What US industrial machinery makers disclose about their AI

Of 71 US machinery makers with $50M to $10B in revenue, 53 name an AI initiative in their own operations, but only 2 credit a cash figure directly to AI.

What did the study find?

Most US industrial machinery makers now name a specific use of AI in their own operations, but almost none says in cash what it returned. Of 71 machinery makers that file 10-Ks, with 2025 revenue of $50 million to $10 billion, 52 (73%) named at least one specific internal AI initiative in public between January 2024 and November 14, 2026. 12 (17%) gave a figure for one, and 2 (3%) credited a cash figure directly to AI, one of them a forecast.

Companies Share
Named at least one specific internal AI initiative 52 73%
Gave any figure for one (cash, hours or operational) 12 17%
…counting only figures the company or the press published 10 14%
Gave a cash figure they credit directly to AI 2 3%
Named AI only in products they sell, or only in general statements 7 10%
Nothing found beyond risk factors 12 17%

The study is the research behind the Roiva AI in Practice Index, US Industrial Machinery 2026, which names ten leaders. Every figure here is a claim as published, by the company itself or, where the table says so, by a vendor or the press; none was audited or measured against a control group.

Do larger machinery makers disclose more?

Yes. Of the 30 companies with $2 billion to $10 billion of 2025 revenue, 29 named an AI initiative; of the 41 with $50 million to $2 billion, 23 did. Every company that gave a cash figure it credits directly to AI is above $2 billion.

$2B-$10B (n = 30) $50M-$2B (n = 41) All (n = 71)
Name at least one specific internal AI initiative 29 (97%) 23 (56%) 52 (73%)
Give any figure for one 10 (33%) 2 (5%) 12 (17%)
Give a cash figure they credit directly to AI 2 (7%) 0 2 (3%)
Nothing found beyond risk factors 1 (3%) 11 (27%) 12 (17%)

The groups are every company that met the rules, not samples, but they are small: in the larger group one company is about 3 points, and in the smaller one about 2.

Which companies credit a cash figure to AI?

Two, and one of the figures is a forecast. Each is the company's claim, as published; none was checked against the company's accounts, and none is a benchmark.

Company Figure Credit to AI Source
BWX Technologies A hyperspectral sensor "combined with a machine learning algorithm" forecasts when a weld will go out of specification, and "that saves us about $2 million a year in rework" Direct Investor Day, February 2024
Manitowoc Its information systems department "began to use AI to convert several computer programs"; "As a result, we can save 2,000 man-hours and avoid spending $400,000." Direct; a forecast Earnings call, May 2025

No other cash figure passed the study's test for crediting it to AI. Brunswick's expectation of "more than $50 million" from its enterprise AI program names no specific use of AI, so it counts as a general statement. A figure counts as AI's only when the sentence with the figure names the AI work, or clearly refers back to AI work named in the sentence just before it. Program-wide savings, operating margins and productivity programs mentioned beside AI don't count.

Why do so many machinery makers talk about AI without using it?

Because for many of them AI is a market, not a tool. 35 of the 71 talk about AI mainly as demand for what they sell, such as cooling, power and pumps for data centers, or equipment for making chips. That demand isn't a use of AI, and the study doesn't count it. In their SEC filings, 38 of the 71 name AI only as a risk.

AI in a product a company sells, such as connected pumps, condition-monitoring platforms or vision in scanners, isn't counted either. The study recorded 77 such rows and left them out of every figure.

Do SEC filings show a machinery maker's AI work?

Mostly not. 13 of the 52 companies that name an internal AI initiative do so in an SEC filing, and 39 name one only outside their filings. Of the 12 that give a figure, 1 gives it in an SEC filing and 11 only outside them.

Found in SEC filings Found only outside SEC filings
Companies naming an internal AI initiative 13 39
Companies giving a figure 1 11

More of the rows with a figure come from investor presentations and conference talks (6) than from earnings calls (4).

How were the disclosures graded?

Each row, one source describing one initiative, got one grade for the strongest outcome that source states. The study recorded 360 rows under 238 initiatives, 139 of them internal.

Grade What the source gives Rows
A A cash figure the source sets beside the initiative. How directly it credits the figure to AI is graded separately 2
B Hours, staff or headcount, with a figure 2
C An operational figure: speed, volume, accuracy or share automated 11
D A specific initiative, with no figure 238
E A general statement about AI, with no specific initiative 107

A count of people using a tool, with no result, is graded D: it shows adoption, not an outcome.

How was the study done?

The group is every company that met the rules, not a sample, as of November 14, 2026.

  • Population: US companies that file 10-Ks, are classed by the SEC as makers of industrial and commercial machinery (other than computer and office equipment), have a US business address, and had 2025 revenue of $50 million to $10 billion: 73 companies. Two were left out because they had stopped being public machinery makers before November 14: TPI Composites deregistered after Chapter 11, and Ekso Bionics was taken over by Applied Digital's cloud business. A distributor and a holding company that the SEC classes as machinery makers were left out before that.
  • Sources searched: for each company from January 2024, its SEC filings, every earnings call, press releases, investor presentations and conference talks, its website, vendor case studies and trade press, plus a search for the names of AI vendors that sell to manufacturers.
  • Who did it: AI research agents (Anthropic's Claude), working to a written protocol under the direction of Roiva staff. Every quote was checked on its source page, by script or by a separate checking agent. A separate agent re-researched a sample of the companies blind, and what it found had been missed was added. Before the index named its leaders, checking agents re-read every row behind each leader, and Luke Wilcox, Roiva's founder, cleared each one.
  • What counts: a specific use of AI in the company's own operations, named in a public source. The source must name AI or machine learning: automation, robotics, smart factories and analytics don't count without it.

What are the study's limits?

Public evidence only: a company that says little may be doing more than one that says a lot. Some companies that make machinery are classed by the SEC in other industries, so they aren't in the group, and neither are machinery makers above $10 billion of revenue, such as Caterpillar, Deere and Cummins. Grades and use-case mapping follow written rules; a person reviewed every row behind the leaders, not every row in the study.

Has this study been corrected?

Five times: four on November 17, 2026, and once on November 18. The figures above are the corrected ones.

  • General statements and governance practices read on their sources (November 17). After the AI in Practice Index launched, checking agents read on their sources the governance practices and general statements about AI that the earlier re-reads hadn't covered. A general statement now counts only when it says what the company is doing with AI now: plans, hopes, hypotheticals and AI as an industry trend no longer count, and neither do risk sentences that name no use, trade-press writers' words or quotes that don't name AI. Oil States' only row was AI as a topic at its women's network events, so it now has nothing beyond risk factors. Companies with nothing found went from 11 (15%) to 12 (17%), and below $2 billion from 10 (24%) to 11 (27%); those naming AI only in products they sell or in general statements went from 7 (10%) to 6 (8%). Rows went from 389 to 386 (grade D from 274 to 273 and E from 99 to 97), and initiatives from 264 to 262, of which internal ones from 163 to 162. No company moved in or out of naming an internal initiative or giving a figure, and the index's leader list didn't change.
  • Every leader item read a second time (November 17). After the AI in Practice Index launched, a second reader, working without the first reading, read again every item behind its leaders that could be opened. Plans and hopes were dropped, programs and training recorded as uses became general statements, and quotes, statuses and dates were brought to the sources' words. Brunswick's expectation of "more than $50 million" from its enterprise AI program names no specific use of AI, so it is now a general statement, and Terex's webinars on the safe use of AI tools and Weatherford's AI training and AI Impact Hub are training, not uses. A vendor's account of Ingersoll Rand's Copilot support assistants now holds its result. Companies that give any figure went from 11 (15%) to 12 (17%), and above $2 billion from 9 (30%) to 10 (33%); those that credit a cash figure directly to AI from 3 (4%) to 2 (3%), now one of them a forecast. Companies giving a figure only outside SEC filings went from 10 to 11, and rows with a figure from investor presentations and conference talks from 8 to 6. Rows went from 386 to 379 (grade A from 3 to 2, B from 3 to 2, C from 10 to 11, D from 273 to 263 and E from 97 to 101), and initiatives from 262 to 253, of which internal ones from 162 to 153. Terex no longer meets the index's leader rule, so the index now names ten leaders.
  • Leaders' uses checked against a sharper test (November 17). After the AI in Practice Index launched, a reader checked every use behind its leaders against a sharper test of what a deployment is. That test applies from the next edition; for these editions we corrected only what the rules they were published under already rule out: plans and hopes, AI as an industry trend, and AI in a product a company sells. Among the machinery makers, rows went from 379 to 373 (grade D from 263 to 257), rows about AI in a product a company sells from 80 to 78, and initiatives from 253 to 251, of which internal ones from 153 to 152. No company moved in or out of naming an internal initiative or giving a figure, and the index's leader list didn't change.
  • The other companies read a second time (November 17). After the AI in Practice Index launched, a second blind sample was read again, and then a second reader, working without the first reading, read again every use and governance practice at the companies that aren't among its leaders. Writers' words, figures credited to AI in general and plans were corrected or dropped. Among the machinery makers, Astec Industries' remaining AI statements name no specific use of AI, so it now names AI only in general statements. Companies that name an internal initiative went from 53 (75%) to 52 (73%), and below $2 billion from 24 (59%) to 23 (56%); those naming AI only in products or in general statements went from 6 (8%) to 7 (10%). Companies naming an initiative in an SEC filing went from 14 to 13. Rows went from 373 to 362 (grade D from 257 to 238 and E from 101 to 109), rows about AI in a product from 78 to 77, and initiatives from 251 to 239, of which internal ones from 152 to 139. No company moved in or out of giving a figure, and the index's leader list didn't change.
  • The last unread documents read (November 18). Investor presentations that no allowed route could open earlier were read from copies of the documents. Two statements in them were plans, which a general statement leaves out, and were dropped. Rows went from 362 to 360 (grade E from 109 to 107), and initiatives from 239 to 238. No company moved in or out of naming an internal initiative or giving a figure.

Who ran the study?

Roiva ran this study. Roiva is AI portfolio management software that tracks every AI initiative's budget, approvals, cost and ROI, so whether companies can show a return on AI is a problem the company sells into. That is a reason to check the work. As of November 16, 2026, none of the 71 companies is a Roiva customer, partner, supplier or investor.

Questions about the data?

Ask us about the method or the per-company records.

Get the next one

New field notes by email as they're published. Or subscribe to the RSS feed.