You chose engineering because it felt safe. Now the design software can do your first-year tasks. Here is what is really happening to engineers.
This in-depth guide covers everything you need to know about will ai replace engineers? what the data says. Based on verified income data and real-world case studies from our database of 138 side hustle tactics.
AI is unlikely to replace licensed mechanical, civil or electrical engineers any time soon. It is already taking over pieces of the work that used to fill a junior engineer's week: first-pass geometry, routine simulation runs, load schedules and report drafting. The official forecasts still show engineering jobs growing. The pressure lands on the first three to five years of a career, the years where you used to learn by doing the boring parts. (If you write code for a living, the picture is different, and we cover it in will AI replace software engineers.)
You did the hard degree. Thermodynamics at 2am, statics problem sets, a capstone that ate a whole year. Maybe you took loans for it, and maybe your parents told relatives at dinner that their kid was going to be an engineer, the safe kind of job. Now you sit in a design review and someone mentions that the new CAD release can generate the bracket you spent a week modelling, and the simulation tool can predict the airflow you were going to spend two days meshing.
You are not imagining the shift. The tools are real, the vendors are spending billions, and the people who sign the drawings are quietly asking which tasks still need a graduate. Here is what the evidence says.
What AI can already do in engineering design
Generative design is older than the current AI boom. In May 2018, General Motors and Autodesk showed a seat bracket that was 40% lighter and 20% stronger than the original, combining eight separate parts into one, after the software produced more than 150 valid design options for engineers to choose from. That was a proof of concept. The engineer set the constraints and picked the winner. The software did the iterating.
What changed in 2025 is the scope. In September 2025, Autodesk announced a family of 3D generative AI models it calls "neural CAD", for its Fusion and Forma products. Autodesk said the models could automate ["80 to 90% of what you [designers] typically do"](https://develop3d.com/cad/autodesk-unleashes-neural-cad-3d-generative-ai-foundation-models/). That is a vendor's claim about routine work, made at a launch event, and the products were announced as coming later. Still, read the number twice. The routine work it describes is the work entry-level engineers are usually handed.
Simulation is moving the same way. Ansys, the biggest name in engineering simulation, now sits inside Synopsys, which completed the acquisition on July 17, 2025 in a deal reported at $35 billion. Its SimAI product learns from a company's past simulation results and predicts how a new design will behave. Ansys says it lets engineers "predict performance in seconds instead of hours or days" and claims it "reduces the overall design cycle by 10-100X". The page gives no independent benchmark for that range, so treat it as marketing. The direction is clear though: fewer full-fidelity runs, fewer people babysitting meshes.
Siemens made a similar bet. It closed its purchase of Altair Engineering for an enterprise value of about $10 billion in March 2025, saying the deal "extends leadership in simulation and industrial AI". On the factory side, Siemens showed an AI copilot for its NX CAM software in November 2025 that it says can cut CNC programming time by as much as 80%. Writing CNC programs used to belong to a manufacturing engineer or a skilled machinist.
Then there are the design firms themselves. In November 2025, AECOM, one of the largest engineering firms in the world, bought the Norwegian start-up Consigli for a reported 4 billion NOK, about US$390 million. Consigli markets itself as "The Autonomous Engineer", an AI agent for tasks such as space analysis, MEP loadings and report generation, and claims it can cut engineering time by up to 90%. Again, a company claim. But a firm that sells engineering hours paid hundreds of millions of dollars for software that promises to remove hours.
The work that is shrinking first
Think about what a graduate engineer actually does in year one. Drafting from a senior's markup. Running the standard analysis on a variant. Pulling load data into a spreadsheet. Writing the first version of the calc report. Checking a model against a code table. Each of those is now a product feature somewhere.
The clearest official signal sits one rung below the engineer. The Bureau of Labor Statistics projects that from 2025 to 2035, mechanical drafters will shrink by 6% and electrical and electronics drafters by 3%. BLS gives the reason plainly: CAD and BIM tools "allow engineers and architects to perform many tasks that used to be done by drafters." That happened with the software of the last decade. The question for you is which engineering tasks the software of this decade absorbs in the same way. (Architects and their drafting teams face the same squeeze, covered in will AI replace architects.)
The professional bodies have started to answer that, in a revealing way. The National Council of Structural Engineers Associations, in guidance updated in July 2025, says that if AI contributes to a design error, "it is comparable to mistakes made by interns or junior engineers." The point of that answer is liability: the licensed engineer is responsible either way. Read it as a hiring manager would, though. The profession's own framing puts the AI tool in the same box as the junior.
The same guidance says "traditional methods like mentorship and manual calculation remain important" for keeping expertise in future engineers. That sentence only needs writing if people worry the old training path is thinning out.
A question to sit with: if the tasks you were hired to learn on are now done by software in minutes, who in your firm is paying for you to learn them anyway?
How graduates are doing right now
The data on recent engineering graduates is mixed, so look at it branch by branch. The New York Fed tracks unemployment for graduates aged 22 to 27 by major, using 2024 Census data released in February 2026.
| Measure | Figure | Source |
|---|
| Mechanical engineers, US jobs (2025) | 298,500 | BLS |
| Mechanical engineers, projected growth 2025 to 2035 | 11% | BLS |
| Civil engineers, US jobs (2025) | 380,600 | BLS |
| Civil engineers, projected growth 2025 to 2035 | 6% | BLS |
| Electrical and electronics engineers, US jobs (2025) | 297,900 | BLS |
| Electrical and electronics engineers, projected growth 2025 to 2035 | 8% | BLS |
| Mechanical drafters, projected change 2025 to 2035 | -6% | BLS |
| Recent grad unemployment, civil engineering (2024) | 2.3% | New York Fed |
| Recent grad unemployment, mechanical engineering (2024) | 4.4% | New York Fed |
| Recent grad unemployment, electrical engineering (2024) | 3.2% | New York Fed |
| Recent grad unemployment, computer engineering (2024) | 7.8% | New York Fed |
| Recent grad unemployment, all majors (2024) | 4.2% | New York Fed |
Civil and electrical graduates are doing better than average. Mechanical graduates sit right around the all-majors rate, with about one in five (20.1%) underemployed, meaning in jobs that do not need a degree. Computer engineering, the branch closest to software, is the outlier at 7.8%, nearly double the average. That is the branch where AI coding tools bite first, which is why it gets its own story in our software engineers post.
These are 2024 figures. Most of the tools in the section above were announced in 2025 or later and some are still not shipping. So the numbers tell you where engineering stood just before this wave, and they say nothing yet about where it lands.
What the industry says about AI and its own people
Autodesk surveys thousands of people across design, engineering and manufacturing every year. In its April 2025 State of Design & Make report, 48% of respondents said AI will destabilize their industry, up seven points from the year before. Trust fell too: 65% said they trust AI in their field, an 11-point decline. At the same time, 46% of leaders said AI skills will be a top hiring priority over the next few years. The report's own summary line was blunt: "The next generation must master AI skills to thrive".
So the employers are saying two things at once. They are nervous about what AI does to their industry, and they want to hire people who already use it. If you are early in your career, that second part is the filter you now have to get through.
A question to sit with: when your manager says "AI skills", do they mean you can use the tool, or that the tool lets them hire one of you instead of two?
Licensing is a real wall, and it protects the signature
Here is where engineering differs from a lot of white-collar work. In most of the US, someone has to stamp the drawings. BLS notes that civil engineers "usually must be licensed if they provide services directly to the public", and only a licensed Professional Engineer can sign off on projects and approve design plans. Software cannot sit the PE exam, carry insurance or lose a licence.
The American Society of Civil Engineers made this explicit in a policy adopted in July 2024: "AI cannot serve as a replacement for the professional judgement of a licensed Professional Engineer." It adds that "AI cannot be held accountable, nor can it replace the training, experience, and judgement of a professional engineer". NCSEA says the same in its own words: "Licensed engineers are ultimately responsible for the integrity of their designs."
That is genuine protection, and it is why "AI replaces engineers" is the wrong headline. But look at who it protects. The licence protects the person holding the stamp. BLS also notes that "licensure is not required for entry-level civil engineers". The years before you are licensed, when you are logging supervised experience, are the years least protected by the licence and most exposed to the tools. A firm can keep every PE it has and still hire fewer graduates to feed them work.
There is also a slower risk. To get a PE licence you need years of qualifying experience. If the routine work that used to count as that experience moves into software, the path to the stamp gets narrower, and the people already holding stamps get more valuable. That is good news if you are 45 and licensed. It is a harder road if you are 23.
What the official forecasts say, and where they disagree
The Bureau of Labor Statistics is the most optimistic voice here. Its 2025 to 2035 projections show mechanical engineers growing 11%, "much faster than the average for all occupations", with about 17,800 openings a year. Civil engineers are projected to grow 6%, with about 22,700 openings a year. Electrical and electronics engineers are projected to grow 8%, with about 16,300 openings a year. Pay is strong: median wages in May 2025 were $104,110 for mechanical engineers, $100,840 for civil engineers and $125,040 for electrical and electronics engineers.
BLS did think about AI. In a February 2025 study of AI's effect on its projections, it wrote that for civil engineers, "strong underlying demand for civil engineering services is expected to offset the potential employment impacts of efficiency gains." In the same paper it admitted that "the magnitude of productivity enhancements offered by various AI tools remains unclear". BLS also says its methods assume technology changes at roughly the pace it has in the past. If the vendors' claims (80 to 90% of routine design work, 10 to 100 times faster design cycles) turn out even half right, that assumption breaks.
That is the disagreement in one line. The government forecast assumes AI arrives like CAD did, gradually, absorbed by growing demand. The vendors are selling it as something much faster. Nobody has 2026 data yet that settles it. Our wider look at how many jobs AI will replace covers why the big forecasts land so far apart.
A question to sit with: are you planning your next five years around the BLS number, or around what the software in your own office can do this quarter?
The counter-evidence: there is a lot to build
The honest case against panic is demand. The International Energy Agency projects that electricity demand from data centres worldwide will more than double by 2030 to around 945 TWh, and that in the US, data centres are on course to account for almost half of the growth in electricity demand to 2030. Every one of those megawatts needs electrical, mechanical and civil engineers: substations, cooling, transmission, foundations.
Infrastructure is the same story. ASCE's 2025 Report Card gave US infrastructure a C overall and a D+ for energy, with a $3.7 trillion gap between planned spending and what is needed. And engineering firms are still hiring: in the ACEC Research Institute's January 2026 survey of 628 firm executives, nearly two-thirds of firms projected increased hiring over the next 12 months, while more than half reported investing in dedicated AI-focused talent.
So the work is there. The open question is how many people it takes to do it once the tools mature, and whether firms keep paying graduates to learn on tasks a model can now handle.
Where this leaves you this month
If you are licensed or close to it, your position is stronger than the headlines suggest. Your stamp, your judgment on site and your name on the risk are what clients pay for. Learn the new tools well enough to check their output, because checking is the job that grows.
If you are in your first few years, take the junior squeeze seriously. Ask for work that builds judgment: site visits, client calls, design reviews where you defend a choice. Push toward your PE on schedule. Look hardest at the sectors with physical demand behind them, such as power, grid and data-centre work. If you want to compare where engineering sits against other careers, our list of AI-proof jobs and the 2026 job market are a fair place to start.
If you want a second income while you wait to see how this shakes out, your degree is worth money outside your day job. Engineering students pay well for help with statics, circuits and thermodynamics, which makes online tutoring a direct fit. If you know the software side, firms that want AI tools set up in their workflow need people who understand both the engineering and the automation, which is the idea behind an AI automation agency. And if you are mid-career and thinking about leaving the field altogether, read career change at 40 before you decide.