Amazon Q is a Generative AI Copilot for Business Untethered from Enterprise Apps
Amazon Q is a Generative AI Copilot for Business Untethered from Enterprise Apps
It is also a sign that the battlefront may be shifting to the application layer
Amazon today announced Q, which is Amazon’s answer to Microsoft Copilot, Google Duet AI, and ChatGPT Enterprise. Not to be confused with the leak around OpenAI’s Q* program or even Q-learning, Amazon’s Q is designed to be a helpful assistant for business users. Amazon referred to Q as “Your business expert,” but there are also technical featues in Q Builder. Four key solutions are baked into Q today and one is listed as coming soon. Amazon positioned these as:
AWS Expert - Help developers and builders
Your Business Expert - Help people across my company with their work
Amazon QuicksSight Assistant - Help people with business intelligence
Amazon Connect Assistant - Help customer service agents provide better support
AWS Supply Chain Assistant - Help inventory managers and supply planners (coming soon)
Questions and Answers
Q is an assistant that can answer questions and execute tasks in each of the domains listed above. Users can upload their business knowledge into a retrieval augmented generation (RAG) vector database, and connect Q to internal systems of record or third-party applications and services.
Among third-party services, Amazon said it offers 40+ built-in connectors to solutions such as S3, Salesforce, Google Drive, Microsoft 365, ServiceNow, Gmail, Slack, Atlassian, and Zendesk. Information is returned to users with references and citations and administrators can add guardrails to proactively moderate system inputs and outputs. Q also offers permission and role-based access control.
Amazon listed Alnylam, Deloitte, Gilead, Repay, Virgin Pulse, and Wunderkind as launch customers. It wasn’t clear precisely which assistant solutions each company is using. However, Wunderkind’s Richard Jones said, “We expect the time spent on content discovery alone to be reduced by over 30%…[and] anticipate the ability to accelerate the content creation process by nearly 50%” for its customer success and marketing teams. Repay has connected Q to its Chatbot UI.
AWS Management
Q was trained on “17 years’ worth of AWS knowledge and experience building in the cloud. Amazon Q is an expert in AWS services, best practices, well-architected patterns, and solutions to help you get started faster, learn unfamiliar technologies, build new solutions, and spend less time on undifferentiated work like maintenance.”
This is primarily a knowledge navigation tool for cloud administrators, but it also can connect to Amazon CodeWhisperer through an IDE to understand code, build features, and generate tests.
Knowledge Assistant
Question answering is the top business use case for large language models (LLM) today. According to Azure OpenAI Service, knowledge assistants represent more than 80% of its LLM projects. AWS enables users to upload data or connect their knowledge assistant to forty third-party applications and databases, including Salesforce, ServiceNow, Slack, Google Drive, GitHub, Jira, and several Microsoft applications. For knowledge search, the solution uses a RAG but may also be able to access data without vector database embeddings when connected to other applications.
Amazon does not say which LLM is behind Q. The PartyRock service allows users to select Anthropic’s Claude or AI21’s Jurassic. It is likely that Amazon’s Titan LLM is also being used in the background or will, at some point, displace third-party models as a cost-savings measure.
By the way, creating a customized Q bot looks about as easy as creating a GPT. d
Business Intelligence
Like a Copilot in the Microsoft application stack, Q is also the assistant application layer that works across numerous AWS services. From a business intelligence standpoint, Q is integrated within Amazon Quicksight. “Analysts can quickly build visuals and calculations and refine visuals using natural language…Users no longer have to wait for BI teams to update the data and dashboards for every new question.”
Customer Service
The biggest impact, however, may be for Amazon Connect users. Agent assist technologies can offer significant time savings, improved customer satisfaction, and speed up customer service agent time to productivity. According to the announcement:
Amazon Q in Connect…uses generative AI to deliver agents suggested responses and actions to address customer questions, providing faster issue resolution and improved customer satisfaction.
Knowledge articles, wikis, and FAQs can be spread across separate repositories, and agents waste time trying to navigate all of these different sources of information. In the meantime, the customer waits for an answer. Amazon Q in Connect leverages the real-time conversation with the customer, along with relevant company content to automatically recommend what to say or what actions an agent should take to better assist customers.
While automated call summaries will be the first metric to show benefits, real-time assistance may ultimately be the most impactful. Amazon made a related announcement yesterday about an upgrade to its Transcribe product.
Amazon Transcribe’s next generation, multi-billion parameter speech foundation model-powered system expands automatic speech recognition (ASR) to over 100 languages. All existing and new customers using Amazon Transcribe in batch mode can realize the accuracy improvements for 100+ languages without needing any change to either the API endpoint or input parameters.
Like OpenAI’s Whisper, Amazon Transcribe is now augmented with an LLM. This is a key enabler for Q Connect as it can be used to monitor contact center agent calls in multiple languages and make real-time suggestions about how best to serve the customer.
Pricing
On the surface, Q Business for the knowledge assistant looks affordable at $20 per month per user. Adding technical solutions in Amazon Q Builder, such as the AWS expert, code explanation, and debugging, raises that price per user to $25 monthly. However, the terms say there is a minimum of 10 users, so that will be a monthly bill of at least $200 to $250. In addition, there is a minimum charge of $100 per month for using the RAG in Q Business Expert. This is less expensive than the generative AI packages for ChatGPT business ($60 per user per month), Microsoft 365, and Google Workspace ($30).
Is there a Constituency for Q?
Microsoft currently has a very strong strategy for generative AI in the application layers. Heavy users of Microsoft 365 and other services such as Teams and Sharepoint get a copilot assistant overlayed on their existing applications. Companies that are not as committed to the Microsoft productivity application stack can access ChatGPT Enterprise through Azure. Google Workspace users have a similar option to Microsoft 365 users, though Google’s solution now consists of a mixture of Duet AI and Bard.
Q looks a lot more like ChatGPT Enterprise. Yes, Q is a feature overlayed on Quicksights and Connect, but most of those users are also customers of Microsoft 365 or Google Workspace. Amazon is likely counting on net new customers for Q.
That leads to the question of whether there is a natural customer constituency for Q, similar to Microsoft and Google. The answer is probably no. Users of CodeWhisperer and CodeCatalyst will surely benefit from the chat assistant, which is similar to the GitHub Copilot assistant. This doesn’t mean Amazon won’t capture a large user base, but it will be harder to develop. The advantage Amazon has over traditional application providers is that it is well-positioned for a customized solution built within AWS, a trusted partner for many organizations.
The Cloud Wars Shift to Gen AI Applications
Q’s introduction also signals the rising shift from the foundation model battles within the cloud computing wars to the application layer. This is not to say the foundation model competition has fully played out. It is still very early. However, you can think of AI foundation models, such as LLMs, as generative AI infrastructure. Now that many companies are looking to take capabilities into production, applications are becoming an area of focus and point of leverage on top of the AI models.
I am skeptical that Q will be a big hit. Its market share will surely trail that of Microsoft and Google and probably OpenAI as well. But, generative AI is the catalyst for the launch of millions of assistants, and Q is likely to capture a share of that demand.
Many thanks to our sponsor Dabble Lab. Learn more about bespoke AI and application development here.
