Bill Gates, Zoho founder Sridhar Vembu and OpenAI CEO Sam Altman have made related but distinct points about AI and work. The available reporting does not show that they jointly agreed AI will “steal most jobs.” Gates was reported to name coding, energy and biology as relatively resilient fields; Vembu forecast that AI could produce much of the code programmers write; and Altman has argued that AI can make software engineers more productive. None of those claims proves that most jobs will disappear.
What the three executives actually said
A March 26, 2025 article in Indian Defence Review brought together comments from three separate people. Its headline turns those comments into a stronger shared conclusion than the evidence supports.
Bill Gates: coding, energy and biology
Gates was reported to have identified coders, energy experts and biologists as fields likely to remain comparatively resilient. The three-job list is supported here by secondary coverage; a primary transcript or video listing all three was not available. Axios reporting separately describes Gates’s view that coding remains a useful skill in an AI economy.
“Resilient” should not be read as untouched by AI. It more plausibly means that work involving discovery, complex system design, real-world experimentation, cross-disciplinary judgment and responsibility for outcomes may remain valuable even as AI changes how it is done. Gates’s broader view, as summarized on Gates Notes, is that AI could help people perform many jobs more efficiently, while also taking on some routine services. That is a less absolute claim than saying jobs are simply stolen.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →#1 Best Overall
Sridhar Vembu: a forecast about code, not programmers
The March 2025 article attributes to Vembu the forecast that AI could write roughly 90% of code, much of it boilerplate or what he described as “accidental complexity.” He distinguished that routine work from “essential complexity”: the difficult, original and system-level problems that still call for human expertise.
The figure is Vembu’s opinion, not a measured statistic about software employment. It refers to code or coding activity, not to 90% of programmers losing their jobs. Writing code is only one part of software engineering, which also includes defining requirements, architecture, security, testing, deployment, maintenance and deciding whether a solution serves users’ needs.
Sam Altman: greater productivity, with possible hiring effects
The same article presents Altman’s view as one in which AI makes individual software engineers substantially more productive and may eventually mean fewer engineers are needed for a given amount of software. That is a productivity argument with a possible employment consequence—not a demonstrated forecast that most engineers, let alone most workers, will lose their jobs.
OpenAI’s 2025 enterprise report says 73% of surveyed engineers reported faster code delivery. This is company-reported survey data about the respondents, not independent proof of productivity gains across the whole economy.
Why task automation is not the same as job loss
AI’s effect can occur at several levels. A tool may help with a task, automate a task, reshape a role or reduce an employer’s need to hire—without eliminating the occupation. A job category largely disappearing is a further and more consequential outcome.
Rank #2
- Task assistance: AI helps a worker complete an existing duty.
- Task automation: AI performs a discrete activity that a person previously handled.
- Role redesign: Workers take on different or broader responsibilities as some tasks change.
- Lower hiring demand: An employer may produce the same output with fewer new hires, or increase output without adding staff.
- Occupation elimination: The job category itself largely disappears. This does not automatically follow from any of the steps above.
The International Labour Organization (ILO) emphasizes that exposure to generative AI does not automatically mean job loss. Its 2025 update finds that many occupations are more likely to be transformed or augmented than fully automated, though clerical work and some other occupations face greater exposure. The ILO’s one-in-four figure describes jobs potentially at risk of being transformed by GenAI, not jobs certain to disappear.
Whether a system augments or replaces work depends in part on how central the affected task is to a role, how an employer adopts the technology and whether human oversight or complementary work remains necessary, according to the ILO’s discussion of AI adoption and its impact on jobs and its AI and employment topic page.
What makes work more exposed—and what may make it more resilient
Job titles alone are poor predictors. Work is generally more exposed when its tasks are digital, repetitive, standardized and easy to evaluate. Examples include data entry, routine translation and summarization, template-based marketing, basic bookkeeping, document review, standardized customer support, routine coding and test generation. The ILO identifies clerical occupations among those with higher GenAI exposure, while cautioning that exposure is not the same as full automation.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsWork may be harder to automate fully when it involves unpredictable physical settings, ambiguous goals, high costs of error, personal trust, negotiation, regulation, proprietary real-world information or accountable decisions. These traits are not guarantees: AI can still change parts of those jobs, and organizations may choose to redesign them.
Why coding may remain valuable—and still change sharply
Software professionals do more than turn instructions into syntax. They clarify what should be built, make architectural trade-offs, integrate with older systems, check security, test generated code and take responsibility when a production system fails. Those activities can preserve demand for experienced judgment even if routine implementation takes less time.
At the same time, entry-level programming, routine maintenance, simple application development and basic test generation may face pressure. If AI takes over tasks that once trained junior staff, employers and workers will need other ways to build practical experience. A portfolio showing tested, useful outcomes can help demonstrate skills beyond familiarity with a particular tool.
Why energy work may remain important
Energy work spans grid planning, generation and storage, infrastructure, safety, compliance, permitting, supply chains and field operations. AI can assist with modeling and analysis, but physical assets still have to be built, inspected, regulated, financed and maintained. The balance between automation and human work will vary by task and employer.
Free tools Windows power users keep installed
One-click scans. No signup required.
Why biology may remain important
AI can support literature review, data analysis, diagnostics and areas such as protein design. Biological work also depends on experiments, sample handling, clinical or field observation and real-world validation. Interpreting uncertain evidence and navigating ethical and regulatory decisions are not the same as producing an analysis; they require context and accountability.
For all three fields, resilience means that human expertise may remain useful and may be amplified by AI—not that the work is immune from automation, competition or changing hiring requirements.
What the “90% of code” claim does and does not tell you
“90% of code” could mean lines of code, boilerplate, routine implementation, coding time or a programmer’s current output. Those measures are not interchangeable. The reported argument concerns AI’s potential to handle much routine code; it does not establish that AI can independently deliver 90% of a production software system’s value or responsibility.
Even if a tool generates a high proportion of code, the effect on employment depends on what happens next. If lower development costs lead companies to build many more products and services, demand for engineers could remain strong. If output demand grows more slowly than productivity, employers could need fewer people for the same volume of work. Vembu’s estimate alone does not settle which outcome will prevail.
Who may face the greatest pressure?
Routine, structured work is more exposed than work defined by a prestigious or supposedly “safe” job title. The ILO’s findings point particularly to clerical exposure; tasks in administration, customer support, document processing, basic research and repetitive digital production can also be affected.
Effects may differ within the same occupation. One worker’s job may consist mostly of standardized document handling, while another’s includes complex client conversations, exception management and responsibility for decisions. The practical question is how much of a role’s work can be automated reliably, and what the employer chooses to do with the resulting capacity.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Productivity gains can create uneven outcomes
AI can make specialized knowledge easier to access, but the gains may not be shared evenly. Highly skilled workers who use AI effectively may produce more and become more valuable. Employers may also expect one employee to handle work previously spread across several people. Workers who lack access to tools, training or good data could fall behind.
Entry-level workers face a particular challenge if routine tasks that once provided practice are automated. That can weaken the path from junior work to senior expertise unless organizations deliberately provide mentoring, supervised projects and other ways to learn.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchBest Value
The IMF estimates that almost 40% of global employment is exposed to AI, with effects that vary across economies and jobs. Exposure includes work that AI may complement as well as work that may be displaced; it should not be read as a forecast that 40% of jobs will vanish. See the IMF’s analysis of AI and the global economy and its staff discussion note on GenAI and the future of work.
How to assess your own role
Instead of asking whether an occupation is “safe,” look at its tasks and constraints:
- How much of the work is repetitive or template-based?
- Does it require physical presence or action in unpredictable environments?
- Are goals ambiguous, and does the work require negotiation or persuasion?
- What is the cost of an error, and can AI-generated output be checked reliably?
- Are trust, personal accountability, regulation or ethical judgment central?
- Does the work produce new knowledge or rely on proprietary real-world data?
- Can an employer legally and responsibly delegate the decision to AI?
These are indicators, not a prediction formula. A role with exposed tasks may still grow if demand rises or if AI creates new work; a role with difficult-to-automate duties may still face changing workloads or fewer openings.
What workers can do now
- Learn AI within your profession. Practice using relevant tools on low-risk tasks, then check their output against reliable standards.
- Build domain expertise. Knowing the field helps you identify incorrect answers, set useful constraints and decide what should be done.
- Make verification a skill. Generated code can be insecure or fail in production; generated analysis can be plausible but wrong. Review, test and document the work.
- Strengthen communication and ownership. Requirements gathering, explaining trade-offs and coordinating people remain useful where goals are unclear or consequences matter.
- Learn privacy and security basics. Do not put confidential employer or client material into an AI service unless the organization has approved that use and its data-handling rules.
- Show outcomes, not just tool familiarity. A portfolio or work record should demonstrate useful results, sound judgment and quality control.
- Keep practicing fundamentals. Overreliance on generated work can erode the knowledge needed to catch mistakes and handle cases the tool cannot solve.
Where AI adoption can go wrong at work
Faster output is not automatically safer or better output. Generated code may pass a superficial review but fail under real operating conditions. Other risks include confidential data exposure, privacy or copyright problems, weakened institutional knowledge when experienced staff leave too quickly, and overreliance in fields such as medicine, energy, finance and public services.
Organizations can also intensify workloads rather than reduce hours, or cut junior roles without replacing the training those roles provided. Human review, clear accountability and appropriate safeguards matter most when errors are costly or difficult to reverse.
The practical takeaway
The three executives’ comments do not establish that AI will eliminate most jobs, nor that coding, energy and biology are guaranteed shelters. They point instead to a workplace in which AI can absorb routine tasks, increase the output expected from each worker and change hiring needs. For workers, the strongest hedge is not a supposedly untouchable job title: it is the combination of AI fluency with domain judgment, verifiable work, accountability and skills that meet real human or physical needs.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
