The Society of Actuaries launched a job analysis survey in early March 2026 to every Associate and Fellow worldwide, and named the technology it expects to find. The press release says there is "an opportunity to tap into AI as part of the journey to become an ASA."

Credentialing bodies do not usually announce a hypothesis before fielding the instrument that tests it. The interesting part is not the signal but the interval between the signal and any change it can produce.

Key Takeaways

  • Every ASA and FSA worldwide received an invitation, with the survey run by Data Recognition Corporation. That is the psychometrically defensible process used to justify changes to exam content, not an informal feedback exercise.
  • 18 to 24 months is the gap the SOA has consistently kept between publishing job analysis results and making new requirements mandatory, with transition rules protecting candidates already in the pipeline.
  • January 1, 2026: both parts of the CAS Property and Casualty Predictive Analytics exam became required for the ACAS, so the comparable associate credential already tests applied modelling.
  • 22% projected employment growth for actuaries from 2024 to 2034 against a 3% average across occupations, with roughly 2,400 openings a year, is being met by more selective hiring rather than broader hiring.
  • The 2018 precedent worked because the technique stabilized. Predictive analytics entered the ASA pathway as a five-hour applied exam and became routine practice; AI governance frameworks are still being written.

What the Survey Is and Is Not

The distinguishing feature is procedural weight. This is the formal research a credentialing body uses to defend a change, which is why the framing around it stands out.

Invitations went to all Associates and Fellows globally beginning in early March 2026. Responses are anonymous and reported only in aggregate, and the stated purpose is to gather information on the knowledge, skills, abilities and other characteristics of a new ASA, so the credential meets evolving market needs. Data Recognition Corporation specializes in credential assessments, and its involvement is what makes the output defensible if the SOA later changes an exam syllabus.

The methodology mirrors the job task analysis that preceded the 2018 curriculum overhaul: survey credentialed professionals, validate the competency domains externally, then propose a revision with defined transition timelines. What is different is that the announcement states a direction. These notices are normally procedural and neutral, and this one identifies the technology before any responses are analysed.

The Cadence Is the Story, Not the Signal

The naming of AI matters less than the machinery it has to travel through, because that machinery runs on a schedule employers are already ahead of.

The prior cycles set the pace. The 2018 changes, which introduced Predictive Analytics as a core ASA requirement testing a real data set within five hours, were announced roughly 18 months before taking effect. The restructuring announced in July 2021 replaced Exam IFM with Advanced Topics in Predictive Analytics and rebuilt LTAM and STAM into FAM plus ALTAM or ASTAM, with candidates choosing between IFM and ATPA in 2022 and only ATPA available from 2023. The FAP streamlining that took effect in January 2025 cut the course from eight modules to five. Across all of them the SOA has held 18 to 24 months between publishing results and mandatory adoption.

Employer demand is not waiting for that interval. DW Simpson's 2026 recruiting report and the Bureau of Labor Statistics outlook point the same way:

SignalSourceFinding
Entry-level hiringDW Simpson (2026)Employers expect 2-3 exams plus internship experience and demonstrated technical/analytical project work
Mid-career differentiationDW Simpson (2026)Machine learning, risk governance, and GenAI knowledge cited as top differentiators
Employment growthBLS (2024-2034)22% projected growth vs. 3% average across all occupations
Annual openingsBLS (2024-2034)~2,400 openings per year through mid-2030s
Role transformationAmerican Academy of ActuariesAI will automate routine reporting; actuaries shift toward strategic and advisory roles
AI model validationNAIC regulatory actionsGrowing demand for actuaries who can validate AI systems used in underwriting and pricing

The competitive placement sharpens the point. From January 1, 2026 both parts of the Property and Casualty Predictive Analytics exam are required for the ACAS, a change the CAS set out when enhancing its exam content, and the CAS ran its own job task analysis with Assessment, Education, and Research Experts, using focus groups to identify the technical and soft skills actuaries will need over the next three to five years. By April 2026 every CAS exam except 6I is offered at least twice a year.

So the associate credentials diverge for the duration of the cadence. One already requires an applied modelling exam; the other has begun the survey that might eventually add comparable content, on a timeline that reaches candidates who have not yet sat Exam P.

What a Credential Can Durably Test

The harder question sitting under the survey is whether AI is testable at all in the form a credential requires, and the 2018 comparison cuts both ways.

Predictive analytics was a successful integration because the underlying technique was stable. Eight years on, testing it looks obvious. The methods a candidate learned for the exam are still the methods in use, so the syllabus did not need continuous maintenance to stay accurate.

AI does not currently offer that. The regulatory frame around it is being built right now: the NAIC's AI evaluation pilot is mid-flight, state model bulletins on algorithmic decision-making are still being adopted, and ASOP guidance on AI model governance is in development. A syllabus that tests governance frameworks likely to shift within two to three years imposes a maintenance burden on the credentialing body and hands candidates content with a short shelf life.

That pushes toward testing principles over tools. Model validation concepts, bias detection methodology and compliance reasoning survive a change of platform; proficiency with a particular system does not. It also constrains where the content can sit. Expanding ATPA's syllabus is the least disruptive route, but the exam already balances several competency areas inside five hours, so added coverage comes out of something else.

The alternative, another required assessment, runs against the direction of travel. The SOA pathway is long, and the CAS is moving toward more frequent windows and more scheduling flexibility. The existing micro-credentials, Pre-Actuarial Foundations, Actuarial Science Foundations and Data Science for Actuaries, are the mechanism that adds a verifiable signal without adding time, which is the only option on the list that does not make the credential longer at the moment its competitor is making theirs shorter.

Further Reading

Sources