WhatsTheBigData

Airlines have used pricing systems for years to change ticket prices based on passenger demand, competition and booking trends. Now, artificial intelligence is helping airlines improve these systems and adjust fares more quickly. This has raised concerns about how airlines set prices and whether passengers could pay more for the same flight.

Startups such as JetBack, Junova and Refare are using AI tools to help travelers save money after booking. These services track flight prices and help passengers claim refunds or travel credits when fares drop.

According to a Reuters report published on October 30, 2026, these companies help travelers take advantage of airline policies that allow them to change or rebook eligible tickets without traditional change fees.

Junova analyzed 10,000 flights between September 2025 and July 2026. It found that around 22% of these flights had at least one price drop after booking. This means travelers could miss a chance to save money if they stop checking ticket prices after making a reservation.

However, the amount passengers can recover depends on the airline’s rules and the type of ticket they buy. As more airlines use AI to set prices, startups are using similar technology to help travelers find lower fares and recover money.

Airlines Are Using AI to Optimize Ticket Prices

Airline ticket prices can change multiple times before a flight departs. A fare that is affordable one day may cost more or less just a few hours later. Airlines have long used revenue management systems to adjust ticket prices based on passenger demand and seat availability.

These systems analyze factors such as travel dates, booking trends, historical data and competitor fares to decide how much to charge for each seat. Artificial intelligence is adding new capabilities to airline pricing systems. Airlines are using AI to track competitor prices, analyze passenger demand and improve revenue management.

However, the growing use of AI has raised concerns among US lawmakers. One major concern is whether airlines could use personal information and customer behavior to set different prices for individual passengers.

There is a difference between AI-assisted pricing and personalized pricing. Reuters has not established that airlines generally charge passengers different fares based on their personal information. Not every change in ticket prices is caused by AI.

As airlines adopt more advanced pricing systems, travelers may find it harder to understand how airfares are determined. Startups are now developing tools to help passengers track changing prices and make better booking decisions.

ALSO READ: Companies Are Building AI Clones of Consumers to Predict What You’ll Buy

Three AI Tools Help Travelers Save Money After Booking Flights

Several startups are using artificial intelligence to help travelers save money after booking their flights. These services track airline ticket prices even after a passenger has paid for a reservation.

If the fare drops, they help customers claim eligible refunds, credits or other savings. This gives travelers another way to reduce flight costs without canceling their existing bookings. Three companies are gaining attention for offering these services.

1. JetBack

JetBack monitors eligible flight bookings and helps travelers recover money when ticket prices fall. Founder Bryan Dubno told Reuters that the company has recovered $2.4 million for customers since launching in early 2025.

Pricing: $100 per year.

2. Junova

Junova tracks booked flights for price drops and helps passengers claim available savings. Its analysis of 10,000 flights found opportunities for travelers to save money after booking.

Pricing: 20% of the recovered savings.

3. Refare

Refare also monitors existing flight bookings and helps travelers find opportunities to reduce ticket costs. Founder Justin Dombrow has warned that changes in airline policies could make it harder for passengers to recover these savings.

Pricing: 25% of the recovered savings.

Although these startups use different pricing models, they offer a similar service. They monitor flight prices after booking and help travelers claim savings when eligible fares become cheaper.

However, savings are not guaranteed. The results depend on whether ticket prices fall and whether the airline allows passengers to claim a refund or travel credit.

ALSO READ: Meta’s New Muse AI Agent Can Send Emails, Book Travel and Make Payments

22% of Flights Became Cheaper After Booking, Junova Analysis Finds

A Junova analysis of 10,000 flights found that around 22% experienced at least one ticket price drop between September 2025 and July 2026. The results, reported by Reuters, show that passengers may find cheaper fares even after booking their flights.

The analysis also found that 40% of the flights with lower fares experienced multiple price drops. This means travelers may have more than one opportunity to save money before departure.

Source: Reuters

This shows why companies are developing services that track airline ticket prices after booking. Most travelers compare fares before purchasing a ticket but rarely check prices again once their reservation is confirmed.

For example, a passenger who buys a $500 flight ticket might later find the same fare available for $400. Depending on the airline’s rules, the passenger could recover the $100 difference through a travel credit or an eligible fare adjustment.

For people who book several flights each year, these savings could add up over time. However, not every ticket qualifies for a refund or price adjustment. Airline policies, ticket restrictions and change fees can also affect how much money travelers can recover.

How Much Money Can Travelers Save From Airline Price Drops?

The amount travelers can save depends on how much the ticket price falls and how much the price-monitoring service charges. JetBack, Junova and Refare use different pricing models, which can affect how much money passengers keep.

For example, consider a passenger who books a flight for $500. If the same flight later drops to $400, the passenger could recover $100 if the airline allows a full price adjustment without additional fees.

The type of refund also matters. Some airlines offer travel credits instead of cash refunds. These credits may have limited value for passengers who do not plan to fly with the same airline again.

Travelers who book flights occasionally may prefer a service that charges only when it recovers money. Frequent travelers may consider an annual subscription if their total savings exceed the fee.

The pricing are based on information reported by Reuters on October 30, 2026. Travelers should check each service’s latest fees, terms and supported airlines before signing up.

Why Airlines Started Allowing More Ticket Changes

The opportunity for flight price-monitoring services grew after US airlines changed their ticket policies during the COVID-19 pandemic. As travel demand fell, several airlines removed change fees on many non-refundable tickets to encourage bookings.

Before these changes, passengers often had to pay high fees to modify their reservations. For example, a $50 drop in ticket prices offered little benefit if changing the booking cost $200.

According to Reuters, US airlines collected around $2.84 billion in reservation cancellation and change fees in 2019. The removal of many change fees made it easier for passengers to benefit from lower fares after booking.

However, removing change fees does not mean passengers automatically qualify for cash refunds when ticket prices fall. Airlines may offer travel credits instead, while basic economy tickets often come with stricter restrictions.

These policies also affect how much money travelers can recover through services such as JetBack, Junova and Refare. Any changes to airline rules could limit the savings these companies help passengers claim.

Alaska Airlines Changes Its Rules for Flight Price Adjustments

Alaska Airlines changed its fare-drop policy in early October 2026, making it harder for passengers to claim savings when ticket prices fall after booking. The airline stopped handling fare adjustment requests through customer service. 

Instead, passengers must cancel their existing tickets and book again under the airline’s cancellation rules. Alaska Airlines said its main cabin and first-class tickets do not have cancellation fees. 

However, basic economy tickets have stricter conditions. Passengers who cancel these tickets at least 14 days before departure receive a credit worth only 50% of the original ticket price. This means passengers could lose money if they cancel a basic economy booking to take advantage of a lower fare. 

For example, a price drop may not offer any real savings if canceling the original ticket means losing half its value. The policy change has also raised concerns among companies that track airline ticket prices. 

Refare founder Justin Dombrow told Reuters that similar restrictions from other airlines could make it harder for these startups to recover savings for customers. Alaska’s policy shows how changes in airline cancellation rules can affect post-booking price monitoring services. 

If more airlines adopt similar restrictions, travelers may have fewer opportunities to benefit from lower ticket prices.

Can Airlines Limit AI Tools That Track Flight Price Drops?

Airlines have control over their ticket cancellation, change and rebooking policies. This gives them the power to decide whether passengers can claim lower fares after booking. As a result, changes to these policies could affect how well AI-powered fare-monitoring services work.

Airlines can introduce additional steps for passengers, restrict fare adjustments or apply stricter rules to certain ticket types. These changes could make it harder for travelers to recover money when ticket prices fall.

However, airlines may also benefit from allowing passengers to claim fare differences. Instead of receiving cash refunds, travelers may get credits that they can use for future flights with the same airline.

For example, a passenger who receives a $100 travel credit may use it to book another flight. This allows the airline to keep the customer within its booking system. This creates a mixed situation for airlines and fare-monitoring startups. While airlines want to protect their ticket revenue, they also want passengers to book with them again.

Does AI Make Airline Ticket Prices More Expensive?

AI is changing how airlines set ticket prices, raising concerns about whether passengers could end up paying more. However, AI does not automatically make flights more expensive. Airlines have used dynamic pricing for years to adjust fares based on demand, seat availability and booking trends.

AI can help airlines analyze more data and adjust prices faster. The bigger concern is whether airlines could use personal information to charge different prices to individual passengers.

US lawmakers have raised concerns about AI-based personalized pricing and how airlines might use customer data to decide what passengers are willing to pay. These practices could raise questions about price fairness and consumer privacy. AI tools can also help travelers track fare drops and find opportunities to save money after booking.

How Travelers Can Save Money When Flight Prices Drop

Travelers do not always need a paid service to recover money when flight prices fall after booking. They can check their existing reservations and contact the airline to see whether they qualify for a fare adjustment.

Before requesting a refund or travel credit, passengers should check their ticket conditions. They should confirm whether the airline allows changes without a fee and whether the lower fare applies to the same flight and itinerary.

Travelers should also check the rules for using travel credits, including expiration dates and other restrictions. A credit may be useful for frequent flyers but less valuable for passengers who rarely travel with the same airline.

AI-powered fare-monitoring tools can help track price changes automatically. However, passengers should compare the service fees and check how their booking information will be used.

A lower ticket price does not guarantee a refund. The amount passengers can recover depends on the airline’s policy, ticket type and fare adjustment rules.

AI Is Changing How Airlines Set Prices and How Travelers Save

Airlines are using AI to improve how they set ticket prices. While, startups such as JetBack, Junova and Refare are developing automated tools to help passengers save money after booking. These services monitor flight prices and help travelers claim eligible refunds or travel credits when fares drop.

Junova’s analysis found that 22% of 10,000 monitored flights experienced at least one price drop. JetBack has also reported recovering $2.4 million in refunds and credits for customers since its launch in early 2025.

However, travelers cannot always claim the full difference. Airline policies, ticket restrictions and service fees can affect how much money passengers recover. Some airlines also offer travel credits instead of cash refunds.

As airlines adopt more advanced pricing systems, automated fare-monitoring tools are giving passengers another way to track ticket prices after booking. How useful these services become will depend on airline policies and whether travelers can continue to claim savings when fares fall.

China has proposed new export controls on 23 artificial intelligence technologies. The rules could change how DeepSeek and other Chinese AI companies sell their technology to international customers.

Under the proposed rules, foreign businesses could continue using Chinese AI models through websites, apps and cloud services. However, transferring the technology behind some advanced models overseas could face new restrictions.

On October 16, 2026, China’s Ministry of Commerce proposed adding 23 AI-related technologies to its export-control catalogue. These include advanced AI models, autonomous agents and computing clusters. Public consultation will remain open until November 15, 2026. Some registration requirements for overseas technology transfers have already taken effect.

A Reuters investigation published on October 28 reported on China’s emerging approach to controlling AI technology exports. The policy could help Beijing protect valuable technology while allowing Chinese companies to earn money from global customers.

The changes could affect companies such as DeepSeek, Moonshot AI and Z.AI, which have expanded their reach through downloadable models and online AI services.

China is not placing a complete ban on AI exports. Instead, the rules could separate access to Chinese AI services from the transfer of technology. Which allows foreign companies to run and control the models themselves.

China Wants to Control How Advanced AI Technology Leaves the Country

China is considering new restrictions on the export of advanced artificial intelligence technologies as part of a broader review of its export control rules. The proposed changes cover 40 technologies across artificial intelligence, biotechnology and solar energy. Of these, 23 are related to AI.

The list includes advanced language models, AI agents that can perform tasks with limited human supervision, chip design tools and technologies used to manage large computing clusters. China’s Ministry of Commerce says the proposed restrictions aim to protect national security and prevent sensitive technologies from being misused.

The ministry has also raised concerns about AI systems that can create or modify malicious software. Such tools could pose risks to critical infrastructure if used for cyberattacks. However, China does not intend to block all AI technology exports. Most AI technologies would remain eligible for export under the proposed rules.

This approach would allow Chinese AI companies to continue serving international customers while keeping tighter control over technologies that Beijing considers strategically important. The proposed restrictions could also help China protect valuable AI research and models while allowing domestic companies to generate revenue from overseas markets.

ALSO READ: China’s AI Industry: Key Statistics and Data (2025–2026)

Foreign Customers Could Use Chinese AI Without Owning the Models

China’s proposed AI export rules could distinguish between using an AI model and owning the technology behind it. Foreign customers may be able to access Chinese AI models through APIs, websites and cloud services. 

This would let businesses use AI tools without owning or running the underlying models themselves. Transferring the technology is different. It could allow a foreign company to run, modify or reproduce an AI model independently.

Under the proposed rules, China would treat these two types of AI distribution differently. This could allow Chinese AI companies to serve international customers while maintaining greater control over their models and underlying technology.

For example, a developer outside China could continue using a Chinese AI language model through an API. The developer would send requests to the provider’s servers and receive responses without downloading the entire model.

However, running the same model independently could require access to its underlying technology. This type of transfer may face additional regulatory requirements under China’s proposed AI export rules.

This approach is similar to the business models used by major AI companies. They charge customers for access to their models while keeping control of the technology and computing systems behind them. China’s proposed rules would add another layer of control by allowing the government to regulate certain technology transfers.

For international developers, this could mean greater reliance on Chinese AI providers. It could also limit their ability to run some advanced models on their own servers. The actual impact will depend on which AI models are covered by the final rules and how Chinese regulators enforce the restrictions.

China’s AI Export Rules Could Affect DeepSeek, Moonshot AI and Z.AI

China’s proposed AI export restrictions could affect how companies such as DeepSeek, Moonshot AI and Z.AI distribute their models internationally. These companies offer downloadable AI models, APIs and hosted services. The new rules could change how they share future models with overseas customers.

According to Reuters, researchers at major Chinese AI labs are seeking clarity on whether advanced models can be trained or run on servers outside China. This matters because some Chinese AI companies rely on overseas computing facilities to access advanced chips restricted by U.S. export controls.

If moving certain models abroad requires government approval, companies may need to change how they train and deploy their AI systems. One possible outcome is that companies develop separate versions for Chinese and international markets. 

However, no such requirement has been confirmed for DeepSeek, Moonshot AI or Z.AI. The impact will depend on the final rules and each company’s business model.

China Proposes Three Tests for Restricting Advanced AI Models

China’s proposed AI export rules include three criteria to identify advanced models that could face restrictions. These tests focus on training computing power, model size and performance.

However, applying these thresholds may not be easy. Some AI models contain trillions of parameters but use only a portion of them for each task. Other models achieve strong results with fewer parameters.

Chinese law firm Han Kun has raised questions about how regulators would measure model size for systems that use this approach. The performance test also remains unclear. The draft does not specify which rankings or assessments would determine whether a model qualifies for restrictions.

This could make it difficult for AI companies to know whether their models fall under the proposed rules. Even a smaller model could qualify based on its performance in a specific assessment.

Liu Jing, a senior partner at Beijing Dacheng Law Offices who took part in policy consultations, told Reuters that companies want clearer guidance on these tests. The final regulations and implementation guidelines will determine which AI models are subject to export restrictions.

China’s AI Export Rules Show Growing Competition With the US

China’s proposed AI export restrictions come amid growing technology competition with the United States. 

Washington has restricted China’s access to certain advanced AI chips and semiconductor manufacturing equipment. These measures aim to limit China’s access to computing technology that could support military applications and advanced AI development. 

China’s proposed rules focus on a different area. Instead of limiting the technology Chinese companies can acquire. Beijing wants greater control over which advanced Chinese technologies can be transferred to foreign organizations.

Yang Jingru, a partner at Chinese law firm Jincheng Tongda & Neal, described the approach as a counterpart to US restrictions on computing hardware, investment and other technologies.

The two countries are focusing on different parts of the AI industry. The US has targeted China’s access to advanced chips. Meanwhile, China is considering tighter controls on the overseas transfer of AI models and technical knowledge.

ALSO READ: China’s AI Bubble Has Burst, Wiping Out $480 Billion in Market Value

Beijing says the proposed rules would allow legitimate technology exchanges to continue. They would also prevent sensitive technologies from being transferred without proper oversight. The measures show that advanced AI models are becoming strategic assets. They are also playing a growing role in the technology competition between China and the United States.

China’s Open-Source AI Strategy Could Face New Restrictions

Chinese AI companies have gained global attention by releasing models that developers can download and run on their own systems. These models allow researchers and businesses to modify the technology and build applications without relying entirely on the original provider. 

This has helped Chinese AI models compete with proprietary systems that customers access through paid APIs. However, China’s proposed AI export rules could make it harder to distribute some advanced models internationally. 

If transferring certain technologies requires government approval, companies may prefer to offer access through APIs and cloud services instead of allowing customers to download the models. This could push some AI companies toward subscription services and API-based revenue.

However, China’s Ministry of Commerce says technology already available through public research or open-source releases would generally remain outside the proposed restrictions. The draft also includes simplified procedures for eligible open-source technology exports.

This shows that China doesn’t plan to stop open-source AI development entirely. Instead, the government wants greater control over the export of certain sensitive technologies.

Some Chinese AI models could remain freely downloadable, while others may become available mainly through hosted services. The final rules will determine how these changes affect AI developers and international users.

ALSO READ: Open Source AI Statistics

MiniMax Completes AI Technology Export Registration in China

Chinese AI company MiniMax has taken an early step to meet China’s technology export requirements. On October 23, MiniMax announced that it had completed a technology export registration with China’s Ministry of Commerce.

According to Reuters, the registration covered the open-source release and overseas licensing of certain recently released and upcoming AI models. The move shows that Chinese AI companies can continue to distribute models internationally while following export requirements.

However, MiniMax has not disclosed full details about the models covered by the registration or the conditions governing their use abroad. This also shows why AI companies may need to consider export rules when planning new model releases and licensing agreements.

How China’s AI Export Rules Could Affect Global Data Centers

China’s proposed AI export rules could change where Chinese AI companies train and run their models. Some developers use overseas data centers to access advanced processors and support international operations. 

The new rules could affect how they transfer AI models between facilities in China and other countries. For example, moving a restricted model to a foreign-owned data center could face different requirements from running it at a facility operated by the Chinese company itself.

China’s proposal offers simplified licensing for certain overseas computing centers run by Chinese companies. However, it does not provide the same clear treatment for transfers that allow foreign customers to build and operate similar systems independently.

This could encourage Chinese AI companies to maintain greater control over their overseas infrastructure. It could also affect partnerships involving AI model hosting, technology licensing and joint development. The final rules will determine how these requirements affect international operations and compliance costs.

Foreign Investment in Chinese AI Could Become More Complicated

The proposed rules could also affect foreign investments and acquisitions involving Chinese AI companies. Reuters reported that China introduced registration requirements on October 16 for certain transfers that put Chinese technology under foreign ownership or control.

This means the rules may cover more than AI models and cloud services. They could also apply to deals involving valuable AI technology and intellectual property. Foreign companies planning to buy a Chinese AI business may need to check whether the deal involves technology covered by the restrictions.

International partnerships could also face additional checks if they involve the transfer of sensitive technology or control over key assets. Governments are paying closer attention to AI investments because of their potential national security risks. This could make it harder for foreign investors to determine which deals need regulatory approval.

However, the proposed rules do not ban all foreign investments in Chinese AI companies. The requirements would depend on the technology involved and the details of each transaction.

How China’s AI Export Rules Could Affect Global Developers

Developers outside China may see little immediate change under the proposed rules. The proposal generally excludes access through APIs, websites and software-as-a-service (SaaS) platforms from the specific controls on AI model technology exports.

This means developers could still use hosted Chinese AI services to build their applications. However, downloading and running AI models independently could become more difficult. Some businesses prefer to run models on their own systems. This gives them more control over costs, data security, customization and access.

These benefits are important for companies that must follow strict privacy rules or industry regulations. If advanced Chinese AI models become available mainly through hosted services, developers may have to rely more on the companies that provide them. They could face changes in pricing, service availability or access policies.

They may also have fewer options to modify models or study how they work. Hosted AI services have their own advantages. They allow businesses to use powerful models without buying expensive computing equipment or managing the infrastructure themselves.

The impact will depend on each company’s needs. Some businesses may see little difference, while those that rely on downloadable AI models may need to reconsider which systems they use once China finalizes the export rules.

Could China Earn More From Hosted AI Services?

China’s proposed rules could encourage AI companies to earn more from hosted services instead of downloadable models.

When companies release model weights, developers can run the models on their own systems without paying for every request. Providers may still earn money through licences and support, but they have less control over usage.

Hosted services work differently. Customers pay to access models, allowing providers to earn recurring revenue while keeping their technology private. This approach could help Chinese AI companies reach global customers while protecting valuable technology.

However, downloadable models have helped Chinese developers attract users worldwide. Strict limits on their distribution could slow adoption.

Companies may choose a middle ground by releasing some models publicly while offering their most advanced systems through paid services. However, this remains a possible business strategy, not a confirmed requirement under the proposed rules.

China Is Changing the Rules of the Global AI Race

China’s proposed export rules could change how its AI companies compete with US developers. Until now, much of the debate has focused on whether Chinese companies can access enough advanced chips to build powerful AI systems. 

Beijing is now looking at another issue: how much of its AI technology foreign companies should be allowed to control. The proposed rules suggest that China wants its AI companies to reach global customers while keeping sensitive technology under tighter control.

This could affect downloadable models, overseas data centers, international partnerships and foreign acquisitions. It could also push more businesses towards hosted AI services instead of running models on their own systems.

However, the rules are not final. Public consultation runs until November 15, and companies will need further guidance to understand how the restrictions will apply. China is not proposing to stop foreign users from accessing its AI services. 

Instead, the focus is on controlling the transfer of technology behind certain advanced AI systems. This approach could influence how Chinese and American AI companies compete in global markets.

Fashion brands are using AI models to create product images and advertising campaigns at lower costs. This helps retailers save time and money. However, it also raises concerns about job losses among human models, photographers and other creative workers.

Retailers can now create images of virtual models wearing different outfits without arranging traditional photoshoots. Some companies report production cost savings of more than 90%.

An October 27 Reuters investigation found that major fashion and beauty brands had sought permission to alter models’ appearances. The proposed terms also included swapping faces and creating digital replicas. 

Reuters interviewed 15 modeling agency directors as part of its investigation. It also reviewed more than a dozen confidential contracts. It involved various brands such as Diesel, Calvin Klein and L’Oréal. Zalando has expanded its use of AI-generated images, while its associated business, Scayle Studios, says the technology can cut production costs by more than 90%. 

The shift raises concerns about models’ jobs, consent and payment. Should models receive additional compensation when brands reuse their digital likenesses, or can a single contract give companies lasting rights to their images?

Fashion Brands Can Now Create AI Models Without Traditional Photoshoots

Traditional fashion photoshoots require several steps, from hiring models and photographers to arranging studios, styling outfits and editing images. For retailers launching hundreds or thousands of products, these tasks can increase production costs and delay marketing campaigns.

Generative AI is giving fashion brands another way to create product and advertising images. Retailers can use AI to generate realistic models in different outfits, change image backgrounds, and produce multiple campaign versions. 

Some brands also use digital versions of existing models to showcase new clothing without arranging a separate photoshoot for every look. This helps companies produce more marketing images in less time and potentially lower their production costs.

In May 2025, Reuters reported that Zalando was cutting marketing production costs by 90% using generative AI. The European fashion retailer said the technology had reduced the time needed to produce marketing images from six to eight weeks to around three or four days.

Zalando also reported that AI-generated images accounted for approximately 70% of its editorial campaign images in the fourth quarter of 2024. This is why fashion retailers are adopting AI image generation. 

Instead of organizing a new photoshoot for every campaign, brands can create different image variations. This allows them to launch marketing materials more quickly as trends change. However, these savings may vary across companies. 

Zalando’s reported 90% reduction does not mean every fashion brand can achieve the same results or cut its overall photography budget by that amount. Actual costs depend on campaign requirements, image quality, model licensing and the level of human editing and review involved.

Zalando’s AI Fashion Business Has Generated 190,000 Outfits in One Month

Zalando and About You are expanding their use of AI-generated fashion images to reduce production time and costs. Zalando, one of Europe’s largest online fashion retailers, has closed its two photography studios in Berlin as it shifts more work to AI-generated content. About You, acquired by Zalando in 2025, has also shut its studio.

Reuters reported the changes were accompanied by more than 340 job cuts at Zalando, although some affected employees moved to other roles. About You is now expanding its AI content services through Scayle Studios, which produces digital fashion images for other retailers. 

The business says its technology can reduce production costs by more than 90%. Its reported output shows how quickly AI fashion imagery is scaling, with 190,000 outfits generated in a single month.

Source: Reuters

Julian Jansen, About You’s vice president of marketing, told Reuters that using AI was helping the company save significant amounts of money.

He said traditional photoshoots often struggle to keep up with online fashion retailers, which regularly launch new products. AI allows brands to produce large numbers of fashion images faster than conventional photography.

This can reduce the need for large production teams and help retailers create marketing content with fewer resources. However, the shift also raises concerns about the future of jobs in fashion photography and related creative roles.

Human Models Face Job Losses as Fashion Brands Adopt AI

AI-generated fashion images could threaten one of the main sources of income for working models. Online retailers regularly hire models to showcase new clothing collections, providing steady work beyond high-profile advertising campaigns and magazine shoots.

Adot Gak, a London-based model who has appeared in Harper’s Bazaar and worked with brands such as Balmain and H&M, told Reuters that AI could reduce job opportunities for models. She highlighted ecommerce photography as one of the areas most at risk.

Retailers can now use virtual models to display clothing across product pages and marketing campaigns. This could reduce demand for traditional photoshoots and affect other workers, including photographers, makeup artists, hairstylists, stylists and studio staff.

However, the impact may vary across the fashion industry. Luxury brands may continue hiring human models and photographers for high-profile campaigns, while mass-market retailers may use AI for routine product images.

As more fashion brands adopt AI, routine modelling jobs could face greater pressure than high-profile fashion assignments.

ALSO READ: AI Job Automation Statistics

Fashion Contracts Could Allow Brands to Reuse Models’ Faces Without Extra Payment

AI is raising concerns beyond job losses in the fashion industry. Some brands are seeking contract terms that give them greater control over models’ digital images and allow them to reuse those images in future campaigns.

A Reuters investigation found proposed contracts that included permission to change facial features, body shapes and expressions, swap faces between models and create digital replicas. Some agreements also sought permission to use model images to train AI systems.

These terms could allow brands to create new advertising images from an existing photoshoot without hiring the model again. Depending on the contract, the model might not receive additional payment for the new images.

Patrick Staal, head of Barcelona-based Uniko Models, told Reuters that brands were increasingly requesting rights to alter models’ faces, bodies, skin, ages and expressions.

Reuters also examined proposed contract terms involving Diesel, Calvin Klein and L’Oréal. Diesel’s casting document included permission for extensive digital changes, while Calvin Klein said it was testing AI technology while respecting models’ rights. L’Oréal said its standard AI editing tools did not affect models’ images but did not directly address the specific contract language.

However, proposed contract terms do not prove that brands have used all the rights described. Some agreements reviewed by Reuters were unsigned. Still, the documents highlight growing concerns about how fashion brands could use models’ digital identities and whether they would receive additional payment.

Digital Clones Could Create a New Problem for Models’ Rights

Digital twins are computer-generated versions of real people that can reproduce their appearance in different settings. In fashion, brands can use this technology to show the same model wearing hundreds of outfits without arranging separate photoshoots.

However, digital replicas raise concerns about consent, ownership and payment. Models may have questions about who controls their digital likeness, whether brands can change their appearance without permission and whether they should receive extra payment when their digital image appears in new campaigns.

Melanie Howard, a partner at law firm Loeb & Loeb who works with brands and talent, warned Reuters that broad contract terms can make it difficult for models to challenge unauthorised uses of their images.

The issue could be particularly serious for new models who may feel pressured to accept wide-ranging licensing terms to secure work. Agreements signed early in their careers could affect how brands use their digital likeness in the future.

As AI-generated models become more realistic, clear contracts, consent and fair payment will become increasingly important to protect models’ rights and future earning opportunities.

ALSO READ: Your Face and Voice May Become the Next Form of Digital Property

New York Has Introduced Protections for AI Fashion Models

Governments and industry groups are taking steps to address concerns about AI-generated models, but legal protections still vary by location.

New York’s Fashion Workers Act, which took effect in June 2025, requires consent for the creation or modification of digital replicas of models. The law is significant because New York is a major center for fashion brands, modeling agencies and advertising.

However, it does not provide a global standard for AI fashion models. Brands and models often work across different countries, where legal protections can vary.

In Britain, the British Fashion Model Agents Association has introduced a voluntary AI code of practice. Its recommendations include obtaining models’ consent before using AI and ensuring they retain ownership of their digital twins. Association head John Horner told Reuters that models remain at risk without legally binding protections in Britain and Europe.

Sara Ziff, founder and executive director of the Model Alliance, has also pushed for stronger legal safeguards and supported New York’s legislation.

The debate is less about banning AI from fashion and more about setting clear rules for consent, fair payment and control over digital likenesses. These protections could help fashion brands adopt AI while respecting models’ rights.

Luxury Fashion Brands Are Taking a More Cautious Approach

Fashion brands are adopting AI at different rates. Fast-fashion retailers have strong reasons to use AI-generated models because they need large numbers of product images and frequently launch new collections.

Luxury brands face different challenges. Their campaigns often rely on well-known models, creative photography and the reputation of the people featured in their advertisements. Replacing human models with digital characters could lower costs but may also affect how customers view a brand.

Some luxury brands have already tested AI imagery. Moncler used AI-generated visuals in a 2023 campaign, while Gucci faced criticism after sharing AI-generated images on Instagram in 2026.

Moncler told Reuters that it does not currently use AI-generated models on its website or in advertising campaigns. Gucci did not respond to Reuters’ request for comment. Meanwhile, fashion retailer Mango has introduced AI-generated models in its marketing.

These examples show that AI adoption varies across the fashion industry. The technology may become more common in routine ecommerce photography before seeing wider use in major luxury advertising campaigns.

Could AI Models Eventually Replace Human Fashion Models?

AI technology can now create realistic fashion images, show clothing on virtual models and produce multiple image variations without arranging new photoshoots. These features help retailers reduce production time and costs.

However, replacing some modelling tasks does not mean AI can replace the entire profession. Human models take part in runway shows, work with creative teams, represent brands and connect with audiences in ways that virtual models may not fully match.

Customer trust is another concern. Some shoppers may prefer images featuring real people. AI-generated pictures can also misrepresent a garment’s fit, texture or proportions, making it harder for customers to judge products accurately.

For now, routine ecommerce photography faces a greater risk of automation than the entire fashion modelling industry. The long-term impact on runway shows, luxury campaigns and celebrity advertising remains uncertain as brands continue testing AI technology.

The Fashion Industry Faces a Choice Over Who Benefits From AI

AI fashion models can help retailers cut image production costs by more than 90% in some cases. For brands that produce thousands of product images each month, the savings can be significant.

However, the growing use of AI raises questions about how these benefits should be shared. Should models receive ongoing payments when brands reuse their digital likenesses? Can new creative jobs make up for the work lost by photographers and studio staff? Should retailers tell customers when they use AI-generated fashion images?

These issues will shape the future of AI in fashion. Brands are already testing digital replicas, closing photography studios and changing how they produce marketing content.

AI is likely to remain part of fashion production, but workers’ rights and fair payment need greater attention. The key challenge is ensuring that fashion brands can benefit from lower costs without leaving models and other creative workers to bear the cost of these changes.

An AI startup has raised $300 million to develop technology that simulates how people shop and make buying decisions. These tools are known as synthetic consumers or AI digital twins. They help businesses test products, advertisements and pricing strategies using virtual customers.

Simile, an AI startup valued at a reported $2 billion, is developing systems that predict consumer behavior. According to an October 27 Business Insider report, the company works with businesses in retail, finance and entertainment. 

Pharmacy chain CVS has used its technology to study customer responses to store layouts and marketing messages. However, questions remain about the accuracy of AI-generated consumers. 

A September study by the Pew Research Center examined nearly 300 questions. It found that AI-generated survey responses differed from real human responses by an average of 12 percentage points. The study raise concerns about whether businesses can rely on AI simulations to predict real consumer behavior.

A $300 Million Startup Wants to Build Digital Versions of Consumers

Simile is building an AI model that aims to predict human behavior. The company wants businesses to use virtual consumers to understand how people might react to different products and prices. They can also use these virtual consumers to test marketing campaigns.

Simile co-founder and CEO Joon Sung Park explained the idea during an appearance on the Sourcery podcast. The system uses information about an individual or group to predict how they might behave in different situations. 

Businesses can then interact with these virtual consumers through a web interface and ask questions about their preferences. For example, a retailer could create a virtual group of fitness enthusiasts to find out whether they would buy a new smartwatch priced at $199 or $249.

A supermarket could test whether moving products to another aisle might increase sales. Similarly, a film studio could use the simulations to understand how different audiences might respond to a new movie or advertising campaign.

Park said businesses can choose the type of people they want to study and ask the simulated group as many questions as they need. The goal is to make market research faster and easier. Traditional research often requires companies to recruit participants, conduct interviews, run surveys and analyze the results. These steps can take time and money.

Simile wants businesses to use AI simulations to explore ideas before deciding whether they need to conduct further research with real consumers. According to Park, Simile has raised $300 million and reached a valuation of around $2 billion. However, this funding does not prove that its AI simulations can accurately predict real consumer behavior.

CVS Is Using AI Consumers to Test Shopping Experiences

CVS Health is among the companies reported to be working with Simile. The pharmacy and healthcare company is using AI simulations to study how customers might respond to changes in store layouts, messages and shopping experiences.

Retailers regularly make decisions about product placement, promotions and store design. Traditionally, they rely on surveys, interviews and customer tests to understand what shoppers prefer.

AI simulations offer another way to test these ideas. Retailers can compare different store layouts and see how virtual shoppers might respond before making changes. Simile also works with financial institutions and entertainment companies, according to CEO Joon Sung Park.

Banks could test customer reactions to financial products, while entertainment companies could study audience preferences before launching campaigns. However, virtual shoppers cannot guarantee real sales. 

Actual buying decisions depend on factors such as personal budgets, changing preferences and in-store experiences. For now, AI consumer simulations can help businesses test ideas, but they cannot replace real customer feedback or reliably predict every purchase.

How Do Companies Create AI Clones of Real Consumers?

AI consumer models use data about people to create virtual profiles that can answer questions and simulate shopping decisions. The data may include survey answers, age groups, buying habits and other information about consumer behavior.

Simile says it works with organizations such as Gallup to collect data from people who have agreed to take part in research. CEO Joon Sung Park said these partnerships give the company access to information linked to more than 40 million people worldwide.

The company also uses transaction and mobility data to improve its predictions. For example, a grocery retailer could use these models to study how people in a particular age group might respond to a new product. The virtual consumers could then provide simulated responses about pricing, product preferences and advertising.

However, AI consumer models do not always create exact digital copies of real people. Some simulate the behavior of broad groups, while others use data linked to specific individuals.

The accuracy of these predictions depends on the quality of the data, how the models are built and whether their results match real consumer behavior. A simulated response does not guarantee that an actual customer will make the same choice.

Companies Are Turning AI Consumer Clones Into a Marketing Business

Simile is not the only company developing AI tools to study consumer behavior. On October 20, Resonate announced Resonate Digital Twins, a product that uses data linked to real people to simulate consumer responses.

The company says its system covers more than 250 million American adults matched to individual records. Its data resources include over 100 million consumer research responses, more than 30 billion observed behaviors daily and over 15,000 attributes.

Resonate says marketers can ask questions about consumer preferences and turn the results into advertising audiences within five days. The company aims to connect AI-based research directly to advertising. 

Businesses could use the system to identify people who may be interested in a product and target those groups with ads. This approach could help companies move from testing marketing ideas to launching campaigns more quickly.

However, the size of Resonate’s database does not mean that every person has agreed to have an AI clone created. Its claims about the accuracy and performance of the system also need independent testing.

ALSO READ: AI In Content Marketing Statistics

Companies Developing Synthetic Consumer Technology

Several companies are developing synthetic consumer technology for product research, audience analysis and marketing. The table below highlights key companies working on AI-based consumer simulations and their main uses.

Advertising agencies and consumer brands are also exploring AI-generated consumers. According to Digiday’s October 12 report, Goodway Group is working with Arima to use synthetic population data for media planning. 

The agency wants to test different advertising budgets and strategies before spending money on actual campaigns. Brands such as Magnum, Vans and Kenvue have also tested synthetic personas for product development and market research.

AI Consumer Predictions Can Be Wrong, Pew Research Finds

Independent research has raised concerns about how accurately AI can predict human responses. On September 30, Pew Research Center published a study testing whether AI-generated respondents could give answers similar to those of real people.

Researchers used AI models to simulate participants from Pew’s American Trends Panel. They gave the models information about real participants and asked them to answer questions from surveys that people had already completed. This showed clear differences between AI-generated and human responses.

Across nearly 300 survey questions, AI answers differed from real survey results by an average of about 12 percentage points. The gaps were larger for some demographic and political groups.

Source: Pew Research Center

Pew found that nearly half of the survey questions had at least one answer option that no AI respondent selected. This suggests that AI models may fail to capture the range of opinions held by real people.

Researchers concluded that the tested methods could not replace traditional public opinion polls. The study did not test Simile or Resonate, so its findings do not prove that all AI consumer models are inaccurate. However, businesses should verify AI predictions with real customer feedback to avoid poor product decisions and ineffective advertising campaigns.

Some Studies Report Better Results From Synthetic Consumers

Despite Pew Research’s study, other research suggests that AI consumer models can perform well on specific tasks. In a May 2026 analysis, Boston Consulting Group (BCG) explored how businesses could use AI-generated consumer groups to test products and pricing. 

One experiment predicted actual consumer choices with 92% accuracy after the model was refined. BCG said synthetic research could help companies test more product ideas and speed up development. However, it also found that these models were less effective at predicting reactions to completely new products.

AI models learn from existing data, so they may predict preferences for familiar products more accurately than reactions to something unfamiliar. For example, a model might compare preferences for two popular beverage flavors but struggle to predict demand for a new type of drink. 

BCG also warned that synthetic consumer panels should not fully replace traditional market research. Pew and BCG studied different tasks, which helps explain their different findings. Pew tested AI responses to public opinion surveys, while BCG examined specific consumer choices.

Could Companies Predict Your Purchases Before You Make Them?

Synthetic consumer technology aims to help businesses predict what people might buy in the future. Retailers already use shopping history, browsing activity and customer preferences to recommend products. AI simulations take this a step further by testing how consumers might respond to new situations.

For example, a clothing brand could compare prices for a new collection. A supermarket could test a new store layout, while an advertising agency could compare campaign ideas before launching ads.

This approach could help businesses test thousands of ideas without conducting a separate survey for each one. However, predicting a group’s preferences is different from knowing what an individual will buy. Personal budgets, changing needs and social influences can all affect purchasing decisions.

Even models trained on large amounts of data cannot account for every factor. There is no established evidence that these systems can consistently predict an individual’s future purchases with certainty.

AI Consumer Clones Raise New Privacy Questions

AI consumer models raise concerns about how companies collect and use personal data. These systems may use shopping habits, location patterns, age groups and personal preferences. Some rely on research from people who agree to participate, while others combine surveys with commercial data and observed behavior. 

The risks depend on how the information is collected, stored and used. Simulating a broad group is different from building a detailed profile linked to a specific person. Resonate says its system connects predictions to audience records that marketers can target with ads. 

This raises questions about whether consumers know how their data is used to make AI-based predictions. Companies could also use this information to estimate personal details that people have never shared directly. 

Even if these estimates are wrong, they could affect the ads, offers and recommendations people receive. Privacy concerns are also growing around AI shopping assistants. 

In September, Reuters reported that major banks had warned about fraud, scams and data privacy risks linked to AI agents that shop for customers. Although the technologies work differently, both could change how AI influences buying decisions and the relationship between businesses and consumers.

ALSO READ: 23+ Alarming Data Privacy Statistics For 2026

Will AI Replace Traditional Market Research?

AI consumer models can help businesses test ideas faster and at a lower cost. Companies can ask virtual customers questions and compare different products, prices and marketing strategies without running a new survey each time.

However, relying entirely on AI comes with risks. Real consumers can change their minds, express unexpected opinions and respond to new trends that AI models may miss. These systems can also reflect biases in their training data.

Pew Research found that AI-generated responses did not always match real public opinion. BCG’s research, meanwhile, suggests that synthetic panels can help with certain consumer decisions when their results are tested.

A better approach may be to combine AI simulations with traditional market research. Businesses can use virtual consumers to test initial ideas and then ask real customers for feedback before making major decisions.

This shift could also change the role of market researchers. They may spend more time checking AI predictions, finding errors and comparing results with real consumer behavior. Whether this technology will create new jobs or reduce demand for some research roles remains unclear.

The Future of Shopping May Begin With Your Digital Twin

Synthetic consumers are changing how businesses use AI. Instead of using it only to create ads, answer questions or recommend products, companies are now using AI to predict how people might respond to different products and offers.

Simile’s $300 million funding and reported $2 billion valuation show investor interest in this technology. Resonate’s digital twins also show how AI-generated consumer profiles could help companies plan advertising campaigns.

Retailers, healthcare companies and marketing agencies are already exploring these tools. However, their long-term success will depend on how accurately they predict real consumer behavior.

Companies must also address concerns about consent, privacy and the use of personal data. These issues become more important when AI predictions influence major business decisions. For consumers, the change may happen without them noticing. 

An advertisement, product recommendation or store layout could be tested on an AI model representing people with similar interests and habits. AI may help businesses predict what customers are likely to buy. But understanding why people make those choices remains a bigger challenge.

Banks still rely on COBOL, a programming language developed in 1959, to process financial transactions. Now, OpenAI, Anthropic, IBM and Amazon are developing AI tools to help banks modernize these aging systems without disrupting essential services.

COBOL remains a key part of banking infrastructure, supporting systems that handle ATM withdrawals, credit card payments and account balances. However, replacing this software is difficult because decades of business rules are built into millions of lines of code, much of which is poorly documented.

AI coding tools could help banks analyze legacy applications, identify connections between systems and translate old code into modern programming languages. An October 27 report from The Verge examines whether these advances could make COBOL modernization easier.

Early results are promising, but significant challenges remain. Translating code does not guarantee that a replacement banking system will work correctly. Even a small error could affect payments, account balances or transaction records.

Banks must therefore verify that AI-generated code preserves the original system’s behavior before putting it into use.

ALSO READ: AI Coding Tools Statistics, Market Size and Growth 2025-2026

Why Banks Still Depend on a Programming Language From 1959

COBOL stands for Common Business-Oriented Language. It was created in 1959 to help businesses and government agencies manage large amounts of data, such as payroll records, financial transactions and administrative information.

Banks adopted COBOL because it handles financial calculations and processes large numbers of transactions reliably. Many banks built their core systems around the language and continue to rely on them today.

More than 60 years later, COBOL remains an important part of the banking industry. According to figures attributed to IBM and cited by CIO Dive in July 2026, the language still supports a significant share of the world’s banking activity.

Source: CIO Dive

These numbers show why replacing COBOL has become a major challenge for banks. Banks cannot shut down systems that handle millions of transactions and replace them with new software overnight. These systems must remain secure, accurate and available throughout the process.

Many COBOL applications have also been updated repeatedly over several decades. Banks have added new rules to meet changing regulations, introduce new products and connect older systems with newer technologies.

Some of these rules are buried deep within the existing code, making them difficult to identify and understand. Modernizing COBOL systems involves more than converting old code into a new programming language. Banks must also understand how their existing systems work and ensure that the replacement software performs the same tasks correctly.

OpenAI Says Its New AI Model Can Translate COBOL With 98% Accuracy

Advances in AI coding tools are creating new opportunities to modernize older software. OpenAI says its GPT-6.1 Sol model achieved 98% behavioral preservation accuracy in a COBOL modernization test when using higher reasoning effort.

The result, reported by The Verge on October 27, suggests that advanced AI models can translate legacy code while preserving how the original software works.

This is particularly important for banks and other financial institutions. Their systems handle payments, account balances, interest calculations and millions of other transactions. Any change in how these systems process information could lead to costly errors.

Behavioral preservation means the translated software produces the same results as the original system under the conditions being tested. For example, an application calculating interest must follow the same rules for interest rates, rounding and account balances after its code is converted.

If the new software changes even one of these calculations, it could create financial discrepancies despite running without technical errors. OpenAI’s reported 98% accuracy is a promising result, but it needs to be understood in context. The figure comes from a specific evaluation. It does not mean AI can modernize an entire banking system with a guaranteed 98% success rate.

The remaining errors also matter. Even a small number of incorrect results could cause serious problems if they affect critical financial transactions.

AI could help banks speed up COBOL modernization and reduce some of the work involved. However, financial institutions must still test the new software carefully to ensure it produces accurate results and works reliably in real-world conditions.

Anthropic Says Claude Can Help Banks Understand Their Old Software

Anthropic says its Claude coding tools can help banks understand old software and simplify COBOL modernization. One of the biggest challenges is identifying how legacy applications work. Many banking systems contain business rules added over several decades. 

Their original developers may have left, and existing documentation may be outdated. Before replacing these systems, engineers must understand how the code works, how data moves between applications and which business rules must remain unchanged.

Claude can help analyze COBOL code, document existing systems, identify connections between applications, generate tests and assist with writing replacement code. This could reduce the time engineers spend examining legacy software.

However, AI-generated explanations and code still require human review. Banks must verify that the tools correctly identify business rules before using them to replace critical financial systems.

Amazon Wants AI to Modernize Entire Banking Applications

Amazon Web Services (AWS) is developing AI tools to help banks replace outdated software. Its AWS Transform service can analyze legacy applications, identify business rules and support code modernization.

On September 29, 2026, AWS announced a partnership with banking technology company Thought Machine to speed up legacy banking system migrations. The companies are combining AWS Transform with Thought Machine’s Vault Forge technology to analyze older mainframe applications and turn existing financial rules into new products for testing.

Thought Machine says this approach could reduce some migration processes from years to days. However, this does not mean an entire banking system can be replaced safely in just a few days.

Banks must still check data accuracy, regulatory requirements, security and connections with other systems before deploying new software.

The modernization process involves three main steps: analyzing existing COBOL code, developing replacement applications and testing the new software against the original system.

AWS also stresses the need for human review before AI-generated code is deployed. Banks must verify that the new applications follow the same business rules and process transactions correctly.

IBM Says Replacing COBOL Is Not the Same as Modernizing a Bank

IBM is also using AI to help businesses modernize COBOL applications. The company has spent decades providing the mainframe systems that banks use to process transactions and manage critical operations.

In 2023, IBM introduced watsonx Code Assistant for Z. The tool helps developers understand legacy applications, convert selected COBOL code into Java and test the results.

However, IBM argues that rewriting COBOL code does not automatically mean banks should replace their mainframes. These systems handle transaction processing, security, data recovery and connections with other business applications.

In February 2026, IBM executive Rob Thomas highlighted the difference between modernizing software and replacing the infrastructure that runs it.

IBM also estimates that around 40% of COBOL workloads run on distributed systems such as Windows and Linux rather than mainframes. This means banks can update their applications without necessarily changing their underlying infrastructure.

IBM has reported improvements from its AI modernization tools. In one customer example, the company said its tools reduced the time needed to identify certain unnecessary COBOL routines by 94%, cutting the task from around eight hours to 30 minutes.

These results apply to a specific customer and do not guarantee similar improvements across all projects.

IBM’s approach shows that banks can use AI to improve legacy software while keeping their existing infrastructure where it still meets their needs.

Why a 2% Error Rate Could Be Dangerous for Banks

AI-assisted COBOL modernization must meet strict accuracy requirements. An incorrect chatbot response may cause inconvenience, but an error in banking software could lead to financial losses, customer disputes and regulatory issues.

Consider a banking application that processes one million transactions. If 2% were processed incorrectly, that would mean 20,000 errors. This is a hypothetical example, not a prediction based on OpenAI’s reported 98% behavioral preservation score.

Banks must test replacement software to ensure it follows the same rules as the original system. Testing should cover unusual transactions, account histories, financial calculations and system failures.

The new software must also work correctly with databases and other connected banking applications. An error in one component could affect several services. Banks can reduce these risks by running old and new systems in parallel and comparing their results before switching over. 

They can also use independent code reviews and formal verification to check specific software properties. AWS has also highlighted the importance of giving AI coding tools enough information about existing code, dependencies and business rules. If a model misses connections between applications, it may produce incorrect results.

For banks, successful COBOL modernization requires more than code that runs. The replacement system must produce accurate results and work reliably across the wider banking infrastructure.

ALSO READ: Bank of America Hires 1,000 Contractors to Fix AI-Detected Security Flaws

AI Could Reduce the Cost of Modernizing Banking Systems

Modernizing legacy banking systems can be expensive. Banks often need large teams of engineers, consultants, security experts and business analysts to understand and replace old applications. These projects can take years to complete.

AI coding tools could reduce some of this work by helping developers identify software dependencies, document business rules, generate tests and translate COBOL code. This would allow engineers to spend more time on complex tasks.

AWS says its modernization tools can shorten some migration projects from years to months. Thought Machine has also claimed that its technology can speed up certain stages of banking modernization.

However, there is no reliable industry-wide estimate of how much AI can reduce the total cost of replacing core banking systems. The savings will depend on each bank’s existing technology, documentation and testing requirements.

Banks must also account for the ongoing costs of new systems. Cloud services can offer flexibility but may introduce additional expenses for computing, data transfers, security and maintenance.

The goal of AI-assisted modernization should be to improve system reliability and reduce long-term costs, not simply replace COBOL code with a newer programming language.

Will AI Replace COBOL Programmers?

AI coding tools are changing how developers work, but COBOL programmers remain important to banking system modernization. Many legacy applications contain business rules that are poorly documented. 

Experienced COBOL developers understand how these systems process transactions and handle financial operations. AI can help analyze and translate the code, but it cannot replace the need for human knowledge and oversight.

In a March 2026 interview with ITPro, AWS executive Asa Kalavade said experienced COBOL engineers would continue to play an important role as AI takes on more modernization work. Developers may increasingly use AI to analyze code, document business rules and generate replacement applications. 

Their work will also involve checking AI-generated results and ensuring new systems operate correctly. AI could reduce the time spent on repetitive coding and analysis tasks, but its overall impact on COBOL jobs remains uncertain.

For banks, one key priority is preserving the knowledge of experienced developers before they retire. AI tools could help document how legacy applications work and make that information easier for future teams to use.

Banks May Not Need to Replace COBOL Completely

Banks may not need to replace every COBOL application to modernize their technology. Many legacy systems still handle financial transactions reliably and can continue to serve their purpose.

Instead, banks can update selected parts of their systems. They might keep existing transaction-processing software while building modern interfaces for mobile banking apps and other services. AI tools can also help document old code, identify unused functions and make software easier to maintain. 

Other applications may need to be rewritten or moved to newer infrastructure. The right approach depends on each bank’s technical needs, security requirements and long-term costs. COBOL modernization does not always mean abandoning the language. It can involve improving existing applications, converting selected code or rebuilding entire systems. 

The main goal is to make banking software easier to maintain, update and connect with modern technology while keeping financial transactions accurate and secure.

AI Could Help Banks Modernize Decades-Old COBOL Systems

Banks have relied on COBOL for decades, but maintaining these older systems can be expensive and time-consuming. Replacing them also carries risks because they handle critical financial transactions.

AI coding tools could help banks address this challenge. Models from OpenAI, Anthropic, IBM and AWS can help developers analyze legacy code, understand business rules and convert applications into newer programming languages.

However, AI cannot replace the need for careful testing and human oversight. Banks must ensure that modernized systems process transactions accurately, protect customer data and meet regulatory requirements.

AI’s biggest contribution may be helping engineers understand and update decades-old software without rebuilding every system from scratch. The goal is to make banking technology easier to maintain while keeping essential financial services reliable and secure.