AWS might have made its foundation model service platform, Bedrock, generally available a few weeks ago, but one customer has had access to it for months.
Customer experience platform provider Alida works with large retail, media and financial brands. By examining quantitative and qualitative data, it helps the brands understand consumers' perspectives about their products.
"That sort of qualitative data is where Bedrock has really been amazing for us," said Dan Charles, senior vice president of product and go-to-market at Alida.
While AWS introduced Bedrock in April, Alida gained access as an early adopter of the service in June. Within a week of starting to use the service, the CX software company found it helpful in meeting the needs of the brands it works with, Charles said.
Although Amazon Bedrock provides access to a range of foundation models from generative AI vendors such as Cohere, AI21 Labs, Stability AI, Anthropic, Meta and Amazon through a single API, Alida found that Anthropic's Claude model was the best fit for it. Alida also considered Microsoft partner OpenAI and its popular GPT generative AI system.
"From a summarization capability, we found that the Anthropic model -- compared with OpenAI, which we had previously experimented with, and some of the Llama and Cohere technology -- provided the best summarization," Charles said. "It has some of the best understanding of language [compared with] those models."
Dan CharlesSenior vice president of product and go-to-market, Alida
For example, when examining qualitative data from a survey provided by a brand such as J.Crew, Claude was better at delivering key topics, sentiment analysis for what consumers think about a particular piece of clothing and other data to help make the product better for the consumer, Charles added.
Data privacy and compliance were also key reasons why Bedrock worked better for Alida compared with a vendor such as OpenAI, he said.
"One of the fundamental premises of Bedrock for our customers is that their data is their data and you don't share it with anyone," Charles said. "Bedrock running within AWS provided us that certainty that we weren't going to be able to share our customers' data with any third parties."
Moreover, since Bedrock launched in both North America and the EU and complies with Europe's GDPR, Alida can run the service in the European regions that the brands they work with want them to.
While Alida continues to gain value from Amazon Bedrock, it has also worked with the tech giant to get faster results than it previously could with other vendors such as OpenAI, Charles added.
Bedrock and new generative AI capabilities
Besides releasing Amazon Bedrock on Sept. 28, AWS revealed that its family of foundation models, Amazon Titan Embeddings, is now generally available. The LLMs convert text into numerical representations for personalization, search and retrieval-augmented generation applications.
Also, Amazon Bedrock will soon be the first fully managed generative AI service to offer Meta's open source Llama 2 large language model through a managed API.
On the generative AI front, Amazon also introduced a new CodeWhisperer capability that uses a customer's internal codebase and resources to provide customized requirements.
Amazon QuickSight -- a business intelligence service offering analytics and natural language querying capabilities -- now has new generative capabilities that can help analysts create customizable visuals, complete complex calculations and answer follow-up questions.
Although Alida has not used any of these new capabilities, the company recognizes the power of generative AI for its business operations, Charles said.
"Generative AI, in general, is changing our approach to a lot of aspects of our business," he said, adding that the technology could also help sales and marketing.
AWS has responded quickly to the generative AI boom, said Larry Carvalho, an analyst at Robust Cloud.
"AWS is firing on all cylinders to bring GenAI value to enterprises," Carvalho said. "The pace of announcements has surpassed anything I have seen recently."
As for the new capabilities in CodeWhisperer and QuickSight, the customization capability is unique, but organizations other than those with a lot of custom code might not need it, Carvalho added. And it could be challenging for companies to connect their internal code repositories to CodeWhisperer.
Nevertheless, the new capability will still be helpful for developers, he said.
"It will help make new developer hires productive much faster by reducing the time for new developer onboarding," Carvalho said.
Esther Ajao is a TechTarget Editorial news writer covering artificial intelligence software and systems.