What Small and Medium Businesses Need to Know About Generative AI in the Cloud

As artificial intelligence (AI) rapidly evolves, it can be hard for small and medium businesses (SMBs) to keep up. Non-technical and technical SMB leaders alike need to navigate the increasing complexity and near-constant change of how businesses can use AI. Currently, most are focused on using resources to focus on security and growth. The buzz around generative AI—and automation in general—can seem overwhelming. It gives way to so many questions and around what opportunities might be available.

To help SMB leaders see beyond the hype, Pierre Semaan, Amazon Web Services SMB Go To Market and Solutions Lead, APJ, sat down with Ben Cabanas, AWS Director of Solutions Architecture, APJ, to define the this new transformative technology and to discuss recent AI developments and what they might mean for SMBs.

Understanding conversational and generative AI

Pierre Semaan: One often hears about “conversational AI” and “generative AI.” How do these forms of AI differ?

Ben Cabanas: Conversational AI and generative AI both involve artificial intelligence that can help automate numerous tasks—but they go about it differently. Conversational AI is focused on natural language processing and understanding, allowing machines to interact with humans naturally, typically with a human voice. Generative AI generates text, images, or other media in response to written directions or prompts.

Conversational AI automates customer experience interactions, provides virtual assistants, and enables natural language search. An example of conversational AI is Amazon Alexa. Alexa can play music, answer questions, and control smart home devices based on human voice. It is powered by natural language processing and machine learning technologies, allowing it to understand and respond to voice commands.

Generative AI systems are a bit more technical. They use generative models such as large language models (LLMs) to sample new data based on the training dataset used to create them. A real example of generative AI is Generative Adversarial Networks (GANs). GANs are a type of neural network architecture used to generate new data based on existing data. Simply put, GANs can help you synthesize images, text, and audio. One of the most recognizable examples of generative AI is Adobe Photoshop’s Generative Fill feature. Practically, if you want to create new photos on your SMB’s website but lack in-house designers, Generative AI can help you build a more compelling web experience.

Harnessing AI to produce text-based content

Semaan: You previously mentioned LLMs. What are LLMs, and how do Generative AI apps use them?

Cabanas: An LLM is a type of AI model that is specifically good at processing, interpreting and generating text- or image-based content. It is a type of AI algorithm that uses deep learning techniques and massively large datasets to understand, summarize, generate, and predict new content. LLMs are trained on huge amounts of data, specifically text data or image data. All of this data, wherever it comes from, is processed through a neural network. These networks continually adjust the way they interpret and make sense of data based on a host of factors, including the results of previous trial and error.

Semaan: How does the machine learning component work?

Cabanas: LLMs don’t really “know” anything, but they are very good at figuring out which word follows another, which starts to look like real thought and creativity when it gets to an advanced enough stage. Most LLMs use a specific neural network architecture called a transformer that can read vast amounts of text, spot patterns in how words and phrases relate to each other, and then make predictions about what words should come next … There is some randomness and variation built into the code. For example, a bot might not always choose the most likely word to come next, but the second- or third-most likely, which could obviously lead to interesting results. This is why LLMs are in a constant state of self-analysis and self-correction. In other words, SMBs need to understand where they are useful and apply to the appropriate business needs.

How to use flexible AI solutions in the cloud

Semaan: Earlier this year, AWS announced the preview for Amazon Bedrock. How does this service make Generative AI technology available to customers?

Cabanas: Amazon Bedrock is a fully managed service that makes machine learning models—which we call foundation models (FMs)—from leading AI startups, and from us at Amazon. Our customers can then choose from a wide range of FMs to find the model that is best suited for their use case without having to manage any infrastructure.

Importantly, Bedrock allows you to train the model with your own private data—and your data stays private to you. Because Bedrock is in the cloud, SMBs with technical teams can get started quickly, privately customize FMs with their own data, and easily integrate and deploy them into applications using the AWS tools and capabilities they are familiar with.

Semaan: How would you describe the advantage of this approach?

Cabanas: While services built on specific models typically run solely on their language model platforms, Bedrock users can perform specific tasks by selecting from a range of FMs. As an example, a content marketing manager could use Bedrock to create a targeted ad campaign for a new line of handbags by feeding it data so it generates social media posts, display ads, and web copy for each product. As an example, Coda AI, an AI document generation firm used by companies like Uber and the New York Times, is using Bedrock to scale its business operation.

Semaan: How might an SMB explore opportunities and get started with AI solutions such as these?

Cabanas: If SMBs have limited in-house tech experience, they should check out our AWS Partner Network consultants who have a critical role to play here. First, they are uniquely positioned as trusted advisors with their customers who use AWS. Their part focuses around education and driving clarity in this area. Answering questions like the ones we are covering in our conversation: what are these new models, why are they different, what outcomes do they help with, how to implement, etc. Then they can integrate their models—presumably models created for specific industries or use cases—into Bedrock. Lastly, consultants can help deliver value, and build or integrate these models into customer applications on their behalf.

Next steps

AI is still in its early days and is bound to change. As our experts shared, this is an evolving space that will require a mindset focused on experimentation, not perfection. SMBs can take advantage of AI as a way to offload predictable tasks so their staff can focus on more strategic efforts.

Want to learn about adding AI to your SMB’s operations? Explore how to get started as well as use cases from customer experience to process automation, and much more in our free eBook.