns the work has not been told this engagement is happening, that is a red flag we raise before any build starts.
§ 2.8 · CONSTRAINT
If this work disappeared tomorrow, what would you actually do with the capacity?
Why we askThis converts a time saving into a business case. "Saves six hours a week" is weak. "Lets you bid four more jobs a quarter at a 30% win rate" is a decision. If there is no answer to this question, the workflow may not be worth building.
[ 03 ] The Reframe

The question isn't whether to use AI.
The question is whose discipline is applied to it.

Consumer-grade AI is a brilliant intern with no discipline. Engineering-grade AI is an instrument that has been characterized, measured, constrained, and deployed only where it has proven reliable. MC Consulting builds the second kind — around a single rule: we automate everything around the expertise, never the expertise itself. Your engineers keep the judgment. What goes is the drafting, the chasing and the re-typing that surrounds it.

What this isn't

"We help companies implement AI."

Unverified output pasted into deliverables.

Single-vendor lock-in on a subscription tool.

A junior consultant with a prompt library.

What this is

Verified workflows for engineering and business tasks.

Every output cross-checked against primary sources.

Multi-model architecture — the right AI for each job.

A principal engineer with 52 years and a shipping record.

[ 04 ] The Services

Six practices we build for engineering, manufacturing, and operating companies. Every one is verified before it ships.

Not one of them touches engineering judgment. Every service below automates the work around your engineers' expertise — the drafting, searching, cross-referencing and formatting — and leaves the deciding where it belongs.

Service 01

AI-augmented engineering documentation

Specifications, submittals, RFQs, technical narratives, engineering change discipline, and vendor documentation packages — drafted in a fraction of the time, cross-verified against your standards and against the source datasheets. Your engineers review and approve; they no longer draft from scratch.

▸ Specs · Submittals · RFQs · Change orders
Service 02

Component & vendor intelligence

Datasheet cross-referencing, second-source identification, obsolescence forecasting, and vendor due diligence — done against real datasheets, not the model's memory. We verify every part number against the manufacturer's published record before it enters a bill of materials.

▸ Cross-refs · Alt sources · Obsolescence risk
Service 03

Institutional knowledge capture

The senior engineer who's about to retire has 30 years of "how we actually solved this" in her head. We build queryable knowledge systems that capture and preserve that expertise — searchable, verifiable, tied to specific projects and decisions. Not a chatbot: a durable engineering record.

▸ Retrievable memory · Auditable · Project-tied
Service 04

Proposal, RFQ & bid processing

For firms with more proposals than time to write them: ingest the RFQ, cross-reference your capability library, extract scope and pricing bases, draft the response, and flag every requirement your team must personally sign off on. Turnaround from days to hours, without lowering quality.

▸ RFQ intake · Draft response · Requirement flags
Service 05

Multi-document diligence & research

Contract review, regulatory synthesis, competitive analysis, technical literature review — across dozens or hundreds of documents in parallel. Every claim traceable to a source, every source verifiable. The workload of a research team, delivered by a small one.

▸ Contracts · Regulatory · Market · Technical
Service 06

Executive AI advisory & training

For owners, CEOs, and department heads: which models to use for what, how to structure a verification discipline your team can follow, how to evaluate vendors selling you AI, and how to build a policy that protects your IP and your reputation. Not theory — the exact playbook we run internally.

▸ Model selection · Verification design · IP policy
[ 05 ] Document & Process Control

Most firms don't have a documentation problem. They have a retrieval problem — and a single source of truth problem underneath it.

The paperwork jungle

  • Datasheets scattered across a shared drive, three engineers' laptops, and a vendor portal nobody has the login for
  • The current revision of a spec is whichever file was emailed most recently
  • Contract terms live in a filing cabinet; nobody knows which agreements auto-renew next quarter
  • Finding what part was used on a 2019 build means asking the one person who remembers
  • The same RFQ answer gets rewritten from scratch every time because nobody can find last time's
  • An obsolete component is discovered at purchasing, not at design

Controlled & searchable

  • Every datasheet indexed to its part, its revision, its vendor, and every assembly it appears in
  • One controlled record per document, with revision history and a defensible approval chain
  • Every agreement extracted to structured terms — dates, renewals, escalators, obligations, exposure
  • Query by part, project, vendor, date, standard, customer, or engineer — and get an answer in seconds
  • Prior responses surfaced automatically when a similar requirement appears
  • Obsolescence and second-source risk flagged at design time, not at procurement
▸ Searchable from every angle — the same corpus, six ways in
By part

Every component, its datasheet revision, approved alternates, lifecycle status, and every assembly and project it has ever appeared in.

By project

Complete document set per job — specs, submittals, drawings, change orders, correspondence, closeout — assembled rather than hunted.

By vendor / customer

Every agreement, quote, PO, certification, and performance record tied to a counterparty — with obligations and renewal dates surfaced.

By date & revision

What was current on the day that decision was made — the question that matters in a warranty dispute or a liability review.

By standard & requirement

Which clause of which standard drove which design decision — and everything in the corpus affected when that standard is superseded.

By plain-language question

"What did we quote Acme for the 480V panel in 2022, and which breaker did we spec?" — answered from the record, with the source cited.

[ Document classes brought under control ]
Technical
  • Component datasheets & cut sheets
  • Bills of material & where-used
  • Specifications & design basis
  • Drawings & revision history
  • Test procedures & results
  • Engineering change orders
  • Certifications & compliance files
  • O&M manuals & as-builts
Commercial — outbound
  • RFQ & RFP responses
  • Technical & commercial proposals
  • Quotes & pricing bases
  • Statements of work
  • Service agreements you provide
  • Warranty terms & exclusions
  • Capability & past-performance library
Commercial — inbound
  • MOAs, MOUs & LOIs
  • Supplier & subcontractor agreements
  • Purchase orders & terms
  • NDAs & confidentiality terms
  • Service agreements you receive
  • Leases, licenses & renewals
  • Insurance & indemnity provisions
[ 06 ] The Method

The verified-AI-workflow discipline — the same four steps we've used to commission nuclear systems, high-voltage instruments, and controlled-growing environments.

01/04
Characterize

Identify what the model actually does well on your class of task — not on published benchmarks. Bench-test against representative work from your firm. Document the failure modes.

02/04
Constrain

Design the workflow so the model can only operate where it's proven reliable. Everything else routes to a human or a second verification pass. No blank-page trust.

03/04
Verify

Every quantitative output cross-checked against primary sources — datasheets, standards, your own historical record. The invariant checklist tests each fix against every constraint already solved.

04/04
Deploy

Only after the workflow has passed the first three steps does it enter production. Your team runs it; we maintain the discipline. Measurement governs documentation.

"I understand why you don't trust it. I don't trust unverified AI output either. I treat these models the way I've treated engineering systems for decades — characterize them, measure them, constrain them, verify the output, and use them only where they prove reliable."

— Edward L. McCammon, CEO & Principal Engineer
[ 07 ] Who We Serve

One primary market, two adjacent ones — all three chosen because we have lived their operating problems.

MC Consulting helps industrial automation and controls companies recover expensive engineering capacity by automating the repetitive work surrounding RFQs, specifications, proposals, datasheets and technical documentation — while engineers retain every consequential judgment and approval.

Privately held industrial automation, controls and systems-integration companies with 20–200 employees.

Machine builders, panel shops, systems integrators, OEM controls groups, process and motion control houses, and automation companies building engineered-to-order equipment. Firms with expensive engineering talent that have not yet systematically integrated AI into their engineering and commercial workflows. This is the ground the founder has personally worked for decades — schematic capture, BOMs and component verification, PLC and control-system architecture, international vendor coordination, specifications, submittals, and engineering change discipline.

Company size
20 – 200 employees
Revenue floor
$1M+ annual
Decision maker
Owner · CEO · VP Eng
Sector trend
11% BLS growth · 2024–34
Adjacent 01 · Strong fit

Specialty electrical & engineered equipment manufacturers

Custom machinery, instrumentation, power electronics, switchgear and assemblies, and other engineered-to-order equipment built in low volume against customer specification. The documentation load per unit shipped is the highest of any manufacturing model — datasheets, BOMs, test procedures, certification submittals and O&M packages — and almost none of it is genuinely one-off. That gap between "written fresh every time" and "actually 70% identical" is the whole opportunity.

Adjacent 02 · Strong fit

Power, commissioning & critical infrastructure

Industrial controls and commissioning firms, power distribution, high-voltage systems, data-center infrastructure, and energy projects. Verified AI for bid and proposal throughput, first-pass spec extraction, submittal preparation and review, commissioning documentation, and standards retrieval — for firms absorbing rising project volume without adding equivalent headcount. The founder wrote operational test procedures for nuclear-grade computer, network and field-device systems, and later designed and built three Network Operations Centers and roughly 350 miles of fiber plant across Western North Carolina — for a Library of Congress program, a sovereign tribal nation, and a regional carrier. This is the same discipline applied to the same class of problem.

[ 08 ] Why MC Consulting

Not a prompt library and a business card. Fifty-two years of shipping engineering that has to work the first time.

The founder, Edward L. McCammon, has spent his career in the domains where getting it wrong is not an option: nuclear power startup engineering, DOD/DOE security systems, carrier-grade fiber and network operations centers, high-voltage cold plasma instrumentation, controlled-environment agriculture, and multi-brand operating companies.

The discipline is consistent across every one of those programs: measure before you commit, verify before you ship, and never accept a fix that hasn't been tested against every constraint already solved. That discipline is exactly what current-generation AI systems need — and almost nobody consulting on AI today comes from a career where that discipline was the price of admission.

At MC Consulting, that discipline is now applied to AI systems on behalf of clients who need the leverage without inheriting the failure modes. We use the tools every day on our own programs — a $75M engineering raise and an active R&D facility specification — and on client engagements, including a six-band Class E resonant plasma platform delivered under contract, before we recommend them to yours.

52+ yrs
Engineering practice
L3
ANSI Senior Nuclear Startup Engineer
350 mi
Fiber deployed · 3 NOCs built · WNC
$4M
Single-fix cost avoidance · E3S project
TS
Top Secret clearance · DOE & DOD · Savannah River weapons site
6
Nuclear sites worked · 5 power stations + DOE weapons facility
[ Frontier AI platforms we operate — tooling, not partnerships or endorsements ]
Anthropic Claude
OpenAI ChatGPT
Perplexity
Google Gemini
Specialist models
Custom pipelines
[ 09 ] The Arc

Four companies. One discipline. A fifty-two-year through-line from the nuclear plant floor to verified AI — every step automated further than the last.

What MC Consulting sells today is not new. It is the same discipline that has been carried, refined, and re-applied through every company the founder has owned and run. Each one solved its era's version of the same problem: how do you take work that costs too much, takes too long, and depends on the memory of one person — and turn it into a system?

1984 – 1995Earlier: TVA / Multi-Amp 1974–1984
MC Consulting, Inc.  — where this company began
CEO & Principal Engineer · Nuclear Power & DOD/DOE Systems

Seven years at TVA's Watts Bar Nuclear Plant — on-the-job training working hands-on across every system in the plant, which is where the fundamentals were learned rather than read. Then three years as a startup engineer with Multi-Amp, bringing the Callaway Nuclear Station online — the first commissioning work, and where the discipline got formalized.

This is the company you are reading about. He formed MC Consulting in 1984, immediately after Callaway, in order to contract directly for the next job: two years at the Palo Verde Nuclear Generating Station for Arizona Public Service, across all three nuclear units, bringing the plant's computer systems online. Balance-of-plant monitoring, reactor control rod systems, radiation monitoring, security systems, and the water treatment facility — plus training Arizona Public Service's in-house personnel to run and maintain them.

Then two and a half years at Plant Vogtle for Georgia Power and Southern Company, as procedure writer and startup work package controls processor through the commissioning of Units 1 and 2. Then the classified work. At the Savannah River Site in Aiken, South Carolina — a Department of Energy nuclear weapons facility — he held Top Secret clearance, worked on the team that designed the E3S Electronic Safeguards and Security System, and was onsite as startup engineer overseeing its installation, startup and the training of site personnel. He served as systems engineer on the E3S maintenance system and group leader for site-wide security access control, wrote the operational test procedures for computers, networks, fiber modems and field devices that governed how those systems could be safely commissioned, and re-designed a fiber-modem circuit that prevented a $4M cost overrun on the project. The last MC Consulting contract was Oconee Nuclear — procedure writer for electrical power, control and metering systems — before he turned to building restaurants.

The record, stated plainly · past project environments, 1974–1995
Past project environments included five nuclear power stations and one nuclear weapons facility — six client organisations, two federal agencies, Top Secret clearance. And after the first assignment, never as plant staff — always the outside specialist brought in to design it, bring it up, and hand it over.
Watts Bar
TVA · 7 yrs
Hands-on across every plant system. Where the fundamentals were learned.
Callaway
Multi-Amp · 3 yrs
Startup engineer. First commissioning assignment.
Palo Verde
Arizona Public Service · 2 yrs
Computer systems startup across all 3 units. Trained their staff to run it.
Vogtle 1 & 2
Georgia Power / Southern · 2.5 yrs
Procedure writer & startup work package controls processor.
Top Secret · DOE / DOD
Savannah River Site
Aiken, SC · Westinghouse / DOE · weapons facility
On the team that designed the E3S Electronic Safeguards system, then onsite as startup engineer over installation, startup and training of site personnel.
Oconee
Final MC Consulting contract
Procedure writer — electrical power, control & metering systems.
The pattern is the point: six operators — including a Department of Energy nuclear weapons site — kept hiring the same outside engineer to design or bring up critical systems, verify them against specification, and train their own people to take over. That is the identical shape of what MC Consulting sells today. Only the technology changed.
▸ What carried forward The nuclear-startup rule: a system is not commissioned until it has been measured against its specification and every deviation reconciled. This is now the operating philosophy of every MC Consulting program — and the reason "verified AI" is not a marketing phrase here.
THE WINNING TEAM
1995 – 200914 years · 2 tenures
The Winning Team, Inc.
Senior Partner → CEO & President · Development, Systems & Technology

The Winning Team was a franchisee — an independent operator building and running restaurants under the Arby's and Bojangles' brands, not the brands themselves. Over fourteen years it went from 13 stores and $7M to 35 locations and $31M. Mr. McCammon's lane was development and systems — site selection, facility design, build-out, and every piece of technology the company ran on. An operations partner oversaw restaurant operations. What he built underneath the growth is why this belongs on the timeline: it is where nuclear-grade engineering discipline first got applied to problems nobody considered engineering problems, and the systems that came out of it were the first automation platform he ever shipped.

He designed and wrote a menu-driven back-office software program for inventory, sale tracking, labor, and scheduling — cutting administrative time by two hours per store per day and saving an estimated $153K per year across 14 stores. He engineered and installed automated energy-control systems across every restaurant, saving 40% on electricity and gas and climbing to $126K in annual savings after amortization. He built a Cost Segregation software on the tax-code allocation model that produced a 75% reduction in taxes across the first five years of any new facility, and the freed cash financed the expansion from $7.5M to $29.2M. Arby's named him Innovator of the Year in 2008 and Bojangles' Rookie of the Year in 2009 — both awarded for the designs and systems, not for the sales line.

▸ What carried forward The Winning Team was the first proof — years before the phrase "digital transformation" existed — that writing custom operating software beats buying it, if you understand the business well enough to define it. Every MC Consulting engagement today is built on that same premise.
2009 – 2013The bridge
ECOS Systems, Inc.
Co-founder, CEO & President · Energy Management · Fiber & NOC Infrastructure · Agri-R&D

Co-founded to turn the Winning Team energy-control systems into a standalone consulting practice — turnkey energy evaluations with 24/7 monitoring for retail and franchise clients. But ECOS ran three workstreams under one roof, and the largest of them was telecommunications infrastructure.

▸ Building the fiber backbone of Western North Carolina · past project environments
NOC 01 · FEDERAL
Education Resource Consortium
Designed and built the Network Operations Center for a Library of Congress program serving Western North Carolina, sited in the Federal Building in Asheville — bringing high-speed fiber connectivity to a region that had largely gone without it.
FIBER PLANT · REGIONAL
350 miles of fiber
Deployment of approximately 350 miles of fiber optic plant across Western North Carolina — route engineering, build-out and commissioning of the physical layer the region's networks still run on.
NOC 02 · SOVEREIGN
Cherokee Nation
Network Operations Center designed and built for a sovereign tribal nation, supporting an additional fiber network across the same region under its own governance and operating requirements.
NOC 03 · CARRIER
BalsamWest
Network Operations Center for a regional carrier, extending the fiber footprint further across Western North Carolina. Three NOCs delivered for three entirely different classes of operator.

The lineage is direct rather than coincidental. Two decades earlier on the E3S nuclear project he had re-designed the fiber modem transmitter/receiver circuit for an access-control system — the fix that averted a $4M cost overrun — and written the operational test procedures for the fiber modems, network equipment and token-ring hardware that system depended on. The Western North Carolina work applied that same fiber and network discipline at regional scale.

And the third workstream — the applied agricultural research inside ECOS — is where the sealed-tunnel architecture, mobile grow-rack concept, and centralized nutrient delivery that would become Agrifacture was first laid down as engineering, not thesis.

▸ What carried forward Two things. First, the NOC discipline itself — centralized monitoring, 24/7 operations, and control architecture over distributed infrastructure — went directly into the Agrifacture platform's control room. Second, the instinct to take an operating system proven in one industry and abstract its principles into a platform for another. Restaurant energy control became indoor agriculture; nuclear fiber-network discipline became a regional backbone. That same move is exactly how MC Consulting applies engineering QA to AI systems.
Agrifacture
2015 – 2023Full automation shipped
Agrifacture™
CEO & President · Controlled-Environment Food Manufacturing

The full-scale execution of what ECOS proved possible. Designed, built, and operated two pilot facilities as fully-instrumented Total Controlled Growing Environments — sealed multi-tunnel structures with centralized nutrient delivery, dock-in / dock-out mobile grow racks, and unified climate, lighting, and CO₂ control across every zone. Chose fruiting mushrooms — the most environmentally demanding CEA crop — as the test article, on the reasoning that any platform holding those tolerances would generalize. Both facilities reached sustained production of ~750 lbs/week per 20-ft tunnel, then were closed on purpose once the architecture, controls, and recovery loops had been validated.

The governing insight was to stop treating a farm as a farm. The rack is the product carrier. The tunnel is the production line. The AGV is the conveyor. Once food production is modeled as an assembly line, every discipline from manufacturing — takt time, work-in-process routing, station-level quality gates, MES-tracked genealogy — becomes directly applicable. That reframe is what made full automation designable rather than aspirational.

▸ The automation architecture · 11 specifications · 400+ pages
ICS-001 · CONTROLS
2,200+ I/O control points
ISA-95 compliant control hierarchy. Allen-Bradley ControlLogix PLCs on an EtherNet/IP backbone, Ignition MES platform above them, 8-VLAN industrial network with strict IT/OT segmentation.
AGV FLEET · LOGISTICS
27-unit shared AGV fleet
20 transport · 4 wash shuttle · 3 packing shuttle, all LiDAR SLAM navigated. Modeling the fleet as shared rather than permanently paired to racks produced an 89% capital reduction against the naive design.
SAS-001 · SEEDING
7-station robotic seeding cell
Substrate preparation, automated tray loading, and sterilization at the head of the line. Cassettes routed upward by AGV with RFID genealogy tracked from seed lot to finished pallet.
HAS-001 · HARVEST
5 robotic harvest cells
Vision-guided pick with soft-gripper end effectors and multi-pass logic for crops that mature unevenly. Strict dirty/clean segregation architecture enforced by the routing layer, not by procedure.
HVS-001 · CLIMATE
24-zone independent climate
600–800 TR central plant with per-tunnel AHUs, CO₂ injection, and a clean-to-dirty pressure cascade that makes cross-contamination a physics problem rather than a housekeeping one.
NPS-001 · NUTRIENT
Switchable-manifold tank farm
Centralized recipe delivery with manifolds that let one facility run multiple cultivars simultaneously. Recirculating hydroponic loop returns over 90% of process water.
WTS-001 · SANITATION
5-stage automated wash tunnel
Pre-rinse, detergent, hot sanitizer, final rinse, air dry — CIP/SIP integrated, with the whole loop instrumented so that a sanitation failure is an alarm, not a discovery three days later.
RACK · MASTER v3.4
Convertible crop-agnostic rack
A single 6×8×9 ft chassis with 5×5 ft male-female AGV nesting and interchangeable crop modules. The hardware is the abstraction layer — switch the recipe, switch the crop, same steel.
The AI operational design layer
  • Machine-vision inference across 30–40 industrial cameras — continuous quality inspection from seed to package, crop maturity classification, disease and contamination detection, and harvest-quality grading, all feeding decisions back into the MES rather than into a dashboard nobody reads.
  • Yield forecasting from live environmental telemetry — models trained against the pilot production record predict tunnel output ahead of harvest, so packing, labor and cold-chain scheduling are driven by forecast rather than by what shows up on the dock.
  • Closed-loop climate optimization — the control system holds setpoints; the AI layer decides what the setpoints should be, tuning light spectrum, DLI, CO₂ and RH per zone against measured crop response instead of against a printed recipe card.
  • AGV fleet routing and traffic optimization — dispatch decisions across 27 vehicles and 300 rack positions solved continuously rather than by fixed schedule, which is what made the shared-fleet capital reduction achievable in practice and not just on paper.
  • AI-augmented specification authoring and cross-referencing — the 11-document engineering suite is drafted, cross-checked and version-controlled with frontier models in the loop, under the discipline that no part number enters a released document unless verified against a datasheet, and that every proposed change is tested against the invariant checklist before it ships.

Now principal author of that eleven-specification engineering suite — contractor-ready — for the Phase 2 R&D facility in College Station, Texas, where the robotics and the AI layer get hardened at full production rate before the platform is templated for global deployment.

▸ What carried forward Agrifacture is where the full stack — hardware, controls, MES, cross-referenced specifications, and eventually AI-augmented documentation — came together as one program. It is where verified AI workflows were first used at scale on real deliverables, before we ever sold them to a client.
ION-FLUX
2025 – presentFull-stack AI integration
ION-FLUX  — client engagement
Contracted design & engineering · Six-Band Class E Resonant Plasma Platform · AI-native from day one

This one is a client's program, not ours. It arrived the way good work often does — a friend of a friend with a hard problem and no obvious place to take it. The brief: take a technology Nikola Tesla patented 135 years ago and bring it into the 21st century on solid-state hardware. Torino Ltd — Engineering · AI · InnovationThe work was contracted through Torino Ltd, the international consulting entity, and has since been consolidated into MC Consulting. The client owns the product and the company — and under the terms of that contract, we retained the intellectual property developed in the course of the work. We own the engineering discipline that got it built, and the design that came out of it.

The deliverable is the Mark V "PIONEER" — a six-band Class E resonant platform spanning 125 to 575 kHz at 15,000 to 25,000 volts per bulb. We took it from concept through schematic capture, bill of materials, international vendor coordination and full transformer characterization: secondary board laws, coupling coefficients, assembled resonances, primary geometry — measured on the bench rather than derived on paper. The invariant checklist that tests every proposed fix against every constraint already solved was written for this program, in response to specific failures, and has since caught others before they reached fabrication.

And this is where the two halves of the practice met. We had already been running AI inside our own work at Agrifacture — specification authoring, cross-referencing, verification. ION-FLUX was the first engagement where we integrated full AI systems into a client's design and build-out from day one, as operational infrastructure rather than an assistant bolted on afterward. Anthropic Claude, OpenAI ChatGPT, Perplexity and specialist models run in production across the entire program, under the same rule that governs the transformer bench: measurement governs documentation.

What we delivered on the ION-FLUX engagement
  • Schematic review and design-rule checking — proposed topology changes evaluated against the six-band resonance constraints, Class E switching conditions and thermal limits before they reach layout.
  • Datasheet cross-referencing and BOM verification — every part number checked against the manufacturer's published record, with second-source identification and obsolescence risk flagged at design time rather than at procurement.
  • Transformer and resonance mathematics — coupling coefficients, assembled resonances and secondary board laws computed and sanity-checked against bench measurement, with disagreements treated as findings rather than rounding.
  • The invariant checklist, enforced — every proposed fix tested against every constraint already solved, so a change that quietly reopens a resolved problem is caught before fabrication instead of after.
  • Parallel business workstreams — investor materials, financial modeling, contract analysis, international vendor coordination and multi-document diligence, all running through the same verification discipline as the engineering work.
▸ Why this matters to you ION-FLUX is the proof that this transfers. We built the discipline on our own programs, then took it into someone else's — an unfamiliar technology, a client's requirements, a client's timeline — and delivered hardware. If you have a product to design, a legacy technology to modernise, or an engineering program that needs AI built into it properly rather than bolted on, that is the same engagement. You are not our first. Ownership terms are set per contract and negotiated up front — on ION-FLUX we retained the IP; on your engagement it is whatever the two of us agree in writing before work starts.
The Convergence

Every company on this timeline solved its era's version of the same problem: turn expert work into a system. AI is simply the newest instrument for doing what MC Consulting has been doing since 1974.

[ 10 ] Market Context

Manufacturers are already spending on this. The winners will be the ones who deploy it with discipline.

$3.44B
US manufacturing-technology orders, H1 2026

AMT reports $3.44 billion in US manufacturing-technology orders in the first half of 2026 — 36% above the first half of 2025, and the strongest half-year since the data series began in 1998. AMT attributes part of the rising order value to additional automation content.

Source · AMT · H1 2026
80%
Manufacturing executives budgeting smart-mfg

Deloitte's 2026 manufacturing outlook reports that 80% of 600 manufacturing executives surveyed planned to put at least 20% of their improvement budgets into smart-manufacturing initiatives, with AI specifically named across supplier engagement, knowledge capture, work instructions, and production-office activities.

Source · Deloitte 2026 Mfg Outlook
11%
Industrial-engineer employment growth

BLS projects industrial-engineer employment to grow 11% from 2024 through 2034 against roughly 3% across all occupations, citing cost reduction, process optimisation and automation as the demand drivers. Electrical and electronics engineering is projected at 7% over the same period.

Source · US Bureau of Labor Statistics
176GW
Projected US data-center power by 2035

Deloitte estimates US data-center power demand could rise from 33 GW in 2024 to 176 GW by 2035. That surge is radiating through industrial supply chains — driving demand for generators, cooling equipment, cables and components. Transformer and grid-equipment lead times now exceed 160 weeks in some cases, and every deferred order still carries proposal and documentation load.

Source · Deloitte · Reuters
[ 11 ] Engagement Process

A defined sequence — diagnosis before prescription. No engagement begins until the fit is clear.

Week 00Discovery
60-minute conversation. No obligation, no pitch deck.

We diagnose whether your firm is actually a fit for verified AI workflows right now. Some are, some aren't. If the timing is wrong or the workload doesn't justify the investment, we say so, and the call ends without an ask.

Week 01–02Assessment
On-site or remote workflow audit.

Two-week engagement to characterize the repetitive engineering and administrative work your team is doing today, map it against what verified AI can and cannot do reliably, and produce a written report ranking candidate workflows by ROI and risk. This is a fixed-fee deliverable, not a lead-in to a bigger sale.

Week 03–08Pilot Build
One workflow, fully built, verified, and deployed.

We select the single highest-ROI workflow from the assessment, build it, characterize its performance against your work, and deploy it with your team trained and the verification discipline documented. You measure the impact before deciding whether to expand.

Week 09+Rollout & Advisory
Ongoing engagement — as much or as little as you want.

Additional workflows, advisory retainer for your executive team, quarterly discipline audits, or a full-time embedded partnership. Every client stays engaged only for as long as the value is measurable. No lock-in, no auto-renewal.

[ 12 ] Engagement Tiers

Three ways to work with us. All three are diagnosed before proposed.

Tier 01 · Assessment

Workflow Audit

Fixed fee2-week engagement
  • Two-week workflow characterization
  • Written report ranking candidate AI workflows by ROI & risk
  • Model-selection recommendations tied to your specific tasks
  • Verification-discipline framework tailored to your industry
  • 60-minute readout with your leadership team
Tier 03 · Retainer

Executive Advisory

Monthly retainer3-month minimum · continues quarterly
  • Ongoing workflow expansion across your firm
  • Executive advisory to owner & C-suite
  • Quarterly discipline audits of AI use in your firm
  • AI vendor evaluation on your behalf
  • IP-protection & policy framework maintenance
[ 13 ] Full-Stack Delivery

Once the engineering core holds, the same discipline extends outward — one provider, one verification standard, across the whole operating layer.

First · Always

The engineering core

Specifications, RFQs, datasheets, BOMs, submittals, document control. This is where the leverage is largest and where the verification discipline has to be established. Nothing downstream is worth building until this holds.

▸ Then outward ▸
Second · When it earns its place

The commercial layer

Agents, CRM, intake, front-of-house. Genuinely valuable once the technical core is under control — and genuinely a distraction before it. We will tell you which side of that line you are on.

AI agents

Autonomous execution of the workflows the audit has already characterized and bounded — triggered by events in your systems, operating inside the verification envelope, escalating to a human at the edges rather than guessing.

▸ Ties into · verified workflows · document control
CRM design & integration

A CRM that actually reflects how engineered-to-order business works — quote revisions, long cycles, spec-driven scope changes, and the relationship between an RFQ, its proposal, and the job it becomes. Built or rebuilt around your process, not a template.

▸ Ties into · RFQ pipeline · quoting · past-performance library
AI receptionist & intake

Voice and message handling that knows what your company actually builds — qualifying an inbound RFQ, routing a technical question to the right engineer, capturing a service call with the right equipment details, and never inventing a capability you do not have.

▸ Ties into · CRM · quoting · service dispatch
Quote-to-cash tie-in

Closing the loop from inbound enquiry through proposal, order, engineering release, and invoice — so the same part number, revision and scope description survive the whole journey instead of being retyped four times by four people.

▸ Ties into · ERP · CRM · BOM · document control
Internal knowledge assistants

The queryable version of your own record — what was specified, why, on which project, against which standard. Answers cited to the source document rather than asserted, so an engineer can check the claim in one click instead of trusting it.

▸ Ties into · document control · knowledge capture
Integration & orchestration layer

The connective work that makes the rest of it hold together — API integration between systems that were never designed to talk, event routing, and the plumbing that stops your staff being the integration layer between four pieces of software.

▸ Ties into · everything above
[ Why one provider, and when it is the wrong idea ]
A turnkey provider is only an advantage when the same verification standard runs through every layer. Otherwise it is just one throat to choke for four separate problems.

The argument for consolidating this under one provider is real: the agent that drafts your proposal, the CRM that stores it, and the receptionist that took the call all depend on the same underlying record. Split across three vendors, nobody owns the seam — and the seams are where these systems fail. Under one provider with one discipline, the record stays coherent from first contact through engineering release.

The argument against is equally real, and we will make it to you where it applies. If your engineering documentation is still uncontrolled, an AI receptionist is a distraction with a monthly fee. If your CRM problem is that nobody enters data, no amount of automation fixes a process problem. We would rather sequence you correctly and earn the second engagement than sell the whole stack on day one and watch half of it go unused.

[ 14 ] Common Questions

The questions every engineering owner asks in the discovery call.

How do I know the AI isn't making things up?+
You don't have to trust it — we don't either. Every quantitative output routes through a verification pass against primary sources: the manufacturer's datasheet, the published standard, your own historical record. The workflow is designed so the model can only operate in a bounded region we've bench-tested. Anything outside that boundary is escalated to a human. This is the same discipline used to commission nuclear systems — you don't trust the instrument, you calibrate it.
What about my intellectual property? Won't the AI train on it?+
We design the workflow so your proprietary data is handled through the appropriate enterprise or private-deployment channel for each model — not a consumer account. Anthropic, OpenAI, and other frontier vendors offer enterprise terms that contractually exclude your data from training. Where the sensitivity warrants it, we architect around locally-hosted or private-inference models. IP protection is part of the assessment, not an afterthought.
My engineers won't adopt it. How do you handle that?+
Engineers resist AI when it's introduced as a threat or a black box. They accept it when it removes work they hate doing and preserves the review authority they need to defend the output. We build workflows the engineer approves before it ships — the AI drafts, the engineer verifies and signs. Adoption follows when the tool respects their judgment.
Which AI platform do you use?+
All of the major ones — Anthropic Claude, OpenAI ChatGPT, Google Gemini, Perplexity, and specialist models — because different tasks call for different tools. Model selection is part of the discipline. A firm that locks you into a single vendor is selling you their preference, not your best solution.
Are you an AI agency, or an engineering practice?+
An engineering practice that delivers AI systems — and the distinction matters. Most AI agencies are marketing or software firms who learned the tools recently; the domain knowledge stops at the prompt. We come from the opposite direction: fifty-two years inside schematic capture, BOMs, control systems, specifications and commissioning, now applying that discipline to AI. That is why we can tell you a workflow is not worth building. An agency paid to deploy has no incentive to say that. We do deliver the full stack — agents, CRM, intake, integration — but the engineering core comes first, and we will sequence you honestly rather than sell you all of it at once.
Can you handle the whole thing — agents, CRM, phones — or just the engineering side?+
The whole thing, under one verification standard. AI agents executing bounded workflows, CRM designed around engineered-to-order reality rather than a SaaS template, AI receptionist and intake that understands what you actually build, and the integration layer connecting it. The condition is sequencing: if your engineering documentation is still uncontrolled, a receptionist bot is a distraction with a monthly fee. Get the core right and the rest compounds on top of it. We will tell you which side of that line you are on before you spend anything.
How much does a real deployment cost?+
A single verified workflow deployment typically ranges from $2K to $20K depending on complexity, integration depth, and verification requirements. The Workflow Audit is a fixed-fee two-week engagement designed to give you the numbers before you commit. If the ROI isn't there, we tell you so — and the assessment fee is the only cost.
How long before we see results?+
First measurable results usually within 30–45 days of the pilot deployment. The Workflow Audit itself produces immediate value — most clients report the report alone shifts how they think about which work is worth their engineers' time.
[ 15 ] Fit Check

Six questions. Thirty seconds. Find out whether a discovery call is worth your time before you spend it.

Fit Check 0 / 6
How many people work in the company?
Q1 · Scale
How much repetitive technical or contractual documentation does your team produce?
Q2 · Workload — the core signal
Of the time spent on that work, how much is drafting from scratch versus reviewing?
Q3 · The most predictive question we ask
When someone needs the correct figure or current revision, can they find it reliably?
Q4 · Source of truth — determines whether verification is possible
What would success actually look like for you?
Q5 · Intent
Could you authorize a fixed-fee engagement, or bring the person who can?
Q6 · Decision path
[ 16 ] Workflow Diagnostic

The Fit Check asks whether you qualify. This asks where the hours are going — and what we should audit first.

Company profile
Establish the operating context before evaluating individual workflows.
Workflow under review
Enter one workflow at a time. The diagnostic will rank several workflows for you.
How much of the work is drafting, assembling, searching or formatting? 60%
The single most predictive input · AI collapses drafting time, not judgment time
Annual hours consumed
1,040hrs / year on this workflow
Recoverable estimate
468hrs / year · drafting × judgment
Indicative annual value
$42kat $90 / hr fully loaded

Ranked candidates

0 workflows scored
◆ Book Discovery Call ◆

Sixty minutes. No pitch deck. No obligation.
A diagnosis — not a sales call.

If your firm is a fit, we'll say so and outline what a Workflow Audit would look like. If it isn't — because the timing is wrong, the workload doesn't justify it, or the discipline isn't ready — we'll say that too. Either way, you leave the call with a clearer picture of where verified AI fits in your operation than you had going in.