AI Career Promises Need Proof That Workers Are Getting Ahead

Bloomberg reported on October 21 that Electronic Arts planned roughly 400 job cuts and the closure of BioWare. [1] I think that news exposes a weakness in optimistic claims about AI careers: evidence that someone can do more tasks tells us little about whether an employer will offer them a job.

A workplace can become more capable while its doors become harder to enter. That possibility deserves as much attention as a faster workflow. My standard for progress is whether people gain access to worthwhile paid work, advancement or a better working day.

The distinction matters when reading both AI usage research and announcements promising wider access to training.

The research begins inside the workplace

OpenAI's September 16 research summary describes an analysis of more than 1.5 million work-related ChatGPT messages from April through July 2026. One analysis concerned roughly 6,200 workers whose use was observed consistently across those months. The company identified occupations using information provided during ChatGPT Business onboarding. [2]

OpenAI reported that some users repeatedly returned to tasks associated with other occupations. Those findings describe activity in its sample. They do not establish how many additional people were hired, whether participants received promotions, or what happened to unsuccessful applicants. [2]

That boundary leaves the findings useful for understanding work design. It also leaves employment outcomes unresolved.

Imagine an employee using AI to take on marketing work that previously required a separate assignment. In one hypothetical company, that helps a growing team launch a product and recruit colleagues. In another, it lets management leave a vacancy unfilled. The same broad pattern of tool use could accompany either outcome.

My concern is what happens when the first outcome is assumed without evidence. A person already inside an organization may gain useful capabilities while someone outside loses a possible route in.

A closure removes the place where skills become employment

The reported EA cuts do not establish that AI replaced those workers. I would not infer a technical verdict on game developers from a decision to close a studio. [1]

What matters for this argument is that a job depends on an organization choosing to fund work. A worker's abilities can remain valuable after that choice changes.

The bleak possibility is that some people become better equipped for work at the same time that fewer employers are willing to pay them to do it. I consider that a serious career problem even if productivity improves for those still employed.

I also think a closure has consequences that a count of individual skills misses: the loss of a team in which people could collaborate, learn and build a record of completed work. A training session can teach a technique. Rebuilding that working environment requires projects, colleagues and someone willing to employ them.

That is why I want discussions of AI opportunity to account for access to organizations where skills can actually become careers.

A training target needs an outcome beside it

In its September 23 announcement with Grab, OpenAI described a plan to reach 30,000 driver, delivery and merchant partners across Southeast Asia over two years. The program includes practical workshops intended to help participants apply AI to their work and business ideas. Those are the organizers' plans, rather than independently demonstrated results. [3]

I favor making useful training accessible. But a participation target cannot tell a job seeker how many people will secure employment or progress in a career. The announcement does not provide those results. [3]

A skills program deserves credit for teaching skills. Broader claims about economic opportunity need broader evidence.

Because this program includes people running businesses or doing platform work, hiring alone would be an incomplete test. I would want to know whether participants save time on recurring work, secure additional paid assignments or develop activities that remain useful after the workshop.

I would also want follow-up to include people who stop using the tools or cannot turn training into an opportunity. Reporting only the most successful participants would leave precisely the people who need help least visible.

A program announced last month deserves time to produce results. I would judge later claims of success by those outcomes, while treating today's target as a target.

An employer partnership deserves closer reading

For a job seeker comparing training programs, I would focus on what an advertised employer partnership actually provides. Recruitment into defined positions, a chance to apply, and a company supplying course material are different arrangements. A familiar logo does not tell you which one is on offer.

If a provider publishes placement results, I would look for the total number of participants, how long they were followed and what counted as employment. A percentage covering only people who answered a survey deserves different weight from a result covering the full cohort.

That approach may lead to an uncomfortable answer: a course can be educationally useful while offering little evidence of a route into work. I would choose it for the skill it teaches, within a bounded learning plan. I would not let an unsupported career promise dictate a major change of direction.

Access matters too. A role requiring prior experience may offer scant immediate help to a newcomer, even when it appears in a program's list of employer connections. Career claims should specify whom the opportunities are open to.

The forecast needs room for fewer openings

My expectation is that some employers will use broader employee capabilities to absorb work within existing teams. Others may use lower production costs to attempt projects they previously could not justify. Both paths are plausible; these sources do not determine which will generate more employment.

I am therefore more confident that AI can change the content of jobs than that it will improve access to them. For someone trying to enter or reenter a career, that is a hard distinction with practical consequences.

I would take claims of an AI employment success seriously when they account for people getting hired, moving into better roles and remaining in paid work. The people who never make it into the workplace deserve to appear in the assessment.