On World Usability Day in 2023, I gathered my User Experience team for an AI Learning session. Members of my team had been experimenting with Adobe Firefly and Midjourney, and we were evaluating whether tools like Wevo could speed up our usability testing of websites. The room was somewhere between thrilled and uneasy — the tools were genuinely good, and the implications were starting to create some anxiety about job security.
So I told them what I believe (and still do) — the software industry will never stop needing people who understand users. As a UX designer, if what you love about the work is producing the artifact — the wireframe, the mockup, the polished comp — then yes, that part is going to be well automated soon, and people without deep expertise will assume those outputs are fine. But if what you love as a human-centered designer is understanding people — how they behave, where they get stuck, what a low-friction path through your product feels like — I think there will always be work for you in the software industry.
I was describing the difference between the task and the purpose. I just didn’t have the language for it yet.
A job’s task and its purpose are not the same thing
Speaking at the Milken Institute Global Conference in May 2026, Nvidia’s CEO Jensen Huang said: “the purpose of a job and the task of the job are related, not the same.” [1]
His example was himself. The task he does all day, he said, is typing and talking — both now automated. But if the role of technology is task enablement, he should be out of a job. Instead he works harder than ever, because he is able to give more attention to his purpose as his job tasks are streamlined. The same holds for a software engineer. The task is writing the code, but the purpose is solving problems and building things that didn’t exist before. AI is getting very good at tasks, but it doesn’t have the humanity to have purpose.
That distinction is the whole game right now, because every senior executive I talk to is evaluating what AI might do to organization structure, the distribution of work, and ultimately to headcount. The reflex is to look at a role, see the task AI can now do, and cut. It’s the wrong reflex, and now we have compelling proof from the field of radiology.
The radiologists were supposed to be gone by now
In 2016, Geoffrey Hinton — one of the people most responsible for modern AI, and later a Nobel laureate — said it was time to stop training radiologists. His words: “People should stop training radiologists now. It’s just completely obvious that within five years deep learning is going to do better than radiologists.” [2] At the time the prediction looked obvious, and it was repeated everywhere.
We know now that Hinton was wrong. Ten years on, there are roughly hundreds of AI tools cleared for medical imaging, and radiologists use them as enablers, not replacements. The Mayo Clinic in Rochester has grown its radiology staff 55% since Hinton spoke. [3] The field is currently living through the largest radiologist shortage in its history, with scans backlogged for months at some centers. [4] The tools that read images faster didn’t empty the department, they raised the volume it could absorb and the demand grew to meet it. Hinton himself has walked back his statement — he says he was wrong on timing, that he was only ever talking about reading images, and he now counts healthcare among the technology’s clearest winners. [2] The radiologist’s role wasn’t erased so much as remade. The familiar was radically changed, and unexpectedly, something more substantive rose in its place.
We’ve been here before, and I wrote about it twenty years ago
As with many things related to technology and society, none of this is new. While working towards my anthropology Ph.D., I wrote about how people make sense of their work inside the tech industry. The fear I documented has been the same at every major socio-technical transition, whether it’s from farming to manufacturing, manufacturing to service work, service work to knowledge work. People have always braced for the thing that would take their place and leave them with nothing. The economist Joseph Schumpeter called the underlying churn “creative destruction,” the repeated dismantling of the old to make room for the new. [5] We notice the destruction, but we’re slower to see what’s being created in its wake.
Back then I wrote about how enterprise software was automating routine tasks so employees could take on a broader range of responsibilities; it was the same pattern in a different era, and with different technology. [6] At the time I noted that these changes often meant longer hours; unfortunately, freeing someone’s capacity and respecting them as a person are not the same.
Researcher Shoshana Zuboff argued that a new technology can do one of two things. It can automate — take over the human’s task and push the person out. Or it can informate — give the person more to understand and make them more capable than before. [7]
The same tool can result in wildly different outcomes. And that’s because the difference isn’t in the technology at all. I think it’s in the choice the organization makes about its people.
What this looks like in software — and across my own career
The senior executive’s instinct is to judge the role by the outputs it’s currently producing. But I’ve never held a role that was only its task — including my most technical ones.
Years ago I ran a UNIX mail server named Bubba. The task was ensuring the server was properly configured and that it remained available at all times. The purpose was making sure tens of thousands of people got the email they were waiting for, every day, without fail. Later I was a test engineer loading SAP HR & payroll data into customer training systems. On paper, I was doing script-generated data entry. In practice, I was building the experience that determined whether a customer could learn the software they’d invested in. As an anthropologist, I joke that my whole job is to explain people to engineers — which is all about purpose.
The same split runs through every function in a software company. The engineer whose value was writing the code is not as valuable as the engineer actively engaged in deciding what’s worth building. The designer who only wants to turn the crank on comps is exposed, but the one who truly takes the time to understand their user is not. I told my designers exactly that in 2023 — and AI’s first huge conquest, writing software, makes the point. Layoffs today are focused on the task layer. The roles defined by production — entry-level coding, routine QA, data entry — are contracting [9], while the work that calls for judgment is not [8]. It is why the most durable technical role in the future may well be a forward-deployed engineer, embedded close to the customer’s problem and shipping against it, rather than someone handed a ticket and asked only to produce code. It’s the same forward-deployed posture I made the case for in my own work [link: The Forward-Deployed Anthropologist — insert URL]. The roles where the task compresses hardest are the ones most likely to need redirection over time.
Do you know your own people?
Which brings me to the real question, and it isn’t really about AI at all. Years ago at ZS, the Managing Director asked us to stop calling our consulting staff “resources.” It was impersonal, he said. It didn’t consider them as whole people — their aspirations, what they were good at, what they wanted next. He was right, and the idea has stayed with me.
AI doesn’t result in layoffs, but it does expose how you already see your people. If they’re resources — interchangeable units that produce a task — then introducing AI to accelerate that work seems to justify headcount reductions. But if they’re people with a purpose, then AI can free them up to do their best work. The technology is identical either way; this is ultimately about leadership.
The data bears that out. Much of what gets called an AI layoff is a budget decision in disguise. Companies are trading talent for token budgets, cutting staff and routing the savings into AI compute. [10] But in one study, 90% of executives said AI had made no difference to employment at their own company, even as they cited it publicly [11] — and Forrester expects about half of AI-attributed layoffs to be reversed within a year. [12]
So before you redraw the org chart around what AI can do today, a more thoughtful question might be: do you and your managers know your people well enough to tell the task from the purpose? Because that’s the judgment this moment is really testing.
References
[1] Milken Institute. “Leading in the Age of AI: A Conversation with NVIDIA CEO Jensen Huang.” Global Conference 2026 (transcript), May 2026. https://milkeninstitute.org/sites/default/files/2026-05/LeadingAgeAIAConversationNVIDIACEOJensenHuang_Transcript_GC26.pdf
[2] Hinton, G. Remarks at the Creative Destruction Lab seminar “Machine Learning and the Market for Intelligence,” Toronto, 2016 (quoted in European Medical Journal, “Artificial Intelligence in Radiology: An Exciting Future, but Ethically Complex,” 2023). Hinton has since qualified the prediction, citing healthcare among AI’s beneficiaries.
[3] Becker’s Hospital Review. “Mayo Clinic Radiology Leads in AI Use.” May 14, 2025 (reporting The New York Times). https://www.beckershospitalreview.com/radiology/mayo-clinic-radiology-leads-in-ai-use/
[4] Quartz. “AI Promised to Revolutionize Radiology, but So Far It’s Failing.” July 20, 2022. https://qz.com/2016153/
[5] Schumpeter, J. A. Capitalism, Socialism and Democracy. Harper & Brothers, 1942.
[6] Hanson, N. D. Consuming Work, Producing Self: Market Discourse in Dispersed Knowledge Work. Doctoral dissertation, Temple University, 2004.
[7] Zuboff, S. In the Age of the Smart Machine: The Future of Work and Power. Basic Books, 1988.
[8] U.S. Bureau of Labor Statistics. “Software Developers, Quality Assurance Analysts, and Testers.” Occupational Outlook Handbook, 2025 (employment projected to grow 15% from 2024 to 2034). https://www.bls.gov/ooh/computer-and-information-technology/software-developers.htm
[9] Tom’s Hardware. “Tech Industry Lays Off Nearly 80,000 Employees in the First Quarter of 2026 — Almost 50% of Affected Positions Cut Due to AI” (citing Nikkei Asia; Stanford findings on entry-level roles). April 2026. https://www.tomshardware.com/tech-industry/tech-industry-lays-off-nearly-80-000-employees-in-the-first-quarter-of-2026-almost-50-percent-of-affected-positions-cut-due-to-ai
[10] Challenger, Gray & Christmas, reported in TechTimes, “Tech Layoffs Hit 1,115 a Day in 2026: Companies Cite AI but Cuts Fail to Boost Returns.” June 16, 2026. https://www.techtimes.com/articles/318466/20260616/tech-layoffs-hit-1115-day-2026-companies-cite-ai-cuts-fail-boost-returns.htm
[11] Yotzov, I., Barrero, J. M., Bloom, N., Bunn, P., Davis, S. J., et al. “Firm Data on AI.” NBER Working Paper No. 34836, February 2026. https://www.nber.org/papers/w34836
[12] Forrester (2026 workforce predictions), reported in JobsPikr, “AI Layoffs 2026: The ROI Reality Check.” March 2026. https://www.jobspikr.com/report/ai-layoffs-2026-roi-reality-check/
Leave a Reply