AI and Virtual Reality in Training: When the Combination Makes Sense and When It Is Tech Theater

How many companies already use AI?

About one in five U.S. businesses, and more than a third of large ones. As of May 3, 2026, 19.8% of U.S. businesses reported using AI, along with 37% of firms with at least 250 employees and 32% of those with 100 to 249 (U.S. Census Bureau, Business Trends and Outlook Survey).

In Mexico, where many U.S. manufacturers run plants, large companies moved faster: in June 2026, 48.5% of firms with more than 100 workers used at least one AI technology, up from 24.3% in September 2025 (Banco de México, regional economic report for April to June 2026, box 4). The bank notes that the survey questions changed, so the comparison is indicative.

Gartner predicts that by the end of 2026, 40% of enterprise applications will include task-specific AI agents, up from less than 5% in 2025. AI reaches training later, and that is the moment to ask what job it does.

What does each technology do in a training program?

Virtual reality solves a specific problem: it lets people practice situations that are expensive, dangerous or impossible to repeat at will on the plant floor. An operator can get a lockout/tagout procedure wrong twenty times inside a simulator without anyone getting hurt.

Artificial intelligence solves another one: personalization at scale. A good instructor adapts the exercise to each participant but cannot be in two hundred sessions at once; an AI system can adjust the difficulty, analyze each decision and give individual feedback.

Inside a training program, AI can do four different jobs, and each one is judged differently:

AI's job	When it is worth it	What to ask the vendor
Producing content: scripts, voices, 3D environments	When it lowers cost or lead time. It is production: the trainee never sees it	How much did the price drop because of it?
Adapting practice to each person	When a competency model exists first: which behaviors matter and what doing them well looks like	Against which behaviors does the scenario adapt, and who defined them?
Assessing each attempt	When someone will read the results and decide something: who needs reinforcement, what to correct	Who reviews the score before it decides anything about a person?
Talking with an avatar	In soft skills: sales, service, leadership, a difficult conversation	How do you check that the avatar does not teach something wrong?

On a plant floor, the value almost always sits in the two middle rows. The first one makes production cheaper; the last one pays off more in offices than on the floor.

Is super intelligence the same as AI?

In U.S. federal usage, yes. Executive Order 14434, signed on September 29, 2026, directs federal agencies to use "Super Intelligence" and "SI" where they used "artificial intelligence" and "AI" in official communications, and it defines the new term with the existing statutory definition of AI (15 U.S.C. 9401(3)). It also gives the president's science and technology adviser 60 days to propose legislative language for a federal definition. The order applies to the executive branch.

In the technical sense, a superintelligence would outperform people at almost any intellectual task, and no system does that yet. If a training proposal promises super intelligence, the questions in this article still apply: what job does the model do, and with what data?

Which companies already combine AI and VR?

At BMW's plant in San Luis Potosí, Mexico, apprentices use virtual reality to learn welding and painting, and quality assurance relies on AIQX, a system developed by BMW that identifies anomalies or defects (Mexico Industry, November 2023). That is the most common pattern in public cases: VR trains and AI inspects, each in its own process.

At Fyware, AI is part of GIRI: its AI authoring (in beta) turns a procedure in PDF, Word or video into step-by-step instructions. In Fyware Safe, the forklift simulator exports each session's results so the instructor can decide who needs reinforcement. Bringing both technologies into the same simulator first requires the design work described in the next section.

Three signs of tech theater

In most of the proposals I have reviewed, the words "artificial intelligence" are there to sell. These are the three signs I look for, in my own projects or anyone else's:

1. The AI never touches the learning experience. It produces scripts, voices or environments, and the trainee lives the same linear experience as with handcrafted content: traditional training made with modern tools, and it should be priced that way.
2. The system adapts, but nobody defined against what. An adaptive scenario needs a competency model, and that diagnostic and instructional work happens before anyone writes a line of code. If the answer to "Which behaviors does the system evaluate, and who defined them?" is vague, the adaptation is cosmetic.
3. The data gets collected but never used. If the "AI-powered analytics" report ends up in a folder nobody opens, it does not exist: someone has to read it to redesign a module, reinforce a person or correct a procedure.

What operator data does it generate, and who sees it?

A performance score that identifies a worker is personal information. In California, the exemption for employment-related personal information under the California Consumer Privacy Act expired on December 31, 2022, so employees have the same rights over that data as any other consumer.

In New York City, Local Law 144 bars employers from using an automated employment decision tool unless it has had a bias audit within the past year, a summary of the results is public, and employees or candidates have been notified. If a simulator's AI scores an operator, have a person review that score before it decides a promotion, a sanction or an internal certification.

If the plant network cannot send data out, Fyware Hub monitors the headsets without internet, and the instructor sees what each operator sees.

The questions that put the decision in order

The right sequence starts with the diagnosis:

• Which competency do we want to develop, and what does doing it well look like on the job?
• Is it best learned by practicing? If not, VR is probably unnecessary.
• Does practice need to vary by person or by scenario? If not, AI is probably unnecessary.
• Who will use the data the system generates, and to decide what?
• Against what baseline will we measure the result?

All five can be answered before seeing a single demo, just like the five questions that separate an operations integrator from a content studio. When AI has an identifiable job, the return holds up in front of the board. When it does not, the company is paying for two buzzwords on the same invoice.

Frequently asked questions

What is AI and VR training?

It is training in which a person practices in a virtual reality simulator while an AI system adapts the scenario, assesses each attempt or holds a conversation with them.

Is super intelligence the same as artificial intelligence?

In U.S. federal usage, yes: Executive Order 14434, signed on September 29, 2026, tells federal agencies to say "Super Intelligence" (SI) instead of "artificial intelligence" and defines it with the same statutory definition of AI. In the technical sense, a superintelligence would outperform people at almost any intellectual task, and no system does that yet.

How many U.S. companies use AI?

As of May 3, 2026, 19.8% of U.S. businesses reported using AI, and 37% of firms with at least 250 employees, according to the Census Bureau's Business Trends and Outlook Survey.

Is an operator's training performance data personal information?

Yes, if it identifies the worker. In California, the CCPA's exemption for employment-related personal information expired on December 31, 2022, and in New York City, Local Law 144 requires a bias audit and notice before an employer uses an automated employment decision tool.

Where does Fyware use AI?

In GIRI, whose AI authoring (in beta) turns a procedure in PDF, Word or video into steps ready to run. Fyware Safe's simulators deliver practice of the dangerous step, and the forklift simulator exports each session's results.