Skip to content

Winning the Global Race for Artificial Intelligence Expertise

How the Executive Branch Can Streamline U.S. Immigration Options for AI Talent

By Doug Rand and Lindsay Milliken

April 9, 2021

New applications of artificial intelligence (AI) are expanding rapidly, while countries all over the world are fiercely competing for the talent necessary to take advantage of these increasingly consequential technologies.1See Appendix 5 for a reading list on the economic and national security importance of AI. The United States is still one of the top destinations for AI students and professionals, but it may not stay that way for long. Many countries, such as Canada and the United Kingdom, among others, are adapting their immigration systems to make it easier for AI experts to study, work, and stay permanently.

Meanwhile, the United States’ often rigid and confusing immigration policies make it difficult for AI professionals and students to stay in the country after they complete their education or try to change jobs. If this continues, countries like China, which is providing direct financial incentives to attract global AI talent, could gain an economic and national security edge over the United States.

The most enduring solutions to these challenges should come from Congress. As a failsafe, however, the federal government can adapt existing immigration policies to make it easier for those with valuable AI expertise to study and stay in the United States.

This paper evaluates the full suite of available permanent residency pathways and recommends concrete policy changes to make global AI talent recruitment more streamlined and predictable. Our analysis focuses on permanent residency because this is the best first step toward U.S. citizenship and full integration into the society and economy of the United States.2Temporary (“nonimmigrant”) work visas come with greater restrictions on when, where, and for whom the individual can work, and have an end date after which the individual must leave the United States. Permanent residency allows the individual to work and live anywhere in the United States, freely travel outside the country, and provides eligibility for U.S. citizenship in three to five years.

The Case for Green Cards Over Temporary Visas

Permanent residents—unlike those on temporary student or work visas—can work for any employer and are eligible for U.S. citizenship and full civic integration after three to five years of obtaining permanent residency (a “green card”).

Under the current immigration system, employment-based green cards are unduly scarce: 140,000 per year, fewer than half of which go to the principal worker. Many U.S. employers are thus compelled to rely on the flawed, controversial, and temporary H-1B visa for “specialty occupation” professionals. The chief critique of the H-1B system is that professionals are “indentured” to potentially exploitative employers for years or even decades at a time, with limited freedom to change employers. If underpaid, these professionals cannot leave their job or bargain for better wages without risking revocation of the employer’s green card sponsorship—or even firing and enforced departure from the United States. The H-1B system is problematic for most employers, as well, with a consistently oversubscribed “lottery” of 85,000 visas each year (of which 20,000 are reserved for advanced degree holders from U.S. universities).

If it were easier for U.S. employers to sponsor global talent for a green card, as opposed to an H-1B, more professionals would obtain permanent residency and the ability to freely change jobs and negotiate on par with U.S. employees. The H-1B program could then perform its originally intended function as a vehicle for truly temporary high-skill work needs.

Overview of Immigration Pathways for AI Professionals

One of the biggest barriers to global AI talent recruitment is the sheer time required to obtain a green card. Even if the Department of Homeland Security (DHS) had instantaneous processing of employment-based green card paperwork—which, to be clear, it does not—there is another lengthy process to navigate at an entirely different federal agency.

The labor certification process, managed by the U.S. Department of Labor (DOL) through its Program Electronic Review Management (PERM) system, is intended to establish that a given applicant will “not adversely affect the job opportunities, wages and working conditions of U.S. workers.”

To obtain this PERM labor certification on behalf of an employee, a sponsoring company must first attempt to recruit alternative candidates (including publishing job advertisements in print newspapers), request a “prevailing wage determination” using DOL data, and submit a complicated form. This entire process can take several months to almost a year. This is far longer than most companies—and most sought-after candidates—can wait to determine whether to engage in an employment agreement. Thus, the PERM labor certification requirement is one major hurdle that has made certain temporary statuses (chiefly the H-1B) into de facto prerequisites for skilled foreign professionals to get in the door with U.S. employers.

But existing immigration law includes several pathways for highly skilled individuals considered so self-evidently desirable for the United States that they can skip the typical labor certification process.

There are at least three pathways for skilled professionals with no labor certification requirement, and two additional pathways with a simplified labor certification.3For those seeking O-1, EB-1A, and EB-1B status, there is no requirement to complete the Department of Labor’s ETA Form 9089, the “Application for Permanent Employment Certification.” Form 9089 does not need to be certified by DOL for those seeking EB-2 status with a National Interest Waiver (as they are exempt from DOL labor certification) or EB-3 status with a Schedule A occupation (as they are “pre-certified” by DOL). In these latter two scenarios, an “uncertified labor certification” (i.e. Form 9089) may be directly filed with USCIS, with no DOL review or recruitment requirements. Some of these pathways also allow the individual to apply for immigration status directly, without relying on an employer to sponsor them (also known as “self-petitioning”).

Table 1. Immigration categories with fast-track potential

CategoryEligibilitySponsor required?Labor certification?CapTarget
O-1   (temporary status)Extraordinary abilityYes    No labor certificationUncappedTop of field
EB-1A   (green card)Extraordinary abilityNo    No labor certification40,040   (across all EB-1)Top of field
EB-1B   (green card)Outstanding professors and researchersYes    No labor certification40,040   (across all EB-1)Top of field
EB-2 with National Interest Waiver   (green card)Exceptional abilityNo    No DOL Review40,040   (across all EB-2)Above- average achievement
EB-3 with Schedule A designation   (green card)Bachelor’s degree or work equivalentYes    No DOL Review40,040   (across all EB-3)Shortage occupations

These five pathways should be appealing for AI professionals with varying levels of experience and expertise, but the eligibility requirements are too broad and vague. The regulations make no reference to particular technologies or research areas; indeed, some eligibility criteria are intended to sweep in applicants across fields as disparate as “the sciences, arts, education, business, or athletics.”

Even without modifying these underlying regulations, U.S. Citizenship and Immigration Services (USCIS) could update its Policy Manual to provide clear guideposts for AI professionals seeking to demonstrate their eligibility for one of these categories. While USCIS is typically constrained from using binary yes-or-no eligibility criteria, there are still numerous ways to provide greater clarity and predictability for applicants.

How Adjudications Work Today

The typical adjudication process for EB-1 and EB-2 green card applications comes from a 2010 USCIS policy memorandum, which is based on the administrative appeals decision made in Kazarian v. USCIS that same year. This guidance states that USCIS adjudicators should use a two-step process for evaluating petitions for individuals of extraordinary ability; outstanding professors or researchers; and individuals of exceptional ability. This process is conducted as follows:

  1. “USCIS officers should first objectively evaluate each type of evidence submitted to determine if it meets the parameters applicable to that type of evidence described in the regulations (also referred to as ‘regulatory criteria’).”
  2. “USCIS officers then should consider all of the evidence in totality in making the final merits determination regarding the required high level of expertise for the immigrant classification.”

The standard of proof for these adjudications is that there must be a “preponderance of evidence” to show that the individual is “extraordinary,” “outstanding,” or “exceptional.” This standard means that the evidence must prove this claim has a “greater than 50 percent chance” of occurring; in other words, that it is more likely than not to be true. After an applicant4While the individual seeking a green card is technically called a “beneficiary” with respect to submitting an I-140 petition, this individual will ultimately be the applicant for a green card and we use this term instead to provide clarity to the lay reader. submits this evidence, and after a USCIS officer determines there is a preponderance of evidence that the applicant meets the specific requirements for EB-1 or EB-2 status, adjudicators must then evaluate the evidence as a whole to ascertain whether the applicant has “the required high level of expertise for the immigrant classification.”

Unfortunately, this guidance is not sufficient to guarantee uniformity in the adjudication process. The denial rate for EB-1s has increased dramatically—by about 25 percentage points—since 2017. If the current guidance were sufficient to appropriately and fairly adjudicate EB-1 and EB-2 petitions, it is unlikely that there would be any dramatic increases in denials based on the administration at the time.

In particular, immigration attorneys have cited several problematic denials, such as USCIS’s conclusion that venture capital does not constitute an “award”;5The receipt of an “award,” which is a requirement for O-1s and EB-1s, includes accomplishments such as a Nobel Prize or other nationally or internationally-recognized achievement that demonstrates extraordinary ability in the applicant’s field. See Appendices 1 and 2 for more information. disregarding detailed and consistent support letters from independent experts because they are not corroborated by “documentary evidence”; determining a role was not critical to a company because the applicant was not senior enough; and insisting that published materials are not about the applicant if they describe the individual’s work and not them personally, among many other causes for concern. In one case, USCIS even rejected a Nobel laureate because the adjudicator claimed to need more information on the interpreter who translated the Nobel citation.

It is clear that more effective guidance is needed to provide for rational and consistent adjudications. To that end, USCIS should:

  • Clarify in the USCIS Policy Manual that “preponderance of evidence” must be from the perspective of a reasonable person.
  • Provide a list of evidentiary possibilities that will per se satisfy the “preponderance of evidence” standard for each eligibility criterion, starting with the field of AI.
  • Review and update this list of evidentiary possibilities, on a regular schedule, with the input from a DHS subcommittee authorized under the Federal Advisory Committee Act (FACA) and composed of relevant experts in immigration law, higher education, and highly in-demand occupations.
  • Create a public-facing resource that summarizes these criteria for a general audience with limited legal experience, comparable to the early-2010s Entrepreneur Pathways guide.

With this proposed framework in mind, the remainder of this paper will focus on the second element above—the development of a list of reasonable evidentiary criteria that should per se meet the “preponderance of evidence” standard for future beneficiaries in the field of AI.

It is important to note that while this paper focuses mainly on lawful permanent residency requirements, nonimmigrant pathways, such as the O-1, are still an option for AI professionals. O-1 temporary status has similar requirements to EB-1s (described in more detail below) and may be a better fit for certain AI professionals seeking to live in the United States, especially those from India and China who face lengthy green card backlogs due to statutory caps based on country of origin. In fact, the final report of the National Security Commission on Artificial Intelligence (NSCAI) recommended that the scope of O-1s be broadened to be more accessible for AI professionals. The current requirements focus on largely academic achievements such as publications in major research outlets which are not “well-suited for people who excel in industry.” Broadening the scope of O-1s, in addition to our recommendations below on EBs, would provide a valuable pipeline of AI talent to the United States.

Defining “Extraordinary Ability” in AI (EB-1A)

EB-1A status is for individuals who possess “extraordinary ability in the sciences, arts, education, business, or athletics through sustained national or international acclaim.” For an AI expert to qualify for an EB-1A, they must typically demonstrate that they are in the top 20 percent of their field.6Outlined in the meeting minutes of the American Immigration Lawyers Association (AILA) Liaison – Nebraska Service Center Liaison Meeting on November 4, 2010 (AILA InfoNet Doc. No. 10121562). To be approved for an EB-1A, applicants must have received “a major, internationally-recognized award, such as a Nobel Prize” or—much more commonly—at least three of the following achievements:

  1. Receipt of nationally or internationally recognized prizes or awards for excellence in the field of endeavor;
  2. Membership in associations in the field for which classification is sought, which require outstanding achievements of their members, as judged by recognized national or international experts in their disciplines or fields;
  3. Published material in professional or major trade publications or major media about the individual, relating to the individual’s work in the field for which classification is sought, which shall include the title, date, and author of such published material, and any necessary translation;
  4. Participation on a panel, or individually, as a judge of the work of others in the same or in an allied field of specialization to that for which classification is sought;
  5. Original scientific, scholarly, or business-related contributions of major significance in the field;
  6. Authorship of scholarly articles in the field, in professional journals, or other major media;
  7. Employment in a critical or essential capacity for organizations and establishments that have a distinguished reputation;
  8. Evidence that the individual has either commanded a high salary or will command a high salary or other remuneration for services as evidenced by contracts or other reliable evidence.

The order of the criteria can be confusing, particularly regarding the requirements for published research. To satisfy criterion 6, an applicant simply has to show that they have published their work in a scholarly journal relevant to their field. However, to satisfy criterion 5, an applicant must prove that their work is of “major significance” to their field, such that it has “provoked widespread commentary or received notice from others working in the field” or a “goodly number” of entries in a citation index that prove the applicant’s work is authoritative—a decidedly more difficult requirement to meet.

Because an applicant only needs to provide evidence for three out of the eight criteria above, there are many ways an AI professional can qualify for an EB-1A. For example, a person who would meet three of the criteria could be someone who is (1) a member of the Association for the Advancement of Artificial Intelligence (membership in an association which requires outstanding achievements of their members), (2) has served on a Ph.D. dissertation committee for AI applications in healthcare (participation on a panel as a judge of the work of others in the same or in an allied field), and (3) has published original work in scholarly journals, such as the Institute of Electrical and Electronics Engineers (IEEE) Transactions on Human-Machine Systems and Artificial Intelligence in Medicine, and has cultivated a time-adjusted h-index in the top 20 percent of their field (original scientific contributions of major significance in the field).

The h-index is a calculation of the impact of a researcher’s publications. For example, having an h-index of 40 means the researcher has had 40 papers which have all been cited 40 or more times. It is a relatively simple measure to determine the impact of a scientist’s work in their field. However, h-indices can vary widely depending on the scientific field. Also, without accounting for the length of time the scientist’s papers have been published, it can skew towards those later in their careers and introduce a threshold that is impossible to meet for younger scientists. Adjudicators could use h-indices to help determine whether an applicant has made original contributions of “major significance” in their field, but the index value must be judged based on the typical range for the applicant’s field and it must also be adjusted for the length of time the applicant’s contributions have been published to ensure that earlier-career scientists are not unduly rejected.

These types of qualifying achievements are not solely reserved for experts in computer science. AI has hundreds of applications and the evidentiary standards suggested in this paper should apply to professionals in these AI-adjacent fields.

Adjudicators should also consider alternative ways that an applicant could be deemed extraordinary, such as a startup founder with venture capital funding. In 2019, 770,609 new businesses were founded in the United States. Out of those businesses, a little more than 4,700 received venture capital investments and about 63,730 received angel funding. This amounts to less than 9 percent of new businesses receiving investment funds per year (though it is likely quite a bit less, since many of these funding rounds go to companies that were founded in prior years). Given that the standard for “extraordinary ability” is being in the top 20 percent of the field, receiving angel or venture capital funding should satisfy the first criterion as a “nationally recognized award.”

Further evidentiary standards for the EB-1A pathway are suggested in Appendix 1.

Defining “Outstanding Professors and Researchers” in AI (EB-1B)

The EB-1B green card category is reserved for academics who have at least three years of relevant teaching or research experience, have received international recognition for their work, and plan to pursue a tenure-track job at a university or comparable research position with a private employer. To qualify for an EB-1B, the applicant must satisfy at least two of the six following criteria:

  1. Evidence of receipt of major prizes or awards for outstanding achievement;
  2. Evidence of membership in associations that require their members to demonstrate outstanding achievement;
  3. Evidence of published material in professional publications written by others about the individual’s work in the academic field;
  4. Evidence of participation, either on a panel or individually, as a judge of the work of others in the same or allied academic field;
  5. Evidence of original scientific or scholarly research contributions in the field; and
  6. Evidence of authorship of scholarly books or articles (in scholarly journals with international circulation) in the field.

Interestingly, to satisfy criterion 5 and show that the applicant has contributed scholarly research to their field, current USCIS guidance notes that applicants can use tools such as GoogleScholar, SciFinder, and the Web of Science to establish the number of citations and the impact factor for the journals in which they have published. This does not take into account, though, that many significant publications for AI are not in peer-reviewed outlets, but are on public-access sites such as arXiv. To appropriately judge an applicant for an EB-1B based on their scholarly work, adjudicators must keep in mind that specific outlets are not exclusive signals of quality by which to judge the significance of the applicant’s publications.

As demand grows for AI professionals, the need for well-developed academic programs and experienced professors in the United States grows as well. Enrollment in AI-focused introductory courses at U.S. institutions of higher education in 2017 was up to five times higher than in 2012. Increases in demand for AI graduate education have occurred as well, but between 2007 and 2017, the capacity of instructors has not grown to accommodate it.

To increase adjudicators’ familiarity with achievements in AI that map to EB-1B requirements, further details can be found in Appendix 2.

Defining “Exceptional Ability” and the National Interest in AI (EB-2)

EB-2 green cards have two different subcategories of eligible beneficiaries: individuals who have advanced degrees and individuals of “exceptional ability.” In either case, such individuals can also seek a National Interest Waiver (NIW), given to those whose work in the United States would have “substantial merit and national importance.” Individuals approved for an NIW do not need an employer to sponsor their application and, if approved, are able to skip the Department of Labor’s PERM process, which other EB-2 beneficiaries must undertake. The PERM process is intended to determine whether bringing a foreign worker to the United States would negatively impact the salary or working conditions of U.S. professionals in that same occupation.

Advanced Degree Requirements

To qualify for an EB-2 green card based on an advanced degree, an applicant must provide:

  • An official academic record proving that the applicant has an advanced degree from an U.S. institution (or foreign equivalent); or
  • An official academic record proving that the applicant has a bachelor’s degree from a U.S. institution (or foreign equivalent) and letters from current or former employers showing that the applicant has had at least five years of “progressive post-baccalaureate work experience.”

Exceptional Ability Requirements

Beneficiaries for the other category of EB-2 green card “must be able to show exceptional ability in the sciences, arts, or business.” According to the USCIS Policy Manual, this means that the applicant must have “a degree of expertise significantly above that ordinarily encountered.” While the guidance does not define “ordinarily encountered,” it can be inferred that this guidance means applicants for “exceptional ability” EB-2s typically require achievement above the median for the applicant’s field. Applicants must satisfy at least three out of the six criteria below:

  • An official academic record showing that the individual has a degree, diploma, certificate, or similar award from a college, university, school, or other institution of learning relating to the area of exceptional ability;
  • Evidence in the form of letter(s) from current or former employer(s) showing that the individual has at least ten years of full-time experience in the occupation for which he or she is being sought;
  • A license to practice the profession or certification for a particular profession or occupation;
  • Evidence that the individual has commanded a salary, or other remuneration for services, which demonstrates exceptional ability which is relative to others working in the same field;
  • Evidence of membership in professional associations; and
  • Evidence of recognition for achievements and significant contributions to the industry or field by peers, governmental entities, or professional or business organizations.

National Interest Waiver Requirements

If an applicant seeks to obtain a national interest waiver (NIW) along with EB-2 status, they must first provide the above evidence that they have an advanced degree (or a bachelor’s with five years of progressive experience) or demonstrate exceptional ability. But to demonstrate “national interest,” there is no similar menu of criteria to be satisfied; instead, there are three required conditions. Prospective NIW recipients must show that:

  • Their “proposed endeavor has both substantial merit and national importance,”
  • They are “well positioned to advance the proposed endeavor,” and
  • “It would be beneficial to the United States to waive the requirements of a job offer, and thus the labor certification.”

Multiple federal agencies have declared AI to be a research and development priority for the United States. It has also been shown repeatedly that there is intense demand for AI professionals in the United States, such that there is likely a worsening labor shortage, which could be alleviated by admitting more AI professionals to the country. Because of this evidence, it would be highly beneficial for the country if USCIS approved more AI professionals for NIWs. The most encouraging approach for boosting the approval rates of AI professionals’ applications would be for USCIS to reaffirm the consensus that AI is a highly important endeavor with “substantial merit and national importance,” creating a presumption that “it would be beneficial to the United States to waive the requirements of a job offer, and thus the labor certification” for AI professionals. It would still be necessary for each NIW applicant to demonstrate that they are “well positioned to advance the proposed endeavor.”

It is important to note that not all AI professionals are renowned Ph.D. scientists performing cutting-edge research. A great many professionals who do not have doctorates are critical to advancing AI applications in various fields, are vital to the successful utilization of AI, and help drive AI-related economic growth. Many of these professionals would more likely qualify for an EB-2 with NIW over an EB-1.

A hypothetical example would be someone who has a master’s degree in computer science from the University of Virginia with a specialization in AI (degree from a university relating to the area of exceptional ability); who is developing computer vision algorithms to detect, via satellites, conditions leading to droughts (letters from peers; substantial merit and national importance for an NIW); and who also is a member of the Crop Science Society of America (membership in a professional association). While this example includes a person with a computer science degree, AI is far broader than just computer science. AI can be applied in all STEM fields and the social sciences, so someone should still qualify as an AI professional with substantial merit even without a computer science degree. Further suggested guidance about how adjudicators could match AI expertise to EB-2 requirements can be found in Appendices 3 and 4.

Defining a Labor Shortage in AI (EB-3)

The EB-3 green card category is open to bachelor’s degree holders and skilled professionals with at least two years of job training, which may describe a relatively large percentage of AI professionals who did not go through advanced academic training. There is currently no way for deserving EB-3 applicants to avoid the time-consuming recruiting requirements of the PERM labor certification process, but the DOL could provide one by updating Schedule A. This is a list of occupations, defined in regulation, where DOL “has determined there are not sufficient U.S. professionals who are able, willing, qualified and available” and that the employment of foreign professionals in such occupations “will not adversely affect the wages and working conditions of U.S. professionals similarly employed.”7The Schedule A list of shortage occupations can apply to EB-2 as well as EB-3 green card applicants. For purposes of this paper, we focus on the interaction between Schedule A and EB-3 applicants, because EB-2 applicants have the National Interest Waiver option available to them. If AI professionals were added to the Schedule A list, however, those with advanced degrees or “exceptional ability” could benefit if using the EB-2 pathway without a National Interest Waiver. 

Currently, Schedule A lists only two specific occupations: physical therapists and professional nurses. This “Group I” list could be updated to include a straightforward definition of “AI professional,” based on a list of degree fields and other objective criteria.

Wage Determinations and Schedule A

When preparing a petition for employment-based immigration status, employers must typically request a prevailing wage determination from the Department of Labor (DOL). This determination informs employers what professionals are paid for performing similar jobs, and any sponsored professionals must be paid similar rates. This wage determination is intended to protect sponsored professionals from being underpaid and to protect U.S. professionals.

DOL determines an appropriate prevailing wage by examining whether the job opening is covered by a collective bargaining agreement or if the listed wage is the “arithmetic mean” of the wages of similar employed professionals. Employers can also determine an appropriate prevailing wage by using either a “survey conducted by an independent authoritative source,” or “another legitimate source of information.”

As described above, the Schedule A list contains specific occupations which DOL determines are in shortage in the United States. Even though Schedule A currently only includes two occupations, physical therapists and nurses, it has been updated numerous times since its creation in 1965. Updating the list to include AI professionals would require DOL to submit a notice of proposed rulemaking (NPRM) to the Federal Register for public comment prior to making the final change. There is evidence that AI professionals are in sufficiently high demand and should be included in Schedule A, particularly given the fact that they are paid high salaries. That said, the question of how to define a labor shortage is far from simple.

The definition of a labor shortage depends in large measure on the researcher studying the topic. A common theme throughout the academic literature is that shortages arise when there are not enough professionals available or willing to work in an occupation at the prevailing wage rates and working conditions for an extended period of time. This can manifest as numerous available job opportunities with rising salaries. It is likely that AI as a field is experiencing a labor shortage because the number of job opportunities has rapidly increased in the past decade and employers are offering significant salaries to attract talent. As of 2019, the number of AI-related job postings on Burning Glass, which scrapes job openings from the major online job boards and categorizes them based on skills needed, geographic location, and wage levels, has tripled since 2010. Meanwhile, 40 percent of states are demonstrating demand for AI-related jobs at more than 1,000 positions per 100,000 working-age people in the state. Salaries for these jobs are quite high, with some companies paying AI professionals hundreds of thousands of dollars. According to a recent Deloitte study, about 20 percent of surveyed companies at every level of AI sophistication said their AI skills gap was either “major” or “extreme.”

On the other hand, many economists dislike using the term “labor shortage,” because in the long run, the cost of labor should rise until demand and supply balance each other out and establish an equilibrium. According to data compiled by Deloitte, for example, the price of labor was not rising enough to merit the designation of a general labor shortage in 2017, although this analysis was economy-wide, not job-specific.

The unemployment rate can also be used as an indicator of a shortage. When the unemployment rate dips below its equilibrium point, where demand and supply are generally equal, it means that there are not enough professionals to fill the available jobs. However, this situation only occurred for about one third of the 20-year period between 1996 and 2016. During the other two-thirds of this period, under this analysis, the overall supply of labor has been greater than demand—again, on an economy-wide and not job-specific basis.

DOL has struggled to quantify the conditions that cause a shortage of professionals, in part because comprehensive employer data is expensive and difficult to collect. In the early 1990s, DOL asked Dr. Malcolm Cohen, a public policy professor at the University of Michigan, to develop a set of indicators which would signal a labor shortage. His set of indicators included recent changes in employment, the occupational unemployment rate, and wage rates; the amount of training needed for the occupation; the demand for replacement professionals; projected increases in occupational demand; and immigrants who have been certified in the occupation in the recent past. Occupations received a score between 1 and 49, with the latter designating an occupation with the most severe shortage. The two occupations that scored the highest (at 39 points) were nurses and physical therapists, which are reflected in the current Schedule A regulations. Unfortunately, DOL never implemented Dr. Cohen’s system to update Schedule A for future shortages.

Given the growing portion of the economy and national security enterprise taking advantage of AI applications, it is advisable that DOL should be guided by Dr. Cohen’s efforts and reevaluate Schedule A, as well as the labor market conditions for AI professionals, to determine whether they should be added to the list.

The next two sections discuss how the UK and Canada periodically evaluate whether specific occupations are experiencing labor shortages, as examples of how DOL could engage in a similar evaluation process in the United States.

How the United Kingdom Measures Labor Shortages

The United Kingdom has a thorough procedure for measuring labor shortages and could be a helpful model for U.S. federal agencies. Labor shortages in the United Kingdom are measured by the UK Migration Advisory Committee (MAC). This committee consists of a chair, five independent economists, and the Home Office, which is in charge of immigration and passports, among other domestic policy priorities. One of the MAC’s duties is to maintain the government’s Shortage Occupation List (SOL) and review it periodically. The SOL is similar to the United States’ Schedule A list but is much more widely used. Employers looking for foreign professionals to fill occupations on the SOL do not have to conduct a Resident Labor Market Test or attempt to recruit domestically. The last full review of the SOL occurred in 2020 and outlined how the SOL would change, as well as the data used to determine labor shortages.

For an occupation to be included on the SOL, there are three requirements: does the job require a high skill level (graduate-level or higher-level vocational qualifications); is the job in shortage; and does it make sense to try to fill the shortage through migration? To determine a shortage, the migration committee uses nine economic indicators from five datasets:

  1. Percentage change of median real pay over one year (Annual Survey of Hours and Earnings (ASHE); Consumer Price Inflation with Housing (CPIH));
  2. Percentage change of median real pay over three years (ASHE; CPIH)
  3. Predicted hourly wage for a set of reference characteristics relative to average predicted wage for the same characteristics over all occupation codes (Annual Population Survey (APS));
  4. Vacancies divided by total employment (European Social Survey (ESS); APS);
  5. Vacancy postings divided by total employment (Burning Glass; APS);
  6. Percentage change of employment level over one year (APS);
  7. Percentage change of median paid hours worked over three years (ASHE);
  8. Change in new hires over one year (APS); and
  9. Weighted stock of unemployment and inactive professionals divided by employed, unemployed, and inactive professionals (APS).

The MAC determines whether an occupation is in shortage with two thresholds, depending on the quality of the data. For each indicator, the MAC considers an occupation to have a labor shortage if it surpasses the median plus 50 percent of the median on the normal statistical distribution for all occupations. In a majority of the indicators, this threshold qualifies about 25 percent of the occupations as experiencing shortages. For cases in which this threshold is not suitable, such as those where the median is close to zero or the shape of the distribution is not approximately normal, the MAC considers occupations which are in the top quartile as experiencing a labor shortage. Additionally, due to data limitations and the nature of certain industries, not all occupations may have data for each indicator. Thus, the MAC considers an occupation in shortage as one that achieves either of the thresholds outlined above in 50 percent or more of the available indicators for that occupation.

According to the committee’s 2019 analysis of these datasets, out of 105 occupations that were in shortage, programmers and software developers, which play a major role in AI, are in the highest demand.8MAC has released their 2020 review of the SOL, which can be found at https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/927352/SOL_2020_Report_Final.pdf, but MAC acknowledged that the COVID-19 pandemic likely affected their data. We chose to use the 2019 review to discuss the demands for AI talent in the United Kingdom because it demonstrates more typical labor market conditions. In addition, the MAC noted in its review that the SOL may be affected by the fact that the United Kingdom is transitioning to a points-based system in 2021 as it exits the European Union. The MAC does not list AI professionals as their own group because they categorize occupations with a Standard Occupational Classification (SOC) system that organizes each occupation based on a set of defined tasks or duties. This system is similar to the one used by the Department of Labor’s Bureau of Labor Statistics. Because AI professionals can work in a range of different disciplines, the UK SOC does not assign them their own category.

How Canada Prioritizes Occupations

Like the UK, Canada has a similar occupation shortage list and immigration fast-track program called the Federal Skilled Worker Program (FSWP). This program allows foreign professionals who are employed in 347 eligible occupations and meet minimum entry criteria to apply for a fast-track permanent residence process through the Express Entry Pool. For a job to qualify as an eligible occupation in the FSWP, it must be classified as a specific skill type as defined by the National Occupation Classification (NOC). The NOC system categorizes occupations in five groups, Skill Level 0 through D. Skill Level 0 jobs are in management, such as restaurant or mine managers. Skill Level A includes positions that require a university degree, such as doctors, architects, and dentists. Level B covers occupations that use either a college degree or an apprenticeship, such as plumbers, electricians, and chefs. Jobs that are categorized in Skill Levels C or D require less training, specifically a high school diploma or on-the-job training, respectively. These jobs include long-haul truck drivers, cleaning staff, and industrial butchers. Occupations which qualify as “skilled” in FSWP must be either Skill Level 0, A, or B.

For a prospective immigrant to qualify for the FSWP, they must:

  • Have worked for at least one year in a full-time, or an equivalent part-time, position in one of the 347 eligible occupations;
  • Have that work classified as managerial, technical, or professional;
  • Pass a background check and medical exam;
  • Be proficient in English and/or French;
  • Have enough money to settle in Canada; and
  • Have scored at least 67 points out of 100 on a skilled worker points grid.

This skilled worker points grid assigns points based on factors such as age, level of education, and language skills to assess how well the prospective immigrant would settle in Canada. The full points grid is as follows:

  • Level of fluency in English and/or French (maximum of 28 points)
  • Level of education either from a Canadian or foreign institution (maximum of 25 points)
  • Number of years working full-time (at least 30 hours per week) or an equivalent amount of time in part-time work (15 hours per week) (maximum of 15 points for six or more years of work full-time or at least 12 years of work part-time)
  • Age (maximum of 12 points received if the applicant is between 18 and 35 years of age)
  • Job offer for full-time, non-seasonal work for at least one year (maximum of 10 points)
  • Adaptability to Canada (maximum of 10 points based on the applicant’s experience in Canada, Canadian family connections, or future employment in Canada)

The Canadian government has recognized the significant shortage of experts in AI-related positions and has listed computer engineers and programmers, as well as software engineers, as priority immigration occupations. For AI overall, Canada was the first country in the world to adopt a federal AI strategy to strengthen the country’s research capabilities. One of the strategy’s main goals is to “attract and retain world-class AI researchers by increasing the number of outstanding AI researchers and skilled graduates in Canada.”

The U.S. has the potential to implement a similar system by collecting data from organizations such as Burning Glass. The Bureau of Labor Statistics also collects granular salary data that can be used to determine shifts in compensation that indicate potential labor shortages. The federal government can then use this information to publish a list of the top skilled occupations and amend the Schedule Alist with these occupations.

Conclusion

Countries all over the world are vying for limited AI talent and implementing transparent, straightforward immigration pathways to attract them. Without a similar system, the United States will fall behind. America can no longer rely on only its reputation for stellar research programs and top-ranking companies to attract the best talent. If it is too difficult to work, stay, and put down roots in the United States, AI professionals will go elsewhere. With AI so inextricably linked with both future economic and national security advances, the stakes are high.

It is possible, however, to utilize existing immigration pathways and clarify adjudication guidance to make it easier for USCIS officers to approve the petitions for eligible AI professionals. This paper proposes objective professional achievements in the field of AI which should satisfy the requirements for EB-1s, EB-2s with NIWs, and EB-3s in accordance with an updated Schedule A list. The Biden Administration has a golden opportunity to consider such reforms so that the United States can establish an enduring global lead in AI research and applications.


Appendices 1-5


Suggested Citation: Douglas Rand & Lindsay Milliken, Winning the Global Race for AI Talent Expertise: How the Executive Branch Can Streamline U.S. Immigration Options for AI Talent, N.Y.U. J. Legis. & Pub. Pol’y Quorum (2021).