AI Jobs Radar: Workforce Trends, Emerging Roles, and Skills Demand (October 29, 2026)

The twelfth week of tracking is defined by two very large announcements that trade people for compute and management layers. Microsoft announced on October 26 a time-limited voluntary exit offer to employees at level 70 and below in the US, Canada, India and the UK, a group of about 60,000 people or roughly 27% of its workforce, in a quarter where capital spending reached nearly $50 billion. American Express told investors on October 23 that it will cut about 3,500 jobs, around 5% of roughly 77,000 staff, with a $336 million third quarter charge and a plan to add more agents to its platform. Amazon eliminated about 250 Devices roles on October 28, mostly software developers, USAA has shed 700 with 2,500 to 3,000 expected by November, and Innodata reportedly dropped nearly 550 workers in India. Challenger, Gray and Christmas reports 43,281 announced US cuts in September, of which 3,961 cited AI, bringing the year to date AI total to 120,136, or about 21% and still the leading reason. On the demand side, SANS counts more than 850 open AI and ML security engineer roles at a $140,000 median, and Axial Search puts the median posted ceiling for AI governance roles at $169,000 across 1,997 postings. Fortinet reported third quarter revenue of $2.08 billion, up 20%, while Zscaler cut 261 roles and Cloudflare removed 1,100, showing that even growing security vendors are reshaping headcount. The University of Rhode Island and Georgia Tech each announced applied AI degrees that require employer practicums, and California's SB 951 will force AI layoff disclosure from January 1, 2027. The week's pattern is that employers are buying out or cutting broad middle layers and funding the premium specialists who govern, secure and supervise the agents that replace them.

Microsoft offers voluntary exits to 60,000 employees as capex reaches $50 billion a quarter

Microsoft announced on October 26 a new voluntary exit offer for employees at level 70 and below in certain organizations in the US, Canada, India and the UK. Sources told CNBC the eligible group is about 60,000 people, roughly 27% of the company's reported workforce of around 228,000. The offer is time-limited and includes financial support and employment assistance, and Amy Coleman, the company's executive vice president of people, framed it as part of how Microsoft is "evolving our approach to change." It follows the company's first voluntary retirement program in 51 years in April, which was open to about 7% of the US workforce and which more than 30% of those eligible accepted.

The buyout sits on top of other reductions. A July announcement will eliminate 4,800 positions during fiscal year 2027, an Xbox reset includes plans to cut 3,200 jobs by year end, and total cuts since 2023 were projected to exceed 40,000 before this week. Capital expenditure reached nearly $50 billion in the September 2026 quarter against nearly $35 billion a year earlier. TrueUp counts nearly 200,000 tech layoffs this year and estimates about 370,000 for the full year, below the 430,000 record of 2023.

Why it matters: A voluntary exit offer to more than a quarter of the workforce is a different instrument from a layoff. It lets Microsoft shed the most expensive, longest tenured layers without a WARN notice and without an AI label, which means the disclosure regime that starts in California in 2027 will not capture it. The ratio is the signal: roughly $15 billion of additional quarterly capex against a payroll that the company is trying to shrink by tens of thousands, which is a direct conversion of salary budget into compute and the clearest version of the bifurcation thread this blog has tracked since August.

Microsoft is also a gateway vendor, since Agent 365 is its control plane for enterprise agents, and every employee who leaves is replaced in part by an agent that needs identity, policy and logging. That puts the buyout in the same competitive frame as Palo Alto Networks with Prisma AIRS and Portkey, Check Point with its AI Defense Plane and Lakera, Fortinet with FortiAI and FortiSOC, and CrowdStrike with SGNL identity. Each is selling the governance layer for a workforce in which software does a growing share of the tasks.

The next test is the acceptance rate. If it approaches the 30% seen in April, roughly 18,000 people could leave, and rivals with similar AI spending, including Google and Meta, will copy the instrument rather than announce layoffs.

American Express cuts 3,500 jobs, 5 percent of staff, in a $336 million restructuring built around agents

American Express announced on October 23, during its third quarter earnings call, that it will cut about 3,500 positions, roughly 5% of a global workforce of about 77,000. The restructuring carries a $336 million charge in the quarter. Chief executive Stephen Squeri said the company needs to be "a more agile and efficient organization" and asked senior leaders to "reduce layers to speed decision-making." AP called it the company's first major job reduction in roughly a decade, and the cuts come with third quarter profit of $3.71 billion, up 28%.

The company did not say which roles are affected, how the cuts are split by country or whether contractors are included. Payments Dive reported that AI is part of the context, including broader AI use and pressure from AI-enabled commercial expense management competitors, and Squeri said the company plans to add more agents to its platform.

Why it matters: American Express is the second large financial services employer in two weeks, after NAB last week, to cut roughly 5 to 6% of staff while posting rising profit and avoiding the word replacement. The tell is the phrase reduce layers: flattening is the first and cheapest use of agentic AI, because agents compress reporting, reconciliation and approval chains before they replace any frontline function. A $336 million charge, about $96,000 per role, shows how much the company is willing to pay to move fast.

Card networks and banks are also the most regulated buyers of agent security. Check Point sells Lakera prompt injection defense and AI Red Teaming to exactly this segment, Fortinet brings FortiGate G Series and FortiSIEM to bank data centers and SOCs, Palo Alto Networks pushes Prisma AIRS across the gateway, and ServiceNow pitches AI Control Tower as the registry for agents. When Amex adds agents to its platform, a gateway that logs every tool call becomes the audit artifact regulators will request.

Watch the fourth quarter call for a role breakdown. A mix tilted toward operations and servicing would confirm that agent platforms are displacing process work, while a mix tilted toward management would reinforce the flattening thesis.

Amazon Devices, USAA and Apple: the unlabeled cuts of the week

Amazon laid off about 250 employees in its Devices organization on October 28, the group that builds Kindle and Echo products, with most of the cuts in software development roles. The company confirmed only "a small number of employees," and an email from vice president Cherie Robinson cited a "comprehensive evaluation of our structure."

A layoff tracker also lists USAA with 700 departures over the past two weeks and 2,500 to 3,000 expected by November, with cuts focused on remote and long tenured employees. Apple has cut more than 400 positions since August across AI, mixed reality, hardware and sales, and Comcast is requiring departures in regional and headquarters staff by year end. None of the entries is labeled AI driven.

Why it matters: The Amazon item is notable for what was cut, not how many. Software development roles inside a hardware group, at a company that spends tens of billions on AI infrastructure, are the category that coding agents target first. Combined with the Microsoft and Meta moves earlier this year, software engineering is now both the largest source of cuts and the largest source of premium postings, which is the labor market bifurcation inside a single job family.

USAA is the insurer version of the Amex story, and its focus on long tenured staff matches the Microsoft pattern of reducing the senior cost base. For security vendors, insurers and defense linked financial institutions are core buyers: Fortinet and Check Point sell the network and email layers, CrowdStrike the endpoint and identity layer through Falcon and SGNL, and IBM the DataPower gateway line. A shrinking human workforce raises the share of traffic that is machine to machine, and that raises the value of the gateway.

Expect the Challenger October report, due in early November, to include Amex and USAA. The question is whether AI is cited for them or whether the reason lands under restructuring again.

Innodata reportedly cuts 550 in India, a warning for the AI data labor market

Innodata, the Nasdaq listed AI data engineering company that provides model training, evaluation and safety testing services, reportedly terminated nearly 550 employees in India on October 26, with additional reductions in the Philippines and no figure given. The reported termination letter cited a "reduction in business assignments rather than individual performance." The company has not confirmed the number, and a WhatsApp group of affected workers grew past 600 members by October 28. Shares fell as much as 28% on October 27.

The business context is concentrated. In the second quarter Innodata reported revenue of $92.1 million, up 58%, and adjusted EBITDA of $25.4 million, but one customer represented about 58% of 2025 revenue and 37% of second quarter revenue after 56% in the prior quarter. The company reports third quarter results on November 5.

Why it matters: The data labeling workforce is the first layer of the AI economy and it is just as exposed to customer concentration as any contractor. A single large customer shifting from human labeled data to synthetic data or in-house teams can remove hundreds of jobs in a weekend, and the India based annotation workforce has no WARN protection. This is a quieter version of the bifurcation: the premium goes to engineers who design evaluation and safety testing, while the volume labor underneath is the most volatile.

The same data supply chain is a security surface. Training data pipelines, annotation tools and evaluation harnesses are inputs that Palo Alto Networks covers through Protect AI, that Check Point addresses with AI Red Teaming and that Fortinet will increasingly see as traffic in its FortiGate G Series AI data center deployments. Abrupt lockouts and mass terminations also create insider risk, and CrowdStrike identity tooling is the control vendors cite for revoking access at scale.

The November 5 print will show whether the customer shift is temporary. If revenue guidance holds at 40% growth while headcount falls, it will confirm that AI services firms are already substituting automation for labor.

Challenger: AI cited for 120,136 US cuts year to date, and California SB 951 will force disclosure in 2027

Challenger, Gray and Christmas reported that US employers announced 43,281 job cuts in September, down 18% from August. AI was cited for 3,961, about 9% of the month and the fifth most common stated reason, while market and economic conditions led with 8,789. Through nine months, AI is cited for 120,136 announced cuts, about 21% of the total and the leading reason of 2026, against 573,195 total announced cuts, down 39% from the same period of 2025. Technology companies announced 10,799 cuts in September, up 77% from August.

On the regulatory side, California SB 951 takes effect on January 1, 2027 and amends the state's Cal/WARN rules. Notices for mass layoffs caused in whole or substantial part by AI or other automated technology must give the number, occupation and work location of affected positions, the job functions being automated and the type of technology used, and must open with the header "This notice is for a technology displacement." The Employment Development Department must publish summaries and report to the Legislature by January 1, 2028.

Why it matters: The Challenger numbers show a market where total cuts are falling sharply while the AI attributed share is rising, which is what a structural rather than cyclical shift looks like. The September drop to 3,961 is also a reminder that the monthly series is lumpy and depends on how employers choose to word announcements. Microsoft's voluntary program and Amex's flattening language would score as restructuring, not AI, which is why the 21% figure is a floor.

SB 951 changes that incentive for the first time. A notice that must name the specific AI technology and the functions automated creates a public database of which vendor products are displacing which jobs, and it will push compliance work onto the same teams that govern AI. Fortinet, Check Point, Palo Alto Networks and ServiceNow all sell governance and inventory features that will be asked to produce that evidence, and an AI gateway log that records which agent performed which function is the cleanest source.

Other states are watching California. If New York and Illinois follow within the next two quarters, the AI layoff disclosure becomes a standard compliance workflow with its own tooling market.

SANS tracker: 850 AI security engineer roles at $140,000, and Axial prices AI governance at a $169,000 median

Cybersecurity Dive reported on October 20, using SANS data from mid-September postings, that there are more than 850 open AI and ML security engineer roles with a median salary of $140,000, from employers including Amazon, Mastercard, PepsiCo and SpaceX. AI red team specialist postings number nearly 100 at a $124,000 median and rose 13% week over week, with Cloudflare among the hirers. AI governance, risk and compliance leads number about 430 at $146,000, with JPMorgan Chase, Leidos and Aramark hiring. Smaller emerging categories include AI supply chain security engineers at 93 postings, AI security specialists at 48 and AI SOC orchestrators at 36. Conventional roles are also growing: compliance analysts exceed 450 postings, up 20%, and identity and access management analysts reach 180, up 51%.

Axial Search, in a page updated October 27, analyzed 1,997 US AI governance postings since January, about 71 new postings per week. The median posted salary ceiling is $169,000 overall, with junior individual contributors at $125,000, mid at $166,000, senior at $208,000, principal at $249,000, managers at $244,000 and directors at $268,000. Professional services accounts for 33% of postings, and observability platforms appear in 46%, NIST frameworks in 29% and Python in 28%. These are the top of posted bands, not typical offers.

Why it matters: The SANS data says nearly every large enterprise deploys AI without dedicated AI security staff, and the postings are the market correcting that. Two details deserve attention. The identity and access management jump of 51% is the clearest hiring signal for agent governance, because every agent is a non-human identity that needs lifecycle control, which is exactly the problem CrowdStrike addressed by paying $740 million for SGNL and Palo Alto Networks addressed through CyberArk. And the Axial result that observability appears in 46% of governance postings confirms that the job is operational: you cannot govern what you cannot see, and the AI gateway is the point where that visibility is created.

Compensation shows the compression of seniority. A principal individual contributor at $249,000 nearly matches a director at $268,000, so the premium goes to hands-on expertise rather than span of control, which is the opposite of what Amex and Microsoft are doing to management layers. A $125,000 junior ceiling next to a $208,000 senior ceiling also shows why companies hire experienced people for these roles: the gap rewards the few who have already supervised agents in production.

For vendors, this is a certification and talent war. Fortinet and Check Point both run training programs and need AI governance specialists inside customers to buy their agent controls, and Cloudflare hiring red teamers while cutting 20% of staff earlier this year is the plainest example of the pattern.

Security vendors split: Fortinet grows 20 percent, Zscaler cuts 261 and Cloudflare removed 1,100

Fortinet reported third quarter results on October 28 with revenue of $2.08 billion, up 20%, product revenue of $765 million, up 37%, and billings of $2.34 billion, up 29%. GAAP operating margin was 33.9% and non-GAAP was 38.2%. Fourth quarter guidance is revenue of $2.155 to $2.245 billion and full year revenue of $8.125 to $8.215 billion, about 20% growth. The release does not mention headcount.

Peers are reshaping staff differently. Zscaler said on September 3 it will cut about 261 employees, roughly 3% of more than 8,700, at a cost of $30 million to $33 million, to redirect savings into sales teams hunting new customers, in a quarter with revenue of $898.2 million, up 24.9%. Cloudflare announced in May that it will cut about 1,100 jobs, roughly 20% of about 5,500 employees, alongside record quarterly revenue of $639.8 million, up 34%. Chief executive Matthew Prince said the internal AI tipping point came last November and that "just because you're fit doesn't mean you can't get fitter."

Why it matters: Every one of these vendors is growing 20% or more and still reshaping headcount, which shows that cutting is a strategy and not a symptom. Fortinet's revenue growth of 20% with product revenue up 37% shows the data center and AI infrastructure refresh is reaching the firewall layer, and its organic approach with FortiAI and FortiSOC rests on keeping engineering headcount focused. Zscaler is shifting to sales and Cloudflare is shifting to AI engineering, so three vendors in the same market have three different labor models.

Against them, Check Point reports its third quarter on November 2 and has built its agent security position through acquisitions including Lakera, so its labor story will be told in integration rather than in cuts. Palo Alto Networks has absorbed Portkey and CyberArk, CrowdStrike cut 5% in 2025 citing AI, ServiceNow has guided to flat headcount while integrating Armis and Veza, and IBM continues to sell DataPower gateways and watsonx Orchestrate. The gateway is where these labor strategies converge: whichever vendor gives a customer fewer people to manage more agents wins the budget.

Expect Check Point's call to be the first in this cycle to quantify AI security headcount, and expect Fortinet's next filing to show whether engineering hiring accelerates as Q4 billings guidance of $2.775 to $2.925 billion implies.

URI and Georgia Tech launch applied AI degrees with mandatory employer practicums

The University of Rhode Island announced on October 27 a new Interdisciplinary and Applied AI bachelor's degree, delivered through the College of Business with the College of Arts and Sciences. It requires 53 to 70 credits, a soft launch this spring and full integration by fall 2027. Students must complete an internship with a company as an AI specialist. The program follows a 75 page Rhode Island AI Task Force report and the university's AI for Everyone course.

On October 6 the Georgia Tech announcement said the Board of Regents of the University System of Georgia approved an Online Master of Science in Artificial Intelligence, with launch planned for 2027. It offers six concentrations: AI in Business, Decision-Support AI, Foundations of AI Theory, Health AI, Human-AI Interactions and Predictive AI, and every student must complete an AI practicum with a company or organization.

Why it matters: Both programs hard-wire employer experience into the credential, which is a direct response to the entry level problem. If agents absorb junior tasks, a degree without applied work will not place graduates, and the practicum requirement shifts the first hire decision into the education period. It also gives employers a pipeline to evaluate candidates at low cost, which fits the Axial pattern in which postings ask for around five years of experience.

The concentration list is the more useful signal. Human-AI Interactions and Decision-Support AI map to the supervisor and governance roles in the SANS tracker, while Health AI and AI in Business map to regulated sectors where Fortinet and Check Point compete on compliance. Universities are producing the people who will write the policy that the AI gateway enforces, and vendors that supply curricula, labs and certifications gain influence over that talent early, as Cisco does through its AI Workforce Consortium.

The first real data arrives in 2027. Placement rates for practicum based programs against traditional computer science degrees will show whether the applied AI credential commands the premium that postings suggest.

Numbers at a glance

This week's data spans Microsoft's voluntary exit offer to about 60,000 employees, 27% of staff, with capex near $50 billion a quarter, 4,800 FY27 cuts and 3,200 Xbox cuts, American Express's 3,500 roles, 5% of 77,000, with a $336 million charge and $3.71 billion quarterly profit, Amazon's 250 Devices cuts, USAA's 700 so far with 2,500 to 3,000 expected, Apple's 400 since August and Innodata's reported 550 in India with shares down 28%. Challenger counts 43,281 September cuts, 3,961 citing AI, 120,136 year to date at 21%, and 573,195 total cuts down 39%, while TrueUp estimates 370,000 tech layoffs for 2026. SANS lists 850 AI and ML security engineer roles at $140,000, 430 governance leads at $146,000, 100 red team roles at $124,000 up 13%, and identity analysts up 51%, while Axial measures 1,997 governance postings at a $169,000 median and $249,000 for principals. Fortinet posted $2.08 billion revenue up 20%, Zscaler cut 261 (3%), Cloudflare cut 1,100 (20%), Georgia Tech adds 6 concentrations, URI requires 53 to 70 credits and California SB 951 starts on January 1, 2027.

References

1. iHeart: American Express to cut 5% of jobs as AI reshapes business
2. iTechGuides: Amex CEO says American Express will cut about 5% of its global workforce
3. TechSpot: Microsoft is offering voluntary buyouts to about 60,000 employees, a quarter of the company
4. APH Networks: 60,000 Microsoft employees offered voluntary buyouts as AI costs bite
5. StartupTalky: Amazon cuts 250 jobs as employees wake up to morning emails
6. Georgia Tech: Georgia Tech announces new online master's degree in artificial intelligence
7. Everhint: Layoff news and hints, October 28, 2026
8. CIOL: Innodata reportedly cuts 550 India jobs, letter cites ramp-down
9. Happened in AI: AI cited for 3,961 US job cuts in September, down sharply, but still the top reason for 2026
10. CDF Labor Law: When robots replace staff, California's new AI layoff disclosures
11. Cybersecurity Dive: AI security jobs hiring, SANS tracker
12. Axial Search: AI governance jobs, what the hiring data shows
13. SEC (Fortinet Form 8-K): Third quarter 2026 financial results
14. GovInfoSecurity: Zscaler to lay off 3% of staff as it shifts spend to sales
15. People Matters: Cloudflare cuts 1,100 jobs despite record revenue as CEO says AI changed workforce needs
16. University of Rhode Island: URI to launch new Interdisciplinary and Applied AI bachelor's degree