Verizon's $70 Million AI Skills Plan Counts Its Layoff Fund

AI-generated editorial illustration.

On September 23, Verizon announced a $70 million AI-skills initiative. The number joined two commitments made at different moments: $50 million in new funding for a public training effort, and $20 million already committed to help employees leaving Verizon. The second commitment began in November 2025, when chief executive Dan Schulman told staff the company would cut more than 13,000 jobs.

The September launch gave those two groups a shared headline, but it did not give them the same route to a job. Verizon alumni can seek career-transition services, courses and coaching under an eligibility-limited fund. An adult who has never worked for Verizon can use a free AI-learning portal. Local partners may help people choose courses and practice skills. None of those steps, by itself, shows that a participant found paid work or kept a job.

Goodwill Industries International has supplied the first concrete local test. In an October account of the partnership, it said 18 local Goodwill organizations would enroll and support learners in a pilot. Goodwill has a network of 150 local organizations. The pilot’s stated job is to test and support training, not to report a placement rate that does not yet exist.

An employee losing a job needs more than a course login. The person needs to know who can help choose a useful class, whether there is time to finish it, and which employer might pay for the resulting skill. Verizon’s announcements describe the funding and the doorway. They have not yet followed people through the next steps.

Twenty million came from a harder decision

Schulman’s November 20, 2025 letter gave two decisions together. Verizon would begin reducing its workforce by more than 13,000 employees and cut spending on outsourced and other outside labor. It would also establish a $20 million Reskilling and Career Transition Fund for people departing the company. The letter said the fund would focus on digital training, skill development and job placement.

Those words matter in the September 2026 arithmetic. Verizon’s new announcement calls the total a $70 million investment, while identifying only $50 million as additional funding. The earlier fund accounts for about 29% of that headline figure. That is a breakdown of the company’s stated commitments, not proof that $20 million had already been disbursed or that $70 million had been spent by launch day.

The November letter framed the cuts around a cost structure that, in management’s view, limited investment in customer value. It also discussed technology and economic change. It did not demonstrate that an AI system replaced each eliminated job. A training offer for departing staff does not establish the cause of the departure, either. Putting the two announcements side by side is useful because the same company made both decisions and later combined their budgets. A causal story about AI layoffs would need a different record.

For a person leaving Verizon, the wording of a fund is only the beginning. The alumni program page says eligible former employees can get free digital training and other resources. It lists more than 30,000 online courses, 80 professional-certificate programs, one-to-one AI coaching through Leland and live classes through General Assembly, the latter on a first-come, first-served basis. The page presents options. It does not publish how many departing employees have enrolled, finished, gained a credential or entered a new job.

Eligibility also narrows the denominator. The 2025 letter referred to people departing Verizon; the public alumni page says “eligible Verizon alumni.” Neither text, on the pages reviewed for this article, supplies a complete count of eligible people or an outcomes table for them. Dividing $20 million by 13,000 would produce a tempting figure of roughly $1,538 per announced cut. It would be a misleading per-person spend: the number of affected employees who qualify, the fund’s timing, the services consumed and the amount paid out are not public here. The arithmetic is possible; the inference is not.

The alumni page also makes access time-sensitive. It says the edX-backed Skill Forward certificates are available through December 2026 for eligible U.S.-based Verizon alumni. It warns that inactive course licenses may be removed after non-use. The live General Assembly series has limited places. A person who leaves after a busy period of job applications may face a very different practical offer from someone who can start training immediately. Verizon does not publish a take-up rate by departure date, so the effect of those terms cannot be measured from the announcement.

A budget line can appear in an announcement before any worker uses a service. Later, Verizon may sign a provider contract; a former employee may enroll; a course may be completed. Even then, an employer has to value the skill. Each event needs its own date and denominator. A future report that repeats only the original allocation would tell shareholders how much was pledged, while leaving former employees’ next jobs out of view.

Fifty million widens the classroom

Verizon used the new $50 million to make AI learning a public program, AI Skills for America. Its learning page says the portal is free for adults aged 18 or older in the United States and its territories and requires no Verizon service commitment. It advertises more than 200 courses across AI basics, productivity, digital safety, ethics, data analysis, marketing, software and cybersecurity.

That is a different catalog from the 30,000-plus courses on the alumni-fund page. A reader should not add the two counts or assume every public learner receives the alumni package. The former refers to a broad learning portal. The latter describes options for eligible former Verizon employees, including separate partners and services. The company has not published a single inventory that maps every course and coaching entitlement to each program tier.

The public portal names content from IBM, Google, Microsoft, Anthropic, Coursera and OpenAI. That tells a learner where some lessons come from. It does not show what a participant learned, whether an employer tested that skill, or what Verizon paid a provider. Verizon says comparable premium training can cost more than $700 per person per year. This is a company-stated retail comparison, not a disclosed average program cost. Multiplying $700 by registrations would turn a marketing price into a spending estimate, even when a person used only one short lesson.

The National Association for Community College Entrepreneurship’s program page adds an important detail about time. It says some courses take as little as 15 minutes, while others involve deeper study. Learners can pause and resume. A fifteen-minute lesson on writing a useful prompt, a professional certificate, a live workshop and individual coaching are all training, but they impose different demands on a worker’s time and may signal different things to an employer. Counting all four as equal completions would conceal the difference.

The menu also serves different people. A job seeker may need a hiring manager to recognize a skill and a way to demonstrate it. A small-business owner may want to save time on marketing or inventory work rather than change jobs. A teacher may need a classroom-ready method and a district’s permission to use it. Verizon’s announcement includes all three groups, alongside displaced workers and students. One participation total cannot show whether the program solved any one group’s problem.

Verizon has put a separate local amount behind the campaign. It announced a $1 million grant to Liberty Science Center in Jersey City on September 23 for interactive AI-skills programming. The announcement describes a planned experience at a science center, with visitors exposed to the training resource. It does not say whether this grant sits inside the $50 million new allocation or adds to it. Nor does a museum visit itself count as vocational training or a job outcome. Reporting those boundaries keeps an attractive public event from silently becoming an employment statistic.

The $50 million can lower the price of entry to instruction. It cannot, on its own, create a vacancy, release time for a worker to study, or make a certificate count in a wage negotiation. Those costs and decisions move to someone else.

Verizon has funded a more job-focused model before. In October 2020, it announced more than $44 million for a partnership with Generation, a nonprofit that trains people for specific occupations. Part of that commitment went to Jobs for the Future for evaluation. The announcement set an ambition to prepare 500,000 people for future jobs by 2030 and gave priority to unemployed and underemployed workers.

That older commitment is history, not an amount to add to the 2026 $70 million. It shows that Verizon already knew a course offer and a job outcome needed different institutions: one to train, another to evaluate. The new plan names more content suppliers and local partners. Its public materials still do not connect an enrolled learner to a later paycheck.

Eighteen Goodwills test the local handoff

Verizon says the national course portal will pair with local support through Goodwill, Local Initiatives Support Corporation and the National Association for Community College Entrepreneurship. In the launch announcement, Verizon calls these organizations a way to provide coaching, curriculum guidance and community access. The announcement says core support services will operate in select regional markets, even as online access opens nationwide. National portal availability therefore should not be read as nationwide, in-person coaching coverage.

Goodwill is the first named partner to describe a test at a specific scale. Its October post says 18 local organizations will enroll and support learners. It uses the verbs “pilot and test.” The post does not identify the sites, set a participant target, state an employment goal or report completions. Goodwill’s wider network of 150 local organizations makes the 18-site figure a pilot footprint, not national implementation through every Goodwill.

The most useful advice may come before anyone selects a course. A job seeker with limited study time needs to know whether a short AI lesson, a certificate, a portfolio project or an immediate job referral fits the role they want. A catalog cannot learn that from a click. Goodwill says its local organizations will support enrollment and learners. Its October post does not say how each adviser will make those choices, so the decisions described here are a proposed service test, not a reported script from the pilot.

Community-college partners face another handoff. NACCE’s page invites adults to free courses and describes certificates in cybersecurity, data analytics, cloud, IT support and digital marketing. It also describes shorter lessons in using ChatGPT, Copilot or Gemini for work tasks. A certificate in a field with vacancies can support a career path, but the page does not promise that a particular college will award credit, that a local employer will accept the certificate or that a student will qualify for an interview. The delivery channel and the labor-market result should remain separate.

The colleges and nonprofits also need employers to say what counts. A hiring manager can value a candidate’s completed project, a recognized certificate or prior work experience. The Verizon and NACCE pages list courses and fields, but they do not identify employers that have agreed to interview program completers or adjust a hiring requirement. If local advisers can identify those employers during the pilot, they may steer participants toward courses with a plausible paid destination. If no employer changes its selection process, the program can still teach useful skills; its employment claim will be harder to substantiate.

Opening the portal widely has a clear benefit: people can try a lesson without paying or waiting for a local cohort. Small-business owners and teachers may gain useful skills without seeking a new job. The 18-site pilot can test where that open door needs human help. Those are access and learning gains. A displaced worker’s return to paid, durable work still requires separate follow-up.

For Goodwill, the pilot also creates a chance to learn what the portal does not show. A person may register and stop after one module because the subject was wrong, because a shift schedule changed, or because an employer offered work. Those reasons have opposite meanings for the program. A completion rate alone cannot distinguish them. Follow-up has to ask, and it must do so without turning a learner’s job search into an intrusive data harvest.

Free access still costs time

The word free describes the price charged to a learner for the portal. It does not describe the total cost of learning. A worker needs a device, an internet connection, time and enough confidence to practice. A job seeker may need transportation to a coaching site and money for the weeks spent building a portfolio. A small employer that wants workers to learn on the clock must cover their paid hours. None of these costs is quantified by Verizon’s announcement.

A 2026 GAO review of federally funded job programs found that job seekers with disabilities did not always receive needed support at local centers, including sign-language interpretation in some places. GAO studied government programs, not Verizon or Goodwill. Its finding points to a test for any local training pilot: an accessible course page alone does not guarantee that a coaching session, assessment or referral works for every learner.

The 18-site pilot could record accommodation requests, whether support arrived and whether participants could finish their chosen path. That would measure service delivery. It would leave employment claims for a later follow-up, when a participant has had time to seek a job.

AI-generated editorial illustration. The separate surfaces mark the unverified steps from learning to paid work.

This matters most for people who are already changing jobs. Verizon’s original $20 million fund was announced on the same day as more than 13,000 planned cuts. A departing worker may have severance or transition support, but the public fund page does not report the duration or dollar value of individual access. Some offerings are explicitly limited: the General Assembly live series is first-come, first-served, while the alumni program is for eligible people. A comparison with the open public portal needs to account for those terms before calling the packages equivalent.

The labor market supplies another limit. Skills can improve a candidate’s chances only where employers need and recognize them. A program may teach a call-center worker to summarize customer histories with AI, yet that skill may be useful in the worker’s current occupation rather than a new technology job. A marketing course might help a small business sell more without producing any new paid job. Both could be worthwhile outcomes, but they belong in different columns from wage gains or placement.

The U.S. Government Accountability Office’s July 2026 training guide offers a useful external standard. It asks whether training aligns with mission needs, is delivered efficiently and improves staff performance. It treats planning, design, implementation and evaluation as a cycle, with coaching and mentoring playing a role. The guide was written for federal agencies, not Verizon. Its distinction between program delivery and improved performance still applies as an analytic test of any large training pledge.

The Department of Labor’s WIOA performance indicators keep credential attainment, employment after exit and median earnings as separate measures. WIOA’s legal reporting rules do not automatically govern Verizon’s private initiative. They show why a certificate count cannot stand in for a job or a paycheck. The federal framework even measures employment in the second and fourth quarters after participants leave a program, when an initial placement can be tested for persistence.

A strong evaluation would still have to be careful. People who choose a long course may already be more motivated or have better internet access than people who never enroll. Local job markets differ. A wage change could reflect a new employer, a negotiated raise or a wider economic change. A fair public report can show participant outcomes without pretending every improvement was caused by training. It can compare similar cohorts where possible and describe what it cannot identify.

That is a harder but more honest story than saying the public received a certain number of dollars of training value. The face-value comparison says what a consumer might have paid for some courses. The outcome record asks what people could do afterward.

A course catalog cannot count jobs

The figures available on launch day describe inputs and reach: $70 million in combined commitments, $50 million newly announced, $20 million previously allocated, more than 200 public courses, more than 30,000 alumni course options, 80 certificate programs and 18 Goodwill pilot organizations. They use different denominators. None counts a paid role filled through the new initiative.

The calendar explains why there is no employment result yet. Verizon announced the public plan on September 23, and Goodwill described a pilot in October. An October learner cannot already have a verified second-quarter employment outcome after leaving the course. The urgent task is to define the first cohort before its trail splits among course providers, local advisers and employers.

The unit of analysis should be a person, with privacy protections and a clear permission boundary, not a bundle of course clicks. If one learner finishes three short lessons, attends a live session and joins a Goodwill workshop, a course-level dashboard might count five activities. The worker is still one person. An employer outcome should attach to that person only after a verifiable event, such as a paid role or documented advancement. Aggregate reporting can protect identity while preserving the distinct stages.

An independent evaluator would also need to know who never started. People who encounter the portal and decide it lacks a relevant pathway are part of the access story. A program that enrolls more people by lowering friction can look worse on raw completion percentage if it reaches learners who previously had no training options. A completion percentage without the entry cohort, course length and support intensity invites the wrong comparison.

The comparison group matters as well. Graduates who accept a new job may stop logging in before finishing a credential; counting them as unsuccessful learners would reverse the program’s purpose. Others may finish because their employer gave them paid time to study, an advantage unavailable to a newly unemployed person. A report should show these routes separately, including whether participants were already employed, recently displaced or seeking first work at enrollment. That division is more useful than a single pass rate and less likely to reward the easiest-to-serve group.

Verizon could report different routes with different endpoints. For departing employees: eligibility, take-up, coaching, credentials, job placement, pay and employment persistence. For job seekers outside Verizon: course use, demonstrated skills, referrals and later employment where a follow-up is permitted. For workers who remain employed: applied projects, manager assessment or wage progression. For small businesses: owner-reported adoption may be a start, but revenue, productivity or reduced cost would require independent methods to separate training from other changes. Teachers need a separate classroom measure.

That list is a proposal for evaluation, not a claim that Verizon has committed to publish those fields. The company has not disclosed an outcomes schema or a target placement rate in the materials reviewed here. Keeping the proposal separate from the announcement prevents a proposed scorecard from being mistaken for company practice.

An outcome file for the first cohort

A useful public follow-up would start with a dated group of participants and preserve each transition. The table below proposes the records an evaluator would need. Verizon and Goodwill have not said they use this dashboard.

Stage	Denominator to preserve	Event to verify	Likely record owner
Offered access	Eligible alumni, public visitors or invited local participants, reported separately	Clear offer and eligibility terms	Verizon and local partner
Began learning	Unique eligible people in each cohort	First substantive lesson or coached session	Portal or training provider
Practiced a skill	People who began a defined pathway	Assessed task, project or coached demonstration	Course provider and local partner
Earned a credential	People eligible for that credential	Issued certificate with level and provider	Credential issuer
Entered paid work	Job-seeking participants with follow-up permission	Paid start date, role and employer or self-employment record	Participant and employer, verified with consent
Stayed or advanced	Participants with a prior paid-work baseline	Second- and fourth-quarter work status and pay band	Independent evaluator or trusted partner

The first row forces a split between Verizon alumni and the general public. The second prevents a single person from becoming several “learners” because they used multiple courses. The third tests a skill instead of a sign-in. The fourth keeps a certificate distinct from employment. The final two require time and consent, and will have incomplete follow-up. An honest report should publish that missingness alongside the result.

For the $20 million transition fund, a further budget table would help: amount committed, amount placed under contracts, amount spent, eligible people, participants served and service mix. It should distinguish the previously announced fund from the $50 million public expansion. For the new allocation, Verizon could disclose the share used for content rights, local coaching, outreach, accessibility and evaluation. Those are suggested categories, not figures in the announcement. Without them, outsiders cannot tell whether the largest expense is a course license, a local adviser or something else.

The 18-site Goodwill pilot is a practical place to test the file. Local organizations could report the number of people offered support, the number who actually meet an adviser, the courses selected, and what happened after participants stopped studying. An independent follow-up might ask whether a certificate helped obtain an interview and whether a paid role followed. The pilot cannot prove national effectiveness, but it can expose which handoffs need repair before wider rollout.

Verizon can reasonably value broad access. A telecommunications company has communities and small businesses that may benefit from training without hiring anyone. Workers and local partners may place more weight on time to a wage, a recognized skill and a reliable adviser when a course is too little. A public report can include both views, provided the measures keep them distinct. A survey about confidence cannot replace a payroll event; a payroll event does not capture every useful lesson a small-business owner learned.

The employment test comes later

Verizon has made two observable moves. In November 2025 it announced a $20 million fund while cutting more than 13,000 jobs. In September 2026 it counted that fund within a $70 million AI-skills initiative and opened a new $50 million public effort. By October, Goodwill had committed 18 local organizations to pilot and test learner support. The public portal lists more than 200 courses. Those facts establish a substantial access offer and an early delivery network. They do not establish employment results.

The next useful dates belong to participants. An eligible former employee will need an access date, a service actually used and a later job status. A public learner’s record may end after a short lesson, or continue through a project, adviser meeting and employer referral. Some people will decline follow-up. A report should show how many could not be reached, rather than quietly removing them from the denominator.

Goodwill’s 18 local organizations are now the nearest visible place to test that chain. For each person they support, the first hard result will be smaller than the September headline: a completed task, a credible referral, perhaps eventually a paid start date. Verizon can publish those dates as the cohorts mature. Until then, $70 million describes the commitment. It does not describe the worker’s next job.