Amazon’s new generative AI tools—what marketers need to know

Amazon last week upped its presence in the growing field of generative AI providers, rolling out a suite of tools specifically designed for enterprise clients. A new AI chatbot and image generator led the charge. The chatbot, dubbed merely “Q,” is Amazon’s take on a business-focused ChatGPT. The platform is trained to learn how a client operates and thus provide more services pertaining to internal work items. “Titan Image Generator” is more like DALL-E, allowing clients to produce high-quality images from a text prompt, as well as edit preexisting images. Check out the latest Ad Age events and award programs The release of Q and Titan Image Generator, which were announced at Amazon’s annual re:Invent conference, could have a significant impact on the sizable number of marketers that do business with Amazon Web Services. The important takeaway for the ad industry is that these new tools have been developed to handle the intricacies of business workflows, which often differ from one company to the next. Such focus has so far been elusive in the generative AI category, said Rowan Curran, a senior analyst at Forrester. Bots such as ChatGPT, he said, might be highly capable, but they are geared more toward general use, hence their public accessibility. Businesses require far more tailored use cases that hinge on digesting internal databases for more relevant outputs. Q, for example, can learn a brand’s style guide, and use that information to turn a press release into a blog post that adheres to brand standards. The platform can then analyze how a brand has shown up on social media, and create new posts around the release that will make sense to its followers. Q can also analyze the results of these posts and summarize them for review. The price to use Q starts at $20 per user, per year. The platform is also able to grasp a business’ organizational structure and alter accessibility based on who is using it. When being used by a high-ranking executive, Q may surface any information that is requested, but it may limit some of that information when corresponding with a lower-ranking copywriter. How brands are using Q Brands are only just beginning to use Q, but how they’ve developed alongside pre-existing Amazon tools may indicate the direction they take with the new chatbot. LG, for example, created an image-to-text captioning platform that is trained on AWS machine learning technology. The AI receives an image from LG clients, teases out its details and outputs relevant captions. LG declined to provide a comment for this story. Delta and Booking.com have created generative AI bots of their own that hinge on AWS AI technology. Both tools are consumer-facing and are meant to streamline the booking process for travelers. Neither Delta nor Booking.com responded to requests for comment. To be sure, Amazon is not the first provider to roll out business-focused AI tools. Microsoft offers Copilot, OpenAI provides ChatGPT Enterprise and Salesforce has Einstein. But Curran sees Q as a more seamless option for marketers given its position inside the AWS ecosystem. Clients can easily create applications around specific Q capabilities, such as a “blog expert,” and connect them with various data sources, including from non-Amazon services such as Slack and Zendesk. “It’s a step in the direction of lowering the barrier to entry for the building of large apps with LLMs (large language models),” Curran said. Also read: Amazon’s disappointing NFL Black Friday ratings aren’t worrying advertisers Amazon’s AI attempts Amazon has been steadily growing its efforts in the generative AI space. In October, the company rolled out a beta version of an image generator that exists within Amazon Ads—specifically designed to help marketers create campaign assets. In September, it made a $4 billion investment in Anthropic to improve the research lab’s AI models and expand access for AWS customers. These new AI tools are largely an effort by Amazon to incentivize companies to join its data-rich ecosystem, said Chris Comstock, chief growth officer at data standards company Claravine. The rationale, he said, is that companies can generate first-party data using services such as its new clean room, then access generative AI tools to develop strategy from that data. Claravine is a business partner with AWS. Of course, marketers generate first-party data with other tech companies as well, some of whom are also active in the enterprise AI space. The jury is still out on whether the wiser strategy is to strike partnerships with numerous vendors or stick with one. Forrester’s Curran sees the need for numerous partnerships dwindling as vendors build out their offerings, which will likely look similar to those of competitors. However, Amazon depends on a network of different providers. Bedrock, the foundational platform upon which all of its generative AI tools are built, is composed of LLMs from Meta, Anthropic, Stability AI and others. Interoperability is an important requirement for enterprise-focused generative AI, said Jon Williams, global head of agency business development at AWS. Image generation for businesses As with Q, Titan Image Generator is intended to consume a business’ proprietary data to create a more bespoke output. A brand wanting to extend a social media campaign, for instance, could train Titan Image Generator on the style of images previously used, such that the platform will produce more images that fit the campaign. Arguably most notable about Titan Image Generator, however, are its guarantees around security. All images created by the platform create an invisible watermark, which cannot be cropped or pressed out, according to AWS’s Williams. This feature is a result of commitments to safety that Amazon, along with other AI developers, signed at the White House this past summer. Security has thus far been a bottleneck for enterprise clients concerning the adoption of generative AI, said Claravine’s Comstock. Various data leaks, the spreading of misinformation and lawsuits about copyright infringement have emerged with the onset of generative AI, spurring the Biden administration to issue an AI-focused executive order last month. But guarantees made by Amazon, such as watermarking and an indemnification policy that defends clients from claims of third-party IP infringement, could help more companies embrace its AI, Comstock said. Small and medium-sized businesses, in particular, may stand to benefit because they are more likely to abstain in the presence of expensive security risks, he said. The price to use Titan Image Generator is $0.01 per image, not including fees for customization and fine-tuning. Amazon hasn’t covered all of its bases, however. The company did not disclose precisely what kinds of datasets it used to train Titan Image Generator, nor its methods of data collection. It is unclear, for instance, how Amazon received permissions from copyright owners, or how it may be compensating creators. Forrester’s Curran said that this lack of transparency could muddy the ownership status of assets produced by brands using Titan Image Generator. Still, Amazon insists that these images will belong to clients.

Amazon’s new generative AI tools—what marketers need to know

Featured Stories