
- AI Hiring
- Lenny Rachitsky
- Product Management
- Software Engineering
- Forward Deployed Engineering
- Career Switching
In this article (4)
Lenny Rachitsky says the biggest AI hiring boom is not in software
Key Takeaways
- Do not treat AI opportunity as synonymous with ML engineering. Scan product, operations, implementation, and domain workflow roles.
- Read job titles skeptically. Forward deployed engineer, AI engineer, and AI PM can describe very different work.
- Build proof around workflows and outcomes, not only certificates or tool names.
- AI Hiring
- Lenny Rachitsky
- Product Management
- Software Engineering
- Forward Deployed Engineering
- Career Switching
Why it matters
- careerSwitchersThe strongest AI opening may be where your domain knowledge meets workflow redesign, not where a title sounds most technical.
- ProductProduct teams are gaining hiring share as companies need people who can turn AI capability into usable products and processes.
- employersClearer role definitions will help candidates understand whether you need model builders, deployment staff, or workflow owners.
Career switchers should read the late 2026 job market as a nudge toward product, operations, and domain workflow roles, not only ML engineering.
The job post says AI engineer. The work, once you read past the title, might be product discovery, customer implementation, workflow redesign, or light data plumbing with a chatbot attached. That is the labor market Rachitsky is pointing at in his late 2026 report, where the headline surprise is not that AI is creating demand, but that the demand may be showing up somewhere other than the software engineering ladder.
In his Nov. 10, 2026 report, Lenny Rachitsky wrote that "The tech job market is growing fast, but mostly outside of software." That line will irritate some engineers and encourage some career switchers, which is usually a sign that a useful distinction is hiding underneath the slogan. The practical question is not whether software still matters. It is whether every AI career plan should start with becoming an ML engineer.
The signal is shifting toward product
Rachitsky says his analysis uses TrueUp job openings across over 9,000 companies and Live Data Technologies job change data covering 124 million professionals, with the hiring numbers focused on the U.S. In that dataset, according to Rachitsky, product manager openings are near a four-year high, 75% above the March 2023 low, and PM openings have grown 56% since the start of 2023 compared with 43% for engineering. He also reports that PM is the only one of five core roles whose hiring share increased, moving from 16% in 2021 to 21% today.
CellCog, which reviewed the free portion of Rachitsky's report, pulled out a useful comparison for job seekers: product management showed 7,303 openings and 32,600 U.S. hires, while software engineering showed 65,351 openings and 113,000 U.S. hires. Those numbers do not say product is larger than engineering. They say the growth signal is different from the volume signal, and applicants who only sort job boards by raw opening counts can miss the change.
Core role signals
| Role | Open roles | US hires |
|---|---|---|
| Product management | 7303 | 32600 |
| Software engineering | 65351 | 113000 |
For career switchers, that matters because PM work is often closer to the messy middle of AI adoption: prioritizing use cases, testing whether a model improves a workflow, translating customer pain into specifications, and deciding what not to automate. That is not a shortcut around technical literacy. It is a different bundle of evidence to show: shipped experiments, adoption metrics, domain judgment, and the ability to work with engineering without pretending to be engineering.
Software is not dead, but the generalist lane is narrower
Rachitsky's report is also blunt about engineering momentum. He says open engineering roles grew 29% through 2024, 15% through 2025, and only 2.5% in the past year. He also reports that U.S. companies hired about 113,000 software engineers from May 2025 to April 2026, compared with a 2022 peak of 239,000.
That is not a collapse. It is a hiring screen getting fussier. The résumés that used to clear because someone could build a web app now compete with candidates who can connect products to data, infrastructure, model behavior, security constraints, and customer deployment. Rachitsky also notes that forward deployed engineer openings are up 4.6x in the past year to around 1,500, while LavX News warns that some of that growth may be companies relabeling sales engineer roles.
That title sprawl is exactly where applicants need to slow down. Forward deployed engineer can mean customer facing implementation, prototype building, technical sales support, or all three. Before spending months chasing the label, read the job post for the workflow: who owns the customer, who writes production code, who measures success, and whether the role is really deployment or just demos with a fresher badge.
Outside software means domain work, not magic
CellCog notes that Rachitsky's seventh takeaway about hiring outside software sits behind the paywall in the version it reviewed, so the detailed evidence for that specific claim is not public in its writeup. That should make readers careful, not dismissive. The broader labor market does show hiring strength beyond tech company engineering teams: the U.S. Bureau of Labor Statistics reported that October 2026 payrolls rose by 165,000 and unemployment was little changed at 4.1%, with gains in professional and business services, health care, and transportation and warehousing.
Those sectors are where AI work often looks less like training a model and more like changing a process. A health care operations analyst may need to evaluate documentation tools. A transportation planner may need to redesign exception handling. A professional services manager may need to turn repeatable client work into a supervised AI workflow. None of those people need to cosplay as research scientists, but they do need to understand data quality, failure modes, privacy boundaries, and how to measure whether the tool actually helps.
This is where credential inflation gets expensive. A certificate that teaches vocabulary but never makes you map a workflow, test an output, or document a human handoff is weak evidence. A smaller project that proves you can take a real process, identify where AI fits, and measure the before and after will often tell a hiring manager more than a badge with a grand title.
How switchers should read the next posting
The useful move is to separate title from function. If a posting says ML engineer, look for model training, evaluation, data pipelines, inference systems, and production ownership. If it says AI product manager, look for prioritization, customer research, experiment design, and model assisted user experience. If it says AI operations or transformation, look for workflow mapping, vendor evaluation, governance, rollout, and adoption.
Rachitsky's data gives switchers a reason to widen the search, not lower the bar. At 25, that might mean taking a product analyst or implementation role where you can build technical fluency on the job. At 45, it may mean using domain depth as the wedge, then adding enough AI literacy to redesign the workflows you already understand. Different constraints, same hype cycle, and the same need to show evidence rather than slogans.
The next hiring wave will probably keep producing fuzzy titles. Treat that as a reading test. If the job is really about applying AI inside a business process, the winning portfolio may be a workflow map, a measured pilot, and a clear explanation of tradeoffs, not a résumé line that says AI engineer and hopes nobody asks which kind.
Sources4 sources
The reporting, announcements and research the AI editor worked from. Links open the original publisher.