David Crowder

Vendors demonstrate on crisp PDFs that have never left a desktop — clean lines, clean text, everything where it should be — while shops receive drawings that have been scanned, faxed, folded in half, written on in pencil and passed through several hands.

• A system that passes the demo meets its first genuine customer drawing when a real job depends on it.
• The practical question for buyers is simple: what did the system get tested on, and did any of it come from a real job?

One incoming drawing looked completely normal on paper, but its file had saved every word twice beneath what the eye sees: a value of 0.5 came back as "0.50.5," and every check failed on a drawing any machinist would have read without squinting.

• On a second drawing, notes printed at an angle weren’t found at all.
• Both files, Crowder argues, are ordinary — the kind that arrive from customers, engineers and purchasing departments.

Read original

A polished demo can read a drawing, run a compliance check, flag an issue and route it to the right person; it doesn’t answer who sets it up, maintains it or handles it when it breaks.

• In a shop with a dozen people, the owner is already quoting parts, running the floor, handling customers and doing the books.
• There’s no spare person to become the AI person.
• Crowder’s priority is operability over capability: whether a tool survives contact with a small team, not whether it performs on a screen.

Crowder cites the scale of the problem directly: 74 percent of US manufacturers have fewer than 20 people in the whole building, against more than 239,000 manufacturers overall, three quarters of them small operations. Vendor answers to questions about setup and maintenance assume infrastructure that exists at a 200-person plant but not at a shop with 12.

Read original

Widening the lens turns the same structure into a resilience argument.

• In World War II, the whole country converted: factories that made consumer goods switched to batteries, tires and tanks, and that is how the war was won.
• Crowder says that couldn’t happen now because those facilities are gone, and notes China’s roughly four times the population and somewhere around 70 percent more manufacturing capacity.
• The position holds two things at once: the United States is still the second-largest manufacturer on earth, and Crowder wouldn’t bet on it in a long manufacturing war.

The reasoning rests on a scale comparison rather than a single incident.

• During COVID, the country needed two products and couldn’t make its own N95 masks — a portion of what arrived from overseas did not meet performance standards — while the national ventilator stockpile held roughly 20,000 units and 200,000 more had to be ordered under the Defense Production Act.
• A long conflict, Crowder argues, demands everything a country can build, all at once, for years.

Read original · Original 2

Their expertise is as a maker and builder of AI systems, and the parallel is deliberate — Crowder calls the work "software manufacturing" because both are constrained, both demand precision, and both break when someone skips a step; the only real difference is a physical constraint versus a virtual one. The stated background runs from Thomas Register, through years at MFG, to founding an additive manufacturing factory recognised by the World Economic Forum, described as 35 years around people who make things — standing next to the work rather than running machines, quoting parts or managing the floor.

Crowder draws the line in their own words: "I am not the shop owner.

• I am also not the tech vendor pitching from a slide deck."
• The positioning is an argument about where useful manufacturing software comes from: close enough to the work to understand it, and experienced enough in software to know what it can’t do inside a real operation.

Read original

Crowder describes a drawing that looked normal on paper but had saved every word twice beneath the visible text.

• A value of 0.5 came back as "0.50.5," and every check failed on a drawing any machinist could read without squinting.
• On a second drawing, the system missed notes printed at an angle.
• The point is testing conditions: most manufacturing AI is demonstrated on crisp PDFs that have never left a desktop, making the first genuine customer drawing the system’s first real test.
• Crowder’s question for buyers is direct: what was the system tested on, and did any of it come from a real job?

For brands planning a campaign

Use this as the technical spine of a co-authored piece on AI failures in small shops. It turns an abstract reliability claim into a testable one: state what the system was trained and tested on, then show it handling scanned, folded, pencil-annotated drawings.

Crowder notes that 74 percent of US manufacturers have fewer than 20 people in the whole building, out of more than 239,000 manufacturers overall, three quarters of them small.

• The demos look convincing: a system reads a drawing, runs a compliance check, flags an issue, and routes it to the right person.
• Then the owner asks who sets it up, maintains it, and handles breakdowns.
• The answer assumes an IT department, compliance team, or dedicated integration person — infrastructure found at a 200-person plant, but not a shop with 12.
• The owner is quoting parts, running the floor, handling customers, and doing the books, with no spare person to become the AI expert.

For brands planning a campaign

Reframe the campaign around operability, not capability. A co-authored piece could focus on deploying AI in a small team — setup, maintenance, and failure handling in a shop with no IT staff — rather than model performance.

Crowder contrasts World War II output — 96,000 bombers and 86,000 tanks — with the present.

• The country converted wholesale then because factories making consumer goods switched to batteries, tires, and tanks; that could not happen now because the facilities are gone.
• The scale claim is specific: China has roughly four times the population and somewhere around 70 percent more manufacturing capacity, compared with 239,000-plus US manufacturers, three quarters of them with fewer than 20 employees.
• The COVID comparison distinguishes a narrow test from a broad one: the country could not produce its own N95 masks, some of what arrived did not meet standards, the national ventilator stockpile held roughly 20,000 units, and 200,000 more were ordered under the Defense Production Act — two products, versus everything a country can build, all at once, for years.

For brands planning a campaign

This is the furthest of the four from the software campaign need, but it shows Crowder’s reach into arguments about capacity, resilience, and the supply base. Cite it if the campaign expands from shop-floor AI to the industrial base that depends on it.

Crowder states plainly that they are not a machinist and never pretended to be one, describing themselves instead as a maker and a builder who builds AI systems and calls the work "software manufacturing," on the grounds that both are constrained, both demand precision, and both break when someone skips a step, with the only real difference being physical versus virtual constraints.

• The stated background runs from Thomas Register, through years at MFG, to founding an additive manufacturing factory recognized by the World Economic Forum, with 35 years spent around people who make things — explicitly standing next to the people doing the work rather than running machines, quoting parts, or managing the floor.
• The positioning is deliberate: not the shop owner, and not the tech vendor pitching from a slide deck.

For brands planning a campaign

Use this to frame any co-authored placement. It addresses the first objection a machinist audience will raise — why a software person is writing about their shop — without overstating hands-on production experience.

They point to small teams, no dedicated IT staff, and owners already handling quoting, production, customers, and bookkeeping. Their recurring test is whether a tool fits those working conditions, not just whether it looks capable in a demo.

Uses the shop owner's practical questions—who sets it up, maintains it, and fixes it—to expose assumptions hidden in product pitches.

For brands planning a campaign

A company targeting small and mid-sized shops could work with Crowder on content about setup, maintenance, staffing demands, and who handles failures. Make the piece specific about what a shop must do to use the product; avoid presenting a polished demo as proof it fits a small team.

Examples: Source 1

They give concrete examples: duplicated text made a value read as “0.50.5,” while angled notes went unnoticed. Their argument is that evaluation should include files from actual jobs, not only pristine examples.

Makes a broad concern about AI reliability tangible through specific input defects and their consequences.

For brands planning a campaign

For a manufacturing AI company, a co-authored piece could explain a practical drawing-test protocol and show how messy inputs affect results.

• Use representative, permission-cleared examples and describe both failures and limits.
• Ask for evidence that the product passes this test.

Examples: Source 1

They contrast the scale of wartime manufacturing with current reliance on overseas supply, and argue that small operations matter to any manufacturing comeback. Their posts use historical and pandemic-era examples to make the stakes legible, rather than treating shop technology as an isolated software issue.

Moves from national-scale examples to the structure and constraints of individual manufacturers.

For brands planning a campaign

This theme may suit an industry or policy-facing campaign about manufacturing resilience, especially if it can connect the argument to concrete shop needs. Verify the figures and historical claims independently before using them as campaign facts.

Examples: Source 1 · Source 2

They present their perspective as that of a maker and software builder who has spent years alongside manufacturing workers, and compare building AI systems with manufacturing through their shared demands for precision and process.

Defines the limits of Crowder's expertise while explaining why they consider their software-building experience relevant.

For brands planning a campaign

A campaign can draw on Crowder to discuss translating shop-floor realities into software requirements, while accurately describing their role. Do not position them as a machinist or as speaking for every shop; their own post distinguishes proximity to the work from operating machines or managing a shop.

Examples: Source 1

They describe it as the start of 35 years working alongside manufacturers.

View example

They describe both companies as the background to building AI tools for machine shops.

View example

Both writers treat adoption as an operating problem rather than a purchase.

Esen covers safety and operations from inside a vendor, focusing on EHS leaders, control rooms and workforce frameworks.

• Crowder focuses on small-shop software fit and AI reading real drawings, writing from outside any single product.
• Esen builds arguments around structured research—a World Economic Forum human-machine collaboration framework and a 22,487-mile customer tour—while Crowder builds them around specific cases.

Read original · Original 2

Crowder makes the same move with demo performance versus shop-floor performance. Both writers trace a misleading metric back to the unglamorous practice that prevents it.

Hutchinson covers B2B revenue operations, CRM data and marketing measurement, not machinery; Crowder covers drawings, shop staffing and manufacturing capacity. Hutchinson writes as an operator auditing hundreds of companies over twenty years, while Crowder positions as a software builder alongside the industry.

Read original · Original 2

The overlap with Crowder is the manufacturing owner's-eye view of AI and a willingness to pose open questions rather than sell an answer.

Matan writes about subjects well beyond the shop, including commercial real estate investing, market timing, retirement and household costs, as well as resource downloads and show promotion.

• The questions stay broad.
• Crowder stays inside the shop, making a specific, testable argument about whether a system has seen a real job's file.

Read original · Original 2

Esen has published directly on that ground: a five-point account of what changes on real shop floors when vision systems, robotics and autonomous equipment arrive, including the argument that automation should remove exposure rather than people; and a separate write-up of three months and 159 customer onsites across 11 states and 12 industries, reporting five patterns heard from 47 EHS leaders, 32 operations leaders, 10 IT leaders and about five executives. A vendor whose product is bought by a committee rather than a single shop owner gets a closer audience match, and a co-authored piece can present those patterns as findings rather than opinion.

The vantage is enterprise and committee-driven.

• The argument runs from boardroom to plant floor, through shared ownership of outcomes among EHS, operations and IT, and toward standardisation across shifts and sites.
• The twelve-person shop with no IT staff—the buyer for a campaign aimed at shops of that size—barely appears.
• Esen’s posts also focus on safety technology sold to larger sites, so a co-authored piece inherits a safety-and-surveillance frame that the drawing-and-AI vendor does not sell into.
• If the campaign’s central question is who sets up and maintains the tool, and who answers when it breaks in a small shop, this alternative takes the discussion elsewhere.

Examples: Source 1 · Source 2

Matan writes as an owner in manufacturing and commercial real estate, has described using AI to save time and become more efficient, and asks peers how they are using it and where they think it is headed.

• The posts also show a working format: an episode of a named show, a downloadable guide and a comment prompt.
• That suits a vendor seeking a podcast-length or conversational placement rather than a written argument about product behaviour.

The published posts do not discuss how software behaves on real drawings or the setup, maintenance and staffing limits of a small shop.

• One turns to commercial real estate investing, market timing and why investors regret decisions they cannot change.
• Both posts link to a lead magnet, a show subscription and videos, placing a sponsored piece inside a real-estate-plus-manufacturing content funnel.
• A reader whose drawing failed every check would not find that problem addressed, and the vendor would pay to reach an audience split between manufacturers and property investors.

Examples: Source 1 · Source 2

Hutchinson has published two worked examples.

• One follows a company told its conversion rate was 80 to 85 percent when its own delivered numbers were 35 to 40 percent, tracing the problem to two honest definitions with different denominators.
• The other covers twenty years of B2B reporting stacks and argues that companies that grow treat marketing numbers as they treat their books, with a named owner and a reconciliation schedule.
• A vendor that wants to pair the drawing test with a buyer-side test of its own claims—or reach sales and operations leaders alongside shop owners—gets a stronger match on that half of the argument.

Esen’s posts do not involve manufacturing, machine shops, drawings or shop-floor work.

• The examples are a referral-partner conversion dispute and an audit of CRM and advertising stacks.
• The framing shifts from whether a system survives a real drawing to whether a company can trust its own numbers—a different buying conversation that the drawing-test angle does not need.
• Ask for evidence of overlap with small machine shops before buying reach into this broader B2B operations audience.

Examples: Source 1 · Source 2