What to expect from Amazon's massive investment in AI

David Paul Morris

Amazon's foray into financial services was already turning up the heat on banks and payment companies, an encroachment that stands to get a major jolt as the e-commerce company pours a small fortune into new artificial intelligence technology.  

Amazon has agreed to invest $1.25 billion in generative AI startup Anthropic and take a minority stake in the company, with an option to invest as much as $4 billion. Generative AI has myriad potential uses, including updating and developing new payment products and financial services in a manner that potentially puts traditional bank technology companies on the defensive. 

"Amazon Web Services hosts not just fintechs but merchants. It is now equipping them to customize their commerce and payment functionality using generative AI in competition against legacy core processors," said Richard Crone, a payments consultant.  

Amazon is battling large technology companies such as Google, Microsoft and Meta to gain users of generative AI, or a more advanced version of AI that can produce original content and fuel business practices such as marketing, customer service, security, product development and other yet-to-be-determined tasks. 

Most banks and payment companies are considering how improved large language models and other emerging generative AI technology fits into their businesses.  

Even given concerns about the tech's accuracy, newer AI tools such as large language models can improve the personalization, flexibility and user experience for the value-added functions that are stacked on top of payments and other financial financial services. That potentially puts large technology companies like Amazon and their considerable generative AI investments in a favorable position to negotiate with banks and other payment companies.

"With generative AI, the more you see the smarter you get — and the smarter you get, the more you see," Crone said.  

What Anthropic offers for Amazon

Anthropic's products include a chatbot called Claude 2, which the firm is reportedly expanding. Claude 2 is similar to large language model programs such as OpenAI's ChatGPT, which has gained attention over the past year for its ability to produce original informed responses based on prior communication and other activity. Google also offers a large language model program.

Amazon, which is in the midst of a major push into generative AI, is not disclosing how or where it will use Anthropic's technology as part of this week's announcement, the e-commerce company's public relations office said in an email (Amazon's PR team also referred questions about the Federal Trade Commission's anti-monopoly lawsuit against the e-commerce firm to an Amazon statement countering the FTC's claims). 

Artificial intelligence

AWS, Google and Microsoft are in an AI arms race. Banks are watching.

July 30, 2023 9:00 PM

"With the rise of generative AI, every Amazon team is using or building AI applications that will reinvent and enhance their customers' experience," Amazon's PR team said. 

Amazon uses AI in its AWS division, which includes a development hub for generative AI, and in other parts of its business. That includes recommendation engines for shopping on Amazon.com, AI-powered robots that operate in its warehouses, the large language models that support Alexa, and algorithms that drive personalized content for Fire TV and Prime Video.  Overall, more than 100,000 clients are using Amazon products that incorporate AI in some manner.

In financial services, lending institutions are expected to be able to fuel economic development by accelerating loan approvals for underserved communities, using generative AI to perform analysis, paperwork and communications, Amazon said. 

Anthropic, in the short term, will use AWS as its main cloud provider, giving AWS users access to new features. AWS in June launched AWS Payment Cryptography, which enables cloud-hosted payment development. Payment Cryptography is in competition with similar systems from IBM and Microsoft that attempt to upgrade payment technology remotely, rather than installing new physical hardware. That is an alternative to traditional payment hardware developers, which have invested heavily in recent years to automate legacy systems.

This year, Amazon also expanded its Buy with Prime feature, enabling consumers who subscribe to Amazon Prime to pay on other retailers' sites, and receive Amazon's two-day delivery. BigCommerce is among Buy with Prime's users, bringing thousands of merchants under the Buy with Prime umbrella, and thus in close proximity to Amazon's other payment and financial services. 

Amazon has entered the merchant cash advance market through a partnership with fintech Parafin and has also moved into buy now/pay later lending. 

And Amazon Go, one of the company's major years-long initiatives to augment or even replace point of sale terminals, uses a mix of AI, deep learning and sensor fusion. Sensor fusion merges data from multiple sensors into a single source for analysis, and deep learning "teaches" computers to recognize patterns in images, text, sounds and other data. 

Amazon Go and other checkout-free retail technologies are designed to enable consumers to pay without a dedicated point of sale; the sensors record purchases that are charged to the users' e-commerce account. 

But such systems are also designed to record and accumulate data such as shopping patterns. The tech also collects data on how long shoppers look at an item, or if they read packaging, or how much time they spend in different parts of the store. Fed into a generative AI program, that can inform marketing, incentives and future product recommendations. 

"To the extent that financial institutions, fintechs and payments companies are building on AWS, this gives them additional options to experiment with the technology and innovate around product delivery," said Aaron McPherson, principal at AFM Consulting, who sees an impact from AWS behind the scenes at payment companies.

Large language models could be useful for risk and fraud management by enabling larger and more diverse data sets to be processed when generating a fraud or risk score, McPherson said. 

"Amazon could incorporate the risk and fraud management technologies into Amazon Pay and the payments it processes on its own behalf," McPherson said. 

Generative AI pulls in the cash

Generative AI has seen major investments in 2023. More than $14 billion has been invested in generative AI companies as of the second quarter of 2023, according to CB Insights, a figure that dwarfs the $2.4 billion for all of 2022 and $4.7 billion from 2019 to 2021 combined. Open AI received $10 billion of the 2023 figure, and Anthropic had raised $880 million before receiving Amazon's investment. 

Some estimates place the investment in generative AI higher. Crone Consulting estimates that Google, Microsoft and Amazon have invested a combined total of over $23.3 billion in GenAI.

Since Amazon has about 310 million active users, there are a lot of different ways that generative AI could be useful to the company, such as adding AI to customer service and enhancing search capabilities and customer queries, according to Thad Peterson, a strategic advisor for Datos Insights. 

"Payments can clearly be a component of that evolution, if added data and intelligence indicate that a customer may be ready to make a purchase, GenAI could preemptively offer to close the deal, offer additional, relevant add-ons or options, or present payment options that are particularly relevant to that individual customer," Peterson said, adding that Amazon is "really good" at optimizing its platform, and with generative AI, payments will be a component of that.