Our read
The record covers three jobs. Cognite has CONFIRMED customer stories at Celanese and NOVA Chemicals for plant data and maintenance work. Augury appears on customer pages at DuPont, IFF and others, all OBSERVED, with one CEO quote on uptime. Neural Concept is for engineering simulation, with one CONFIRMED press-release mention (Multi-Wing) and OBSERVED mentions at Eaton and GE. Pick by job; a logo is not usage.
Which tool for which job?
| Your job | Documented example | The reference to read | What it documents |
|---|---|---|---|
| Build a single data portal for plant experts |  Cognite |  Celanese | Celanese uses Cognite Data Fusion to build an integrated manufacturing portal that serves as one repository for its domain experts. 1 |
| Improve operator rounds | Cognite | Celanese | Celanese uses a digital worker application built on Cognite to improve operator rounds. 1 |
| Trend and analyze industrial data without code | Cognite | Celanese | Celanese uses no-code data analytics on Cognite to explore, trend and analyze industrial data. 1 |
| Find plant files and plan maintenance | Cognite |  NOVA Chemicals | NOVA Chemicals ingested one plant's data into Cognite for smart maintenance management and says locating a file takes about a minute. 2 |
| Keep machines running (uptime) |  Augury | Circulus | Circulus's CEO reports uptime of 85 to 90% today versus 65 to 70% before Augury, a vendor-published statement. 3 |
What do manufacturers publicly do with industrial AI?
The named examples fall into three different jobs. Cognite is documented in customer stories as a way to unify and contextualize plant data for maintenance and operations work. Augury is presented as machine-health monitoring with predictive alerts for rotating equipment. Neural Concept is an engineering tool that plugs into CAD and simulation software to predict design performance early. These are not three versions of the same product, so the first decision is the job, not the logo count.
The proof levels differ sharply. As of 2026-11-10, usedby's Manufacturing view of industrial AI lists 48 companies: 9 CONFIRMED by a case study, quote or press release, and 39 OBSERVED, meaning the vendor shows the company on its own pages. An OBSERVED line says a company appears as a customer. It does not say which plant, which machines or what result.
How do Celanese and NOVA Chemicals use Cognite for plant data?
Celanese uses Cognite Data Fusion to build an integrated manufacturing portal that serves as a single repository for its domain experts. Its customer story lists a digital twin with recommendation models, a digital worker application for operator rounds, no-code analytics to trend industrial data, and robotics enablement. The company's Director of Digital Manufacturing describes the aim as putting the right data in experts' hands so they spend less time identifying problems.
NOVA Chemicals describes a narrower start. It focused on one plant, ingested data into Cognite and linked the records, and its digital transformation manager says locating a file now takes about a minute instead of hours of calling subject-matter experts. The company applies this to smart maintenance management, and a usedby summary of its story also lists isolation planning, troubleshooting and operations support, with AI for routine inspections still being explored.
What changes for the team is mostly retrieval and context: drawings, operating data and maintenance history sit in one place before any model is applied. For a smaller plant, that points to a data-readiness question before an AI question. Cognite's privacy notice adds that Atlas AI agents query the knowledge graph and can store conversation history that is shareable with project members, and that a data processing agreement is available.
What do Augury customer mentions tell a maintenance team?
Augury describes Machine Health diagnostics for rotating equipment, predictive alerts with root-cause analysis and a Reliability Agent for maintenance workflows. The manufacturers on its customer pages, including DuPont, IFF, Oil-Dri Corporation of America, Advancion and Baldwin Richardson Foods, are all OBSERVED: the vendor shows them, and the pages name no task. A logo says nothing about asset type, coverage or outcome.
Two named quotes add detail. A senior reliability engineer at DuPont Nomex calls the technology a break from the past, without stating a task. Circulus's CEO says average uptime was around 65 to 70% before Augury and is 85 to 90% today; that is a vendor-published customer statement, not an independent measurement, and it does not isolate Augury from other changes.
Where does Neural Concept fit, and how strong is the record?
Neural Concept targets engineering teams at OEMs and suppliers, with models that predict multi-physics behaviour from CAD and generative CAD. Its usedby profile lists NX, CATIA, SolidWorks, Ansys and Abaqus among tools it works with. In the sector, Multi-Wing is CONFIRMED through a 2025 press release that names it among customers who scaled their use; the release gives no task for Multi-Wing. Eaton and GE are OBSERVED only.
The same release reports, as vendor claims, that MAHLE achieved 15% more efficiency and 4 dB less noise in a blower by embedding the AI in its existing systems, and that General Motors built a crash-safety model. Those are other companies' results, not evidence for Multi-Wing, Eaton or GE.
What do the English-language peers imply for a reader in the US or Canada?
Celanese (US) and NOVA Chemicals (CA) have CONFIRMED lines with Cognite and a described workflow. DuPont, Baldwin Richardson Foods, IFF, Oil-Dri Corporation of America and Advancion (all US) are OBSERVED with Augury, and GE (US) is OBSERVED with Neural Concept. The OBSERVED names appear only as named customers, not as documented projects, so they show who to ask about, not what was done.
What should a smaller plant check before copying a workflow?
Editorial recommendation, not a vendor capability: before reusing the Celanese or NOVA Chemicals pattern, list where your drawings, historian data and maintenance records live and who maintains the links between them. Before pursuing machine-health monitoring, name the rotating assets and the person who acts on an alert. Before engineering AI, confirm that your CAD and simulation tools are among those the vendor supports and that you have past simulation results to learn from. The public record does not show the effort, cost or results of any of these in a plant your size.
Method and limits
2026-11-10
Scope: 48 companies classified in Manufacturing on usedby with a public line on a Industrial & construction AI tool. Every count on this page covers this sector only, not the tools' whole customer base. Company size is not recorded and is not claimed.
This dated snapshot counts distinct companies in public customer evidence. CONFIRMED means a case study, named testimonial or press source; DECLARED means the company itself told usedby; OBSERVED means a vendor listing; SIGNAL means a job post. All four count, but their strength differs. Removed and unlevelled lines are excluded. Examples are a capped selection verified within 90 days. Vendor facts are sourced records checked within 60 days. Coverage is uneven and missing evidence is unknown. These figures measure neither market share nor product performance; capture dates are not adoption dates.
