Artificial intelligence—the great job maker or taker?

The machines are coming. Electrons and silicon are being combined with increasing sophistication to replace human cognitive work. Artificial intelligence—loosely defined as computer systems performing tasks that historically required at least human intelligence—is starting to change society, including the world of chemistry.

Most developers of AI for chemistry expect their technologies to unleash a wave of discovery of useful molecules that will require many more chemists to develop and test. The mainstream view is that jobs in chemistry and the chemical industry will change in the next few years but that no massive cull within the sector is in store. Look a little further ahead, though, and it is less clear whether this positive impact will continue.

OpenAI’s AI model known as GPT (generative pretrained transformer)—a type of large language model (LLM)—can now get top marks on high school exams such as the college-level Advanced Placement Chemistry test in the US. Yet just a few years ago, it could barely join a few sentences together, according to former OpenAI researcher Leopold Aschenbrenner in his 2024 report, Situational Awareness: The Decade Ahead.

“Another jump like that very well could take us to . . . models as smart as PhDs or experts that can work beside us as coworkers,” Aschenbrenner writes. He expects that AI will soon interact with a computer “like a human” does. “That means joining your Zoom calls, researching things online, messaging and emailing people, reading shared docs, using your apps and dev tooling [computer coding], and so on,” he writes. All remote jobs will be completely automatable by about 2027, Aschenbrenner predicts.

Algorithms—the heart of all AI systems that people like Aschenbrenner have been developing—are a set of instructions or rules that enable computers to learn, analyze, and make decisions from available data. Today’s AI can slash the time it takes to discover useful chemical compounds, like new drugs or lithium-ion battery electrolytes, from years or months to weeks or days.

AI is also substantially improving the efficiency of chemical processes. For example, it can fine-tune temperatures so that no excess energy is consumed. More sophisticated algorithms—including those that can themselves program new algorithms—allied to developments in computer hardware and robotics are accelerating AI’s advance.

Some chemists believe that AI could significantly affect chemistry-related jobs, especially when combined with robotics. Christopher J. Collison, a chemistry professor at Rochester Institute of Technology who applies AI models to challenges in chemistry and chemistry education, expects that lab technicians “may be on the way out” as robots take over simpler, repetitive tasks. “This means chemists can focus on what we do best—thinking deeply about chemistry rather than spending hours on tedious processes like running a column or babysitting reactions at the bench,” he says in an email.

Looking at a higher level of sophistication, Collison says he wonders if robots might also one day replace wet-lab chemists. “That makes a lot of sense to me, at least for certain tasks,” he says. “But, as with most AI tools, it can only get you about 80% of the way there. The expertise of humans will always be essential. The danger lies in over-relying on AI—it’s just a tool, like a calculator in mathematics.”

Moreover, Collison says AI continues to be limited by a lack of data on which to train models. The software industry, including the likes of Google, Meta, and Microsoft, is throwing billions of dollars at developing models that might overcome such limitations. They include LLMs that can handle larger subject fields and have emerged as powerful tools in chemistry, significantly accelerating molecule design, property prediction, and synthesis optimization.

AI is advancing so quickly that data on how it affects chemistry-related jobs are sparse. It is “complete speculation at this point” to know whether AI will create—or take—chemists’ jobs, the American Chemical Society (ACS), an organization for chemistry professionals, says in a statement to C&EN. ACS also publishes C&EN but is not involved in editorial decisions.

In a report published in January, the World Economic Forum does forecast what might happen to jobs within chemical companies. It concludes that some positions will be lost to AI and other emerging technologies but that, overall, jobs in chemistry will increase about 10% between 2025 and 2030.

The World Economic Forum predicts a more modest net increase in the number of jobs for chemical plant operators and little change in the number of jobs for chemical engineers. The study also says surveyed employers predict that 44% of workers in chemical and advanced materials companies will need to be either retrained in their current roles or retrained and then redeployed.

Across all sectors, some 92 million jobs around the world will be displaced as a result of AI and other technological advances by 2030—equivalent to 8% of all employment, according to the World Economic Forum. Simultaneously, 170 million new roles will be created, resulting in a net increase of 78 million jobs, the organization predicts.

Major chemical companies around the world, such as BASF, Evonik Industries, and Syensqo, are not just watching the AI space but already using AI day to day in their R&D, manufacturing, and business operations. Most chemical firms appear to be hiring AI specialists. At the same time, though, many are in the process of cutting staff in a bid to reduce costs.

“There will be winners and losers,” says Richard John Carter, an independent consultant and former senior executive at BASF. “We’re not talking about the same people switching from doing a more traditional role with traditional skills to new AI-connected activities—but different people,” Carter says. “I don’t see anything fundamentally changing in terms of loss of jobs, but there’s going to be a shift.”

That chimes with Evonik, a specialty chemical maker that says it is not looking at AI as a tool for replacing staff. “AI and digitalization are about more than just a purely technological substitution,” Henrik Hahn, Evonik’s chief digital transformation officer, says in an email. The company’s plan is “not to eliminate the human factor but to augment human intelligence, making AI a companion, for example in the lab,” Hahn says.

Evonik’s use of AI includes classic and advanced machine learning algorithms, such as learning from labeled data, identifying patterns in unlabeled data, and learning through trial and error to achieve a specific goal, Hahn says. “The areas of application for AI in specialty chemicals are almost limitless.”

It can be hard to tease out jobs created by AI from ones destroyed by it. For example, the specialty chemical firm Syensqo plans to hire 15 AI-related staffers this year. Meanwhile, Syensqo CEO Ilham Kadri announced in November that the company will cut up to 350 jobs from its workforce of 13,000.

Syensqo started looking seriously at AI when it began working with Microsoft in March 2024. Just 3 months later, the Belgian company rolled out SyGPT, an intracompany chatbot based on an LLM. SyGPT uses internal data to enhance efficiency in business, R&D, and production. “It enables us to rethink the way we carry out scientific research and the way we interact with our customers and suppliers,” says Vincent Colegrave, head of AI at Syensqo.

In an upcoming version of SyGPT, Syensqo plans to add a reasoning model. “Here we see tremendous opportunities for folks in the labs and in the plants to really start looking at big, complex scientific or mathematical problems they want to solve,” Colegrave says.

The midsize German materials company AMSilk, a developer and supplier of synthetic spider silk proteins, sees AI clearly as a scientific opportunity rather than a way to reduce costs. “It’s really exciting,” says Gudrun Vogtentanz, AMSilk’s chief scientific officer. “I think it’s superpowerful, and I’m happy to see the way it is progressing.”

A few years ago, AMSilk might have taken a whole year to provide its customers with a protein for a certain application. Thanks to AI, that process now takes 2 months, Vogtentanz says. “If we don’t have to verify and retrain the model, it could soon be half that time.”

AMSilk uses the Nobel Prize–winning AlphaFold and Rosetta AI models to predict protein structures on an immense scale. AlphaFold can predict millions of proteins’ structures partly because the program, which Google DeepMind owns, uses a neural network—an AI system that emulates aspects of the human brain.

AMSilk is also using AI to enhance its production processes. In a recent project, it reduced manufacturing costs by 40% after it identified potential improvements to a dewatering step.

Even though AI is enabling AMSilk to do more research with the same number of scientists, the company is still hiring. A chemist with deep AI knowledge started at the firm in January, working alongside an existing AMSilk protein scientist with expertise in AI, Vogtentanz says. “For us, we are hiring because the company’s growing.”

BASF, the world’s largest chemical company, says it is using LLMs from OpenAI and other providers for most AI projects. The LLMs are enabling BASF to make existing data more accessible and to augment and automate processes, Marcus Pospiech, the firm’s corporate AI program lead, says in an email. “We also work on novel AI concepts that have the potential to deliver groundbreaking solutions and redefine industry standards like agentic AI,” Pospiech says, referring to humanlike AI systems that can undertake a variety of tasks in the same way a person can.

BASF announced a major restructuring in 2024 that includes the elimination of hundreds of jobs. But Pospiech isn’t willing to say that some of these jobs will be replaced by AI. “As with previous major technological leaps that were relevant to the competitiveness of the chemical industry, this technology will also change job profiles, create new positions, and render others obsolete,” he says. Overall, AI will augment rather than replace staff at BASF, he says. “We see AI mostly as a complement to humans in R&D, sales, production etc., with humans carrying the responsibility.”

Fast-forward a few years, though, and the case for hiring a human versus AI will likely tilt toward AI, many AI experts think. “Rather than having to completely remake some workflow to harvest a 25% productivity gain from a GPT-chatbot, instead you’ll get models that you can onboard and work with as you would a new coworker (e.g., just directly substitute for an engineer, rather than needing to train up engineers to use some new tool),” Aschenbrenner writes.

As big chemical companies adopt AI, with an uncertain impact on employment, a wave of AI start-ups creating chemistry technologies is emerging. Many of them are securing substantial funding—and hiring.

Founded in 2019 by University of Glasgow chemistry professor Lee Cronin, Chemify has hired more than 100 scientists, many of them chemists.

Chemify scientists have built a machine learning model that is based on a programming language and that takes characteristics from molecules and translates them into code; the model also runs automated synthesizers to make these molecules. “It’s literally a universal process language for doing chemistry,” says Cronin, who has named the language χDL (pronounced “chi-DL”). “What the programming language allows us to do is remove repeatability, remove lack of reproducibility, and also do quite complicated reactions that would take a lot of labor time.”

Depending on the atomic complexity involved, it takes Chemify hours, days, or weeks—rather than months or years without AI—to discover and start making a novel molecule. The pharmaceutical industry is already taking notice. “We are finishing deals with 6 of the top 20 Big Pharma,” Cronin says.

Cronin is convinced that the emerging digital approach to chemistry will require more—not fewer—chemists. “It’s very much a dance with the chemist,” he says. “Chemify is going to create more organic chemistry jobs tomorrow than there are today.”

Jasmine Hume, founder and CEO of Shiru, a California-based start-up focusing on AI protein development, takes a similar view. She expects AI to make chemists’ jobs better rather than replace chemists. “We think digital tools like these will make chemists’ jobs that much more interesting and challenging in new and exciting ways,” Hume says in an email.

Insilico Medicine, a powerhouse in AI-driven pharmaceutical development, is another start-up that expects AI to create a wave of chemistry-related jobs. About a decade old, Insilico has upward of 200 employees, many of whom are chemists. It has developed a machine learning model that maps the chemistry landscape to generate novel molecular structures with desired properties. To date, it has discovered 10 developmental drug compounds that are now in clinical trials.

The firm’s approach has enabled it to reduce the average time it takes to bring a novel drug compound to clinical trials from 3 years to 13 months, founder and CEO Alex Zhavoronkov says.

Zhavoronkov doesn’t think this acceleration of drug development will threaten chemists’ jobs. “The first people who are going to lose their jobs are programmers who are inefficient. They will be the first ones to be replaced because I’m firing them,” he says. “And also, I would say product managers. They should be downsized to the level where you should have AI supervisors.”

As for medicinal chemists, “I actually need more of them,” Zhavoronkov says. “When it comes to medicines, you need 800 or more experimentally validated predictive models [per medicine]. There is no way around that. And you need to have a bunch of medicinal chemists, or synthetic organic chemists who have insights into how to validate and select the top candidate molecules,” he says. “Compared to experimental validation, chemists are supercheap. So it’s better to have really, really good ones working for you, and as many as possible.”

Insilico is also looking at how it might reorient its AI models toward the design of sustainable materials. In a bid to fill knowledge gaps and accelerate the adoption of AI in materials science, the company launched the Generative AI for Environmental Sustainability Consortium last year.

There are 10¹⁰⁸ potential carbon-based molecules that could be used to make sustainable processes and products, and finding the best ones will require AI, Buff López, an analyst with the market research firm Cleantech Group, says in a 2024 blog post.

With a business that combines machine learning models with robotic lab technology, the German start-up Dunia Innovations is also harnessing AI to develop more sustainable ways to manufacture materials. “We want to innovate for the earth—for the planet,” says cofounder and CEO Alexander Hammer, a former PhD student of Cronin’s. Like many AI start-ups, Dunia is hiring. It had about 12 employees at the end of 2024 and plans to increase that to about 20 by around September, Hammer says.

The challenge is recruiting staff with deep expertise in both AI and chemistry, Dunia’s Hammer says. “Everybody’s doing something these days with AI, but finding the right people—the ones that are really qualified and really understand the space—I think that’s quite rare,” he says.

Universities are trying to respond to the skills shortage in AI, though. “What we have been seeing is that unis are starting to put AI into their [chemistry] curricula,” Hammer says.

One of Dunia’s recent hires is Ruard van Workum, who graduated from Imperial College London with an MSc in digital chemistry in 2022. Learning skills such as lab automation programming, machine learning for simulations, and multifidelity models “made all the difference for me and helped me land a role at a start-up that’s breaking new ground in chemistry,” van Workum says.

A lot of good schools are starting to teach AI as part of their chemistry syllabus, but making this trend more widespread will be difficult, says Christophe Copéret, an inorganic chemist at the Swiss Federal Institute of Technology (ETH), Zurich. While chemistry professors are critical thinkers and exponents of scientific practice, many have been in their roles for decades and are not experts in the emerging field of AI. “We have new tools, but we are not going to integrate them as fast as we wish,” Copéret says.

While a lack of AI experts in chemistry education could slow the technology’s integration into the field, for Rochester Institute of Technology’s Collison, the key thing is the way chemistry is taught. “I believe it’s critical to keep teaching chemistry the way we’ve been teaching it. A deep understanding of the subject is vital as we develop better algorithms. Right now, the level of sophistication in AI for chemistry is low—and superlow in automated chemistry,” he says.

In the meantime, Copéret believes that most chemists’ jobs will be safe despite the emergence of AI. “AI is not going to replace a plumber or cabinetmaker,” he says. Similarly, the chemical sciences will still need chemists to make drug candidates, catalysts, and more sustainable materials. “I don’t believe there will be job losses among chemists—but there will be different jobs.”

The notion that chemists will retain their jobs despite AI seems to be the dominant narrative in these early days of applying AI in chemistry. Go beyond the near future, though, and AI’s potential impact on chemistry jobs becomes less clear. It is this subsequent wave of AI that has some experts worried.

In the coming years, we’re going to develop AIs that are smarter than people, Geoffrey Hinton, a recipient of the 2024 Nobel Prize in Physics for his work in AI, told BBC Radio 4 recently. “And that’s a very scary thought.”

AI systems are on a trajectory to outperform PhDs in specialist fields such as chemistry by 2027, according to Aschenbrenner. Look ahead perhaps another 10 years, and AI models will be even more intelligent, and many thousands of them could be working 24/7 without the need to sleep or take a vacation. Unknown is whether these future AI tools will vastly enable chemists or replace them.