Manufacturing Technology Adoption: 8.7% of Firms, 45.1% of Workers

Abstract

Two figures from the United States Census Bureau's Annual Business Survey are routinely read as one story. 8.7% of manufacturing firms use robotics. 45.1% of manufacturing workers are exposed to robotics. The gap is roughly five-fold and it is not a contradiction: adoption concentrates in the largest firms, and a small number of very large manufacturers employ a large share of the workforce, so a technology present at fewer than one firm in eleven reaches nearly half the people. The common reading, that manufacturing is automating, is true of the workforce and false of the firms. This paper argues that firm-level adoption and worker exposure are different measurements answering different questions, that conflating them renders the majority of manufacturers invisible in the discussion of their own sector, and that for the nine firms in eleven outside the robotics figure the available lever is not advanced manufacturing technology but the commercial record: whether a quote is a document or a record, whether an order carries a stage, and whether anyone can state what was promised to a customer without asking the person who promised it. The productivity evidence for firms that do adopt is taken seriously rather than dismissed (Guo 2025), and the argument states its own limits, including that no causal claim is made about commercial systems and firm performance.

Two measurements, one sentence

The Census Bureau's Annual Business Survey, conducted with the National Center for Science and Engineering Statistics, measures the diffusion of advanced technologies across United States firms, including artificial intelligence, cloud computing, robotics and the digitisation of business information (United States Census Bureau 2020).

Its manufacturing findings are usually quoted in a single sentence that contains both of the following. 8.7% of manufacturing firms use robotics. 45.1% of manufacturing workers are exposed to robotics (United States Census Bureau 2023).

Those are not two views of one quantity. The first counts firms and weights a plant employing eleven people identically to a plant employing eleven thousand. The second counts workers and weights by employment. When adoption concentrates in the largest establishments, the second figure rises far above the first without a single additional firm adopting anything.

Census reports the concentration directly: adoption of advanced technologies remains low, varies substantially across industries, and concentrates in large and young firms (United States Census Bureau 2022). Worker exposure across advanced technologies generally runs between 22% and 72% depending on the technology, against firm-level adoption that stays in single digits for robotics and artificial intelligence.

The distinction determines which sentence is true. If the question is what proportion of manufacturing work happens near a robot, 45.1% is the answer. If the question is what proportion of manufacturing businesses have adopted robotics, 8.7% is the answer, and the first figure says nothing about it.

Who disappears in the conflation

Reporting that leads with worker exposure describes an automating sector (United States Census Bureau 2023). Policy, vendor marketing and trade press all lean on that reading, and the firms it describes are real: large, capital-intensive, frequently young, and genuinely deploying advanced technology.

The nine firms in eleven outside that figure are also real, and they are the majority of manufacturing businesses in the country. They are small, frequently old, and running on equipment and processes that predate the discussion. When the sector's narrative is written from the 45.1% figure, those firms read coverage of their own industry and find nothing in it that describes their situation.

The practical consequence is a misdirected question. A manufacturer of forty people reading that manufacturing is automating concludes that the relevant decision is which advanced technology to buy. That is the decision facing the 8.7%. The decision facing the rest is whether the basic commercial machinery of the business is recorded anywhere.

Guo (2025) examines Industry 4.0 adoption against total factor productivity with firm-level evidence and finds a positive productivity association for adopting firms. That finding should be taken at face value rather than argued away: for firms in a position to adopt, the technology pays. It is also precisely why the conflation matters, because a real benefit available to a minority is being presented as a description of the whole.

What the majority is actually missing

The gap at a small manufacturer is rarely robotic and rarely exotic. It is generally the absence of a commercial record, and it shows up as a set of questions the business cannot answer from any system.

What was quoted, to whom, and when. At many small manufacturers a quote is a document produced in a spreadsheet, attached to an email and stored in a sent folder. It exists as a file rather than as a record, which means the set of all open quotes cannot be listed, their total value cannot be summed, and their age cannot be measured.

Which quotes were lost, and why. A quote that does not convert usually generates no entry of any kind. The firm therefore cannot distinguish losing on price from losing on lead time, which are different problems with different remedies.

What was promised. Delivery commitments made on a phone call and honoured through the memory of the person who made them are a functioning system until that person is unavailable, at which point the commitment is unrecoverable.

Which customers have stopped ordering. A manufacturer with a few hundred accounts and irregular order cycles cannot detect a customer's absence by intuition. Detecting it requires order history in a form that supports the question, and an interval definition against which absence can be judged.

None of those is a manufacturing problem. All of them are record-keeping problems, and none is addressed by any technology in the Annual Business Survey's advanced category.

Why the sequence matters

There is an ordering argument for addressing the commercial record first, and it does not depend on dismissing advanced technology.

Advanced manufacturing technology reduces the cost or increases the consistency of production. Its return is realised on volume, which means it is most valuable where demand is known and stable enough to plan against. A firm that cannot say which customers are slipping, which quotes are open or which orders are late does not have a reliable demand picture, and is therefore poorly placed to size an investment whose payback depends on one.

The investments also differ by an order of magnitude in both cost and reversibility. A commercial record is inexpensive, quick to establish and can be abandoned. A robotics cell is neither.

The claim is not that a small manufacturer should never automate. It is that the sequence favours knowing what the business has promised and to whom before committing capital to producing it faster.

What this argument does not establish

The 8.7% and 45.1% figures come from the same survey programme and are self-reported by firms (United States Census Bureau 2020). Self-reported adoption is imprecise in both directions, particularly for a category like robotics where the boundary between automated equipment and a robot is not obvious to a respondent.

The concentration interpretation is Census's own and is well supported, but this paper does not decompose the exposure figure by firm size, because the published tabulations cited do not report it at that granularity. The five-fold gap is consistent with concentration and is not, by itself, proof of it.

No causal claim is made that improving a commercial record improves manufacturing firm performance. The argument defended is about capability and sequence: certain questions cannot be answered without certain records, and an investment whose return depends on demand stability is better made by a firm that can see its demand. Whether a given firm profits from either is not established here.

Guo (2025) concerns firms that adopted Industry 4.0 technologies, and its productivity findings apply to them. Extending those findings to the majority that has not adopted would be unsupported, and no such extension is made.

No manufacturer's results appear in this paper, and no figure derives from any engagement.

Conclusion

8.7% and 45.1% are both correct, and they answer different questions. One describes an automating workforce; the other describes a sector where most firms have adopted nothing of the kind. Quoting the second as evidence of the first makes the majority of manufacturers invisible in their own industry's account of itself.

For those firms, the useful question is not which advanced technology to buy. It is whether the business can state what it quoted, what it lost and why, what it promised, and which customers have quietly stopped ordering. Those answers come from records rather than machines, and they cost very little by comparison.

References

Guo, L. (2025). Adoption of Industry 4.0 technologies and total factor productivity: Firm-level evidence. Journal of Manufacturing Technology Management, 36(3). https://doi.org/10.1108/jmtm-08-2024-0439

United States Census Bureau. (2020). Advanced technologies adoption and use by U.S. firms: Evidence from the Annual Business Survey (CES-WP-20-40). https://www2.census.gov/ces/wp/2020/CES-WP-20-40.pdf

United States Census Bureau. (2022). A firm-level view from the 2019 Annual Business Survey (CES-WP-22-12R). https://www2.census.gov/ces/wp/2022/CES-WP-22-12R.pdf

United States Census Bureau. (2023). Three results from recent research on advanced technology use and automation. https://www.census.gov/newsroom/blogs/research-matters/2023/09/advanced-technology-use-and-automation-results.html

United States Census Bureau. Measuring technology and the economy: working papers. https://www.census.gov/topics/business-economy/tech-stats/library/working-papers.html

Conflict of Interest Statement

RevOps HQ is a HubSpot Solutions Partner and is paid to build the commercial systems this paper argues most manufacturers lack. The argument rests on published Census Bureau data and on peer-reviewed productivity research rather than on the firm's own client results, and the records described are platform-independent, but the conclusion favours work the firm sells and should be read accordingly.

Acknowledgments

Adoption and exposure figures are published by the United States Census Bureau through the Annual Business Survey, conducted in partnership with the National Center for Science and Engineering Statistics.