• 21 people were injured by pepper spray at London Heathrow Airport on Sunday, police said.
• A woman was robbed of her suitcase in a parking garage elevator.
• Passengers faced chaos as roads were closed and trains stopped serving some terminals.

Travellers faced chaos at London Heathrow Airport on Sunday, where police said some 21 people were apparently injured by pepper spray.

London's Metropolitan Police said they were called to a parking garage at Terminal 3 around 8 a.m. after a number of people were sprayed by a group of men who left the scene.

Armed officers arrived and arrested a man on suspicion of assault within nine minutes of the first report, it added.

"At this stage, it's understood that a woman was robbed of her suitcase by a group of four men, who sprayed a substance believed to be pepper spray in her direction," said Commander Peter Stevens.

He added that it occurred in an elevator, and those directly involved are believed to be known to each other.

21 people were treated by the London Ambulance Service, including a three-year-old child, the police said. Five of them were taken to hospital.

While the terminal remained open, the disruption put many people at risk of missing their flights.

Heathrow is the busiest airport in Europe, serving over 80 million passengers last year.

In an X post, Heathrow Airport advised passengers to allow extra time when travelling to the airport and to check with their airline for any questions.

A highway into Terminals 3 and 2 was closed for about an hour before reopening, according to an X post from National Highways.

It then said it closed it again on the airport's request, "due to the amount of vehicles and pedestrians within the tunnel," but reopened within 30 minutes.

The BBC reported that some passengers were seen getting out of cars and walking down a road with their luggage, towards signs that warned "no pedestrians beyond this point."

There were also delays of 45 minutes approaching the airport, according to National Highways, while Elizabeth Line trains stopped serving the terminals for over an hour.

• Student-loan borrowers are facing major repayment changes.
• Democratic lawmakers said the changes put borrowers at heightened risk of customer service errors.
• Servicing errors can be costly, leading some borrowers to put off saving for their futures.

Changes are coming to student-loan repayment, and Democratic lawmakers are warning that it could lead to costly servicing mistakes for borrowers.

On Sunday night, Democratic Sens. Elizabeth Warren, Ed Markey, Jeff Merkley, Richard Blumenthal, Cory Booker, and Chris Van Hollen sent letters, exclusively viewed by Business Insider, to all five federal student-loan servicers regarding their abilities to assist borrowers amid major changes to repayment.

The letters come as servicers are working with the Department of Education to roll out changes signed into law in President Donald Trump's "big beautiful" spending legislation, including creating new repayment plans and borrowing caps. Additionally, collections resumed on defaulted borrowers' student loans after a five-year pause, all while the administration is working to dismantle the Department of Education.

The lawmakers wrote in their letters to the federal servicers that, given the slew of changes, it's imperative that borrowers can receive help from their servicers, along with accurate paperwork processing.

"Student loan servicers' administrative errors have significant financial consequences for borrowers, including delayed rent payments, loss of mortgage eligibility, postponed retirement contributions, and even risk of homelessness," the lawmakers wrote. "Borrowers pay the price for servicers' neglect and incompetence."

The Consumer Financial Protection Bureau wrote in its 2024 annual student-loan report that servicer errors, such as incorrect payment processing and difficulties contacting customer service agents, led to "significant downstream consequences" that caused borrowers to be late on their monthly bills and even put some at risk of homelessness.

With the administration halting the CFPB's work, the lawmakers wrote that borrowers are at a greater risk of being scammed by actors posing as federal servicers, enhancing the need for clear servicer communication. They requested that the servicers answer a series of questions by December 22, including data on call center volume, staffing numbers, and communications to borrowers in delinquency or default.

"As loan repayment due dates and requirements have changed repeatedly for many borrowers under the Trump Administration, borrowers need clear and timely communication from federal student loan servicers," they said.

The consequences of servicer errors

Customer service challenges have plagued student-loan borrowers for years.

Former President Joe Biden's Education Department released a memo in 2023 detailing the errors borrowers have experienced with their servicers. It included incorrect monthly bills, late or no billing statements, mistakenly taking borrowers out of forbearance, and long hold times with customer agents. The mistakes were costly — some borrowers told Business Insider at the time that interest grew on their balances while they attempted to resolve the errors, and they could not afford the incorrect payment amounts.

The challenges persisted into the Trump administration. While Democratic lawmakers said that Trump's attempts to dismantle the Department of Education will harm borrowers due to even less staffing and oversight, the department said it plans to expand the ombudsman's office to ensure borrowers are better equipped to manage their debt.

The department is also soliciting feedback on creating a "common manual" to serve as a centralized place for servicing practices, intended to help the department bolster servicer oversight.

• One of Warren Buffett's key deputies is leaving Berkshire Hathaway.
• Todd Combs, the CEO of Berkshire-owned Geico, is joining JPMorgan to head up a new unit.
• Combs' departure comes after five years as head of Geico.

One of Warren Buffett's top lieutenants is leaving Berkshire Hathaway to join JPMorgan.

Todd Combs, one of Buffett's two investment managers and the CEO of Berkshire-owned Geico, will lead JPMorgan's $10 billion Strategic Investment Group, part of its new Security and Resiliency Initiative.

The initiative is aimed at helping companies to accelerate growth, boost innovation, and bolster manufacturing, particularly in the US.

This is a developing story.

• Gisselle Hernandez founded Glamlite when she was home with her baby.
• She grew up undocumented in New York City and considered becoming a nurse for stability.
• Today, she's a multi-millionaire and working to break family cycles of abuse and poverty.

This as-told-to essay is based on a conversation with Gisselle Hernandez, founder of Glamlite. It has been edited for length and clarity.

Five years ago, I was running a cosmetic business that had generated millions in sales. And yet, my bank account only had a few hundred dollars in it. I felt trapped in a relationship with my romantic partner because I didn't think I had the financial or emotional resources to leave.

That's when I realized that, despite building a successful company, trauma from my past was still holding me back. With a friend's help and encouragement, I left my relationship and took control of my company's finances — and it changed my life.

My family didn't teach me to dream big

My daughter and I moved into a beautiful six-bedroom house. The thing I was most excited about was having a closet to hang my clothes in: before that, even with my success, I'd been living out of a suitcase.

For me, the trauma was generational. My mom was an entrepreneur in the Dominican Republic. She was an orphan who ran a farm and a day care center. Eventually, she made enough money to invest in real estate.

When I was a toddler, we moved to the US. We were undocumented, and life was very, very, very difficult. My family never taught me to dream big: They just told me to get a job that paid enough to cover rent and food.

A college assignment helped me revisit my dreams

I dropped out of school in 9th grade and started working as a freelance graphic designer. I loved it, but my family didn't encourage me. I obtained my GED and enrolled in a community college to become a nurse.

One of my classes required me to write an essay about the concept of the glass ceiling. As I wrote about how women are disadvantaged in business and entrepreneurship, I realized I was part of the problem: I was settling for a career that felt safe, rather than the one I wanted. Soon after, I switched my major to business.

My business took off when I started taking care of myself

I graduated with an associate degree, and a few years later, I was a stay-at-home mom in LA. I knew my relationship with my daughter's father wasn't healthy, and I wanted a way out. I thought money would give me that, so I spent $1,900 on a credit card to buy selfie lights, which I sold to makeup influencers. At the time, no one else was doing that, and it took off. I did about $80,000 in the first year.

I used that money to launch my beauty brand. In June 2021, Glamlite achieved more than $2 million in sales in one day, with $1.5 million generated from the website and the remainder from wholesale orders.

My daughter and I are living a life I couldn't imagine

When I first became a multimillionaire, I had the money just sitting in my account. Because of my past, I wanted to save everything.

A wealth advisor explained to me that's not how you grow wealth. She worked with me at a slow pace. First, we put money in a CD account, rather than a savings account. Once I was comfortable with that, we moved on to stocks. I'm still learning as I go.

Today, my daughter chooses whether she wants to ride to school in my pink Mercedes G Wagon, my blue Corvette, or one of my other cars. She's living a life I couldn't imagine at her age — even if the cars are mostly a business tool for when we do influencer marketing shoots.

I always encourage my daughter's dreams. Currently, she wants to create content online, so I purchased the necessary equipment for her. She and I are both in therapy, and we're doing the work to break family cycles of domestic violence. I'm proud of everything I've accomplished, but finding the strength to break free is my greatest success.

• Canada is expanding trade ties with Asia due to stalled US negotiations and Trump's tariffs.
• Recent deals include agreements with Indonesia, the United Arab Emirates, and ASEAN talks.
• International trade experts said it may not be worth it for Canada to make a new deal with the US.

Canada has had a very busy year.

Within a year, the US's second-largest trading partner has rapidly ramped up efforts to explore new deals with regional powers in Asia, or to revive talks that had been previously put on ice.

As of November, the US has imposed a 35% tariff on all goods from Canada that are not covered by the USMCA. Canadian exports of steel and aluminum to the US also face a 50% duty. The northern neighbor has since matched some of Trump's levies, and Trump has repeatedly threatened to end trade negotiations with Canada.

International trade experts in both Canada and the US told Business Insider that the pivot is spurred by a lack of progress in negotiations with the Trump administration, which has made tariffs a cornerstone of its policy, and doubt about whether a good deal with the current administration is still possible.

Phil Luck, the director of the Economics Program at the Center for Strategic and International Studies, told Business Insider that he observed similar efforts to diversify trade during the first Trump administration, and that the current trend is "an indication" of how trade discussions have progressed with the US this year.

"Canada took our extreme turn in our trade relationship with them pretty seriously," said Luck, "because this is a very big change in how the US has sort of traditionally treated Canada, and it has really drawn the ire of the Canadian people as a result."

"So even if there was a somewhat palatable sort of middle ground, our partners also have democracies, and they need to find a solution that's palatable to their population," Luck added.

The diversification push

In September, Ottawa signed a comprehensive free-trade agreement with Indonesia, the first with the Asian Pacific country. Two months later, Canada secured a bilateral investment treaty with the United Arab Emirates that came with an expanded air-services pact. Canada is also fast-tracking a previously stalled trade agreement with India and pushing for a free trade agreement with the entirety of ASEAN by the end of 2026.

A campaign page promoting foreign investment had appeared by August on the Canadian government website, featuring a slogan that reads, "Diversification is a national imperative."

Meredith Lilly, professor of international affairs at Carleton University and a former International Trade Advisor to former Canadian Prime Minister Stephen Harper, told Business Insider that even though some negotiations with Asian countries have been in the works for a long time, the current government of Canada is touting the efforts that the country is making and accelerating them as a "response to slow and difficult negotiations" with the US.

"It is a very difficult administration to work with at the moment," said Lilly of Trump's team. "But we should have no expectation that these diversification efforts can move quickly — it's a project that we should always be undertaking, and we should be working on it consistently."

There is more that Canada can do, too, Lilly said.

One thing Canada can do more quickly to make up for the US tariffs' impact on the economy is to increase trade relations with countries it already has agreements with. That includes Korea, Japan, and the 12 Pacific Rim countries that are part of the Trans-Pacific Partnership, Lilly said, all of which are highly interested in Canada's natural resources.

According to data from the Canada Energy Regulator, Canadian oil exports outside the US reached a record 525,000 barrels per day in July 2025 and remain elevated throughout the third quarter of 2025.

Canada, while initially expecting a larger hit on the economy due to Trump's tariffs, has also reported surprisingly robust GDP growth. In the third quarter of 2025, the country reported a 2.6% GDP growth rate, according to Statistics Canada, surpassing a previously revised 1.8% projection.

A decline in confidence in the US

Carlo Dade, the Director of International Policy at the School of Public Policy at the University of Calgary, told Business Insider that it may not be worth it to make a deal with the US now, because most Canadian goods are covered by the existing USMCA trade agreement, and because many of Trump's foreign policy moves seem "personal."

"I argue that we shouldn't be trying to get an agreement with the US," said Dade. "If you look at what happened to Malaysia, Vietnam, the UK — if that's a deal, do you want that?"

Malaysia and Vietnam, despite agreeing to remove most tariffs on the US, still ended up with 19% and 20% of tariffs, respectively. The UK, despite having secured a deal with Trump fairly early in the year, is not exempt from a general 10% tariff on most goods and a 25% duty on steel and aluminum.

"No one gets a good deal from Trump. 'The Art of the Deal' says there are winners and there are losers, and the US is going to be the winner," Dade added, referring to Trump's book on dealmaking strategies. "So what you're looking for is not a good deal. You're looking for the least worst deal."

Luck said that even though the US has a lot of leverage over its closest trading partners, he is concerned about the long-term reputational damage the country will suffer among allies.

"Just because you have the biggest stick doesn't mean you have to use it," said Luck. "If you insist on cashing in on all your leverage today, you have a situation where you have the most ability to harm and therefore the most leverage over your closest allies and partners whose trade is most integrated with yours."

• Rahul Kasanagottu, 32, spent 2.5 years upskilling to transition to an AI role at Google.
• He started reading AI and machine learning books during paternity leave to kick off his job pivot.
• He said continuous learning, hands-on projects, and support were key to his transition.

This as-told-to essay is based on a conversation with Rahul Kasanagottu, a 32-year-old customer engineer at Google, specialized in AI and machine learning, and based in Austin. His identity and employment have been verified by Business Insider. The following has has been edited for length and clarity.

For a few years, I worked in a customer success role at Google, but when the generative AI train came along, I realized I wanted to get into that.

During pivotal moments in technology, a lot of people go into it with the goal of making money. I think more people should think about getting into AI to help influence how it will be used by others — and that's what I wanted to do.

My daughter was born in April 2023, and the AI boom hit right around then. Google offers generous parental leave, and I figured it would be a good opportunity to spend time with my daughter and start reading books about AI.

Paternity leave definitely helped me start the journey, but it took me about two and a half years, 11 books, and hours of watching videos to land a job on an AI team. I interviewed for around four to five different roles, and six months ago, I transitioned from a senior technical account manager to a Google Cloud customer engineer specialized in AI and machine learning. In this role, I build demos and show customers how to use Google's AI products.

I'm still up-skilling continuously. The product is changing every day. Today I'm working with one kind of customer, but tomorrow I might be working with a totally different customer whose needs are completely different. The learning curve is continuous.

Here are the 11 books and courses that helped me skill up.

Textbooks:

• "AI Engineering" by Chip Huyen
• "Designing Machine Learning Systems" by Chip Huyen

• "Designing Data-Intensive Applications" by Martin Kleppmann
• "Hands-On Large Language Models" by Jay Alammar and Maarten Grootendorst
• "Generative AI on AWS" by Chris Fregly, Shelbee Eigenbrode and Anje Barth
• "Deep Learning" by Ian Goodfellow, Yoshua Bengio, Aaron Courville

Other Books:

• "Competing in the Age of AI" by Marco Iansiti and Karim R. Lakhani
• "Prediction Machines" by Ajay Agrawal, Joshua Gans, and Avi Goldfarb
• "Power and Prediction" by Ajay Agrawal, Joshua Gans, Avi Goldfarb

• "Genesis" by Henry Kissinger, Craig Mundie, and Eric Schmidt

• "Deep Medicine" by Eric Topol

Courses and YouTube Channels:

• Deep Learning Specialization taught by Andrew Ng
• 3Blue1Brown videos on YouTube about visualizing mathematics and ML fundamentals

Books

The two books I read end-to-end were "Designing Machine Learning Systems" and "Generative AI on AWS." The latter has an accompanying course in deep learning that was very pivotal in my early learning.

The two books by Chip Huyen were my favorite. He explains things in a really approachable way and he opened my understanding of how organizations use and implement AI. At first, it was hard to understand the difference between the research side and the applied side of AI. These books helped me realize my interest lied in applied AI.

"Power and Prediction" was another favorite. It talks about how technology has to scale economically to make a difference. For example, if the electric light bulb still cost thousands of dollars, it wouldn't electrify every house today. The books talks about AI in similar terms.

"Genesis" also stood out. It talks about the future of AI and the challenges it will pose.

Andrew Ng's courses were also really helpful. He is an amazing teacher and the founder of Google Brain.

The culture at Google also helped. Without support of my manager and teammates, I wouldn't have had the time for personal growth. I had to prioritize my job and personal learning while also taking care of a new daughter. My wife also had to sacrifice a lot of hours for me to work on my own stuff.

Solo projects

The books don't typically come with assignments, but the courses come with a lot of hands-on exercises. Over time, I realized that was the missing piece in my résumé. It was getting difficult to convince hiring managers that I could do the job because it required building demos and doing hands-on projects.

I realized I had to do my own projects and the AI tools were incredibly helpful for that.

For other people who want to transition, I would say you need to keep working hard. It takes time for you to connect the dots on complex problems and sometimes you have to read the same thing again and again to nail that concept entirely.

A lot of people who want to get into AI, including me, are in a rush to land an AI job after six months. But a lot of concepts in machine learning take time to internalize. So persistence is necessary.

• Roblox CEO David Baszucki said that a recent attempt to understand all of the latest AI research was "humbling."
• "It is hard to get to a position where you can understand all of those research papers," he said on the "Access" podcast.
• Baszucki, who did the reading while on sabbatical, concluded that AI research in the 3D space was still very early.

New AI research is published at a steady clip, sometimes daily, and often featuring increasingly technical terms and concepts. How do you keep up?

Roblox CEO David Baszucki recognizes the struggle. On the "Access" podcast, the executive was asked about his recent sabbatical, where he'd spent some delving into AI.

"It was very humbling," Baszucki said. "It is hard to get to a position where you can understand all of those research papers."

Baszucki founded Roblox in 2005. In the company's early days, the CEO said that he read "all the research," from physics simulation to rendering technology.

"People were thinking about all of these new ways of doing gaming, and generally I could understand all of it," he said.

Then came the "horizontal wave" of AI research, which Baszucki called "so massive and so fast." From transformers to diffusion and world models, there's "a lot going on," he said.

Over the past few years, AI research has expanded from an academic subject to a national interest. Major tech companies, such as Meta and Microsoft, have developed in-house research labs — and awarded lucrative compensation packages to top AI researchers.

There are signs that AI research is becoming more private. In 2023, Google informed its staff that it would reduce the amount of AI research it published.

"Now it's time to compete and keep knowledge in house," a Google Brain staffer told Business Insider in 2023 of the company's philosophy.

While much of the AI industry is now focused on scaling compute, OpenAI cofounder Ilya Sutskever recently said that research itself was the key to unlocking the tech's future.

"It's back to the age of research again, just with big computers," he said.

As for Roblox, Baszucki's conclusion from his deep dive into the latest research was that it's still "very early in the 3D space."

Baszucki called AI a "very physically unnatural space," fed on human-made text and images.

"We're training AI models on this thing we made up," he said, rather than "on the 3D raw material of the world itself."

• After years of writing about Temu, I finally tried to buy a Christmas gift on the app.
• I thought it would be as simple as getting a gift from Amazon or Walmart — now I'm not so sure.
• The products were interesting, and the prices seemed low, but the shopping experience was chaotic.

Temu is not exactly new news, I know.

Chinese e-commerce giant Pinduoduo launched the marketplace in the US back in 2022, and it has since taken the retail world by storm. [user

As a retail reporter, I've written about Temu and its impact on US companies, but I never got around to actually trying it myself.

That changed on the Sunday after Thanksgiving when I wanted to find a specific Christmas gift inspired by "KPop Demon Hunters."

In the Google search results, Temu had a better-looking option than what I saw in the official Netflix store, so I decided to give the site a try, assuming it would be as straightforward as buying from Amazon or Walmart.

It was not.

When I clicked on the listing, a prompt suggested I could have the item for free if I used the app.

Doubtful, yet curious, I downloaded the app. Then it wanted my phone number — OK, fine.

Then it asked me to spin a wheel to get a deal. (The fine print says everyone gets the best deal, regardless of the apparent spin.)

Then another bonus unlocked, and another.

Soon, the app prompted me to fill a cart with additional items, assuring me that they would be sharply discounted.

Faced with an endless scroll of apparel and kitchen listings, I tried to build an order of stuff I would actually use.

But every time I got nearly ready to check out, the app offered yet another deal that required the addition of still more items.

The experience was somewhere between gambling and gaming, and I found myself losing touch with any sense of price rationality.

Exhausted, I closed the app for the night without buying anything at all.

On Monday morning, I checked the app and saw the myriad offers from the day before had expired. I had to drop the kids off at school and get to work, so I emptied my cart of everything except the original KPop shirt that got me into this situation, declined several more Cyber Monday bonus offers, and bought the one product.

I've since received two text messages a day with various offers of rewards, credits, reimbursements, and other phrases that sound like money, but involve a lot of steps to redeem.

As I was writing this article, I decided to test one of the bonus offers, a $800 coupon bundle that I could "earn" for the trouble of making a single purchase. I picked a $12 sweater and unlocked a stack of coupons for 20% off various transaction sizes, all of which expire in two days.

Meanwhile, my original order from Monday is still being prepared for shipping, and I have yet to see anything else that I actually want or need on the app. A lot of it honestly reminds me of a Sky Mall brochure.

I asked Temu about this highly gamified app experience, and a spokesperson said many users have shared appreciation for the features' added engagement.

"The interactive elements of our app are intended to enhance the shopping experience and allow customers to unlock additional discounts," the spokesperson said.

I believe I could eventually find something interesting at a great price and wait for it to arrive, but I'm not sure the time and effort required to shop on Temu's chaotic app are worth the potential savings.

I'll be sticking with my familiar e-commerce options for the foreseeable future — at least until I'm much older.

• Humanoid robots are having a moment.
• Companies are racing to build the market for robots, with Tesla's Optimus set to enter mass production next year.
• Morgan Stanley analysts identified 25 companies poised to dominate the humanoid robot boom.

The race to get in on the humanoid robot boom is heating up.

In a research note released earlier this month, analysts at Morgan Stanley broke down a list of the 25 companies best positioned to dominate the market for humanoid robots, which the investment bank estimates will be worth more than $5 trillion by 2050.

The list focuses on companies with expertise in AI and computing chips, cameras and perception, and sensors and movement technology, the analysts said.

They added that the list is intended to help investors look beyond humanoid robot manufacturers and focus on the foundational component suppliers who stand to benefit as robots become mainstream.

Morgan Stanley is not the only one making ambitious predictions about the potential for a robot boom. Elon Musk said last month he thinks Tesla's Optimus robot will be able to "eliminate poverty" and grow the global economy by a factor of 10.

Tesla is set to begin mass production of Optimus by the end of next year, but has not said how many it expects to build.

Other companies are also racing to build their own bipedal bots, including Chinese Tesla rival Xpeng, which unveiled its creepily lifelike "Iron" robot last month.

However, Morgan Stanley analysts wrote that while they estimated that more than a billion humanoid robots would be deployed worldwide by 2050, adoption of humanoid robots would be "relatively slow" until at least 2035 as the technology continues to develop.

It comes after China issued a warning last week that a bubble risked forming in its own robotics industry, with more than 150 companies competing to roll out humanoid robots.

Industry heavyweights like Nvidia, Samsung, and AMD all feature in Morgan Stanley's "Humanoid Tech 25" list. The list also includes lesser-known names such as Hesai, a Chinese lidar maker that analysts said stands to benefit as its sensors could be used to help humanoid robots improve their navigation and situational awareness.

Morgan Stanley analysts also highlighted California-based semiconductor design firm Synopsys as another potential winner, noting that its semiconductor designs had applications in humanoid robot brains. Nvidia announced on December 1 that it would make a $2 billion investment in Synopsys.

Here's Morgan Stanley's list of the 25 companies at the forefront of the humanoid robot boom:

• Baidu
• iFlytek
• Desay
• Horizon Robotics
• Alibaba
• Samsung Electronics
• NVIDIA
• Cadence
• Synopsys
• ARM
• AMD
• Texas Instruments
• Samsung Electro-Mechanics
• Onsemi
• Microchip
• Sony
• Ambarella
• NXP
• ROHM Semiconductor
• Melexis
• STMicroelectronics
• Infineon
• Renesas
• Joyson
• Hesai

Do you work in robotics and have a story to share about your company's development or workplace culture? Get in touch with this reporter at tcarter@businessinsider.com, or tcarter.41 on Signal.

• Banks are working to reskill developers as they race to adopt AI.
• Efforts include capstone-style courses, video modules, and voluntary initiatives, like hackathons.
• Tech leads said English communication with AI agents is key in the new "continuous learning" era.

Not long ago, bank software developers only had to adjust to major technological shifts every few years. Today, those shifts are arriving every other month.

Engineers are in "continuous learning mode" as they race to keep up with the rate of change, Trevor Brosnan, Morgan Stanley's global head of technology strategy, architecture, and modernization, told Business Insider.

Technologists on Wall Street today are often grinding through capstone-style courses, working on basic communication skills, or watching a bunch of YouTube-style videos as many banks help them upskill to keep pace with AI advancements.

"Something you thought you knew about AI three months ago might be out of date by now," Jonathan Lofthouse, the chief information officer at Citi, said.

Banks from JPMorgan to Citi are pouring billions into AI to make their workforces more efficient, deliver better client experiences, and ultimately cut costs. They're also scrambling to recruit the talent needed to bring those strategies to life.

The engineers who do best at these banks will be those who learn how to use AI effectively — and the banks are going to great lengths to help them learn fast, according to Alexandra Mousavizadeh, the cofounder and co-CEO of Evident, which tracks AI use in the financial industry.

And when it comes to recruiting new talent, a willingness to take advantage of the "continuous learning" is crucial.

"We are hiring for potential upskilling," Nish Rana, the senior director of Enterprise Data at Capital One, said of his hiring for the bank's AI and machine learning capabilities. Business Insider spoke with technology leaders at Citi, Morgan Stanley, and Capital One about their strategies for upskilling their engineering ranks. Citi and Morgan Stanley respectively employ 30,000 and 15,000 developers, and Capital One has around 15,000 engineers.

"It's a lot of time spent now on helping accelerate developers' adoption, so that they can become more aware of these and give back some of their capacity," Dov Katz, a distinguished engineer at Morgan Stanley, said.

Communication is key

One key part of that upskilling involves learning how to talk to agents, much as they would with another engineer.

Developers need to communicate with the newest generation of generative AI tools in English, not in code, "and not everybody in technology has a reputation for being an excellent communicator," Katz said. As Citi's Lofthouse put it, most developers' second language is Java.

"And actually, Java is sometimes a bit easier to express problems in than a first language," he said, especially when it comes to describing the complexities of the financial markets.

It's important to give an AI agent as much guidance as possible on the task to perform and the desired output, said Brosnan.

"That is a shift from, 'Okay, I'm the person writing all the code, but have a little assistance,' to now, I'm giving much bigger tasks and delegating them to an agent," Brosnan said. Everyone he works with is learning that skill in real time to use the tools most effectively. It is, he said, a "fundamental shift" in what it means to work as a developer.

That's not to say that coding is irrelevant — all of the technology leads said it's still important to know foundational coding languages, especially since humans have to check all of the code that AI agents write.

Banks are creating in-house AI courses

Capital One has developed its own AI Academies, Rana said, which are "considered a one-stop solution to grasping fundamentals, all the way up to advanced learning." The courses cover tech initiatives within the bank, including AI/ML Foundations, ML Modeling, and Research & Prototyping.

"New engineers who are coming into our company don't necessarily need to have an extensive background," Rana said. "We have formal training that can bring them up to speed in a very consolidated and accelerated way."

Morgan Stanley also offers a mix of outsourced and in-house courses, and Citi recently ran a "Techflix" series, where anyone on the technology and business enablement team could watch videos and participate in associated challenges, some of which related to AI.

Each of the banks Business Insider spoke to said they offer engineers video modules to master emerging AI skills. Brosnan said they can be as short as five minutes. He added that many employees learn a lot from YouTube, whether for work or personal interests, and that videos have proven to be a powerful tool.

Older educational efforts continue

Beyond their AI-specific upskilling efforts, the banks Business Insider spoke with have continued with older programs, like hackathons and technology all-hands meetings. Opt-in programs are hugely popular, Brosnan and Lofthouse said, and have proven to be crucial educational architecture as the pace of learning continues to accelerate.

Learning opportunities aside, the rate of change is a source of anxiety for many engineers, as some worry the shine of a computer science degree is fading in a world where bots can spit out code at record speed. Yet the tech leaders said that AI will ultimately let their engineers, equipped with the right skills, focus on higher-order thinking and higher-impact work.

"There's anxiety about the change, but it's balanced by how exciting it's going to be," Lofthouse said.

Rana said that some veteran engineers are anxious about keeping pace with the rapid shifts in AI.

Everything his team is doing — from AI academies to online videos to tech talks — is aimed at giving engineers not only the tools to succeed, but the "psychological safety net" of learning in a comfortable corporate sandbox.