Customize Amazon Quick embedded chat into your application
Amazon Web Services Chetan Makvana
AWS explains how to customize its embedded Quick chat tool so it blends into your app's look and voice. No more generic chatbot bolted on—now it can match your brand and even answer like your team.
Based on reporting by Amazon Web Services, Chetan Makvana — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Amazon Quick embedded chat lets you drop a conversational AI panel straight into a web app, so users can ask questions and dig into data without ever leaving the page. The catch, as AWS walks through in a new post, is that out of the box the thing looks and sounds like a stranger. Default styling, generic branding, a tone that reads like a textbook definition of revenue rather than something your finance team would say. AWS's fix is a two-part customization approach: visual theming and persona configuration.
On the visual side, the embedded chat lives inside an iframe, so you can't reach in and restyle its guts directly. Instead you style the container wrapping it and lean on SDK frame options — things like a custom className for border-radius and shadows, a withIframePlaceholder flag so users see a loading state instead of a blank box, and footerOptions to strip out the default "Powered by" attribution and usage policy link. AWS's example turns a generic chat widget into a slide-in side panel matching a finance dashboard's purple gradient header, complete with responsive CSS for mobile screens.
The tone problem is separate, and arguably the more interesting one. Ask the default chat agent "What was our revenue last quarter?" and you get a bland explainer about what revenue even is — no numbers, no context, nothing tied to the actual company. AWS solves this by having you build a custom chat agent in the Quick console with persona instructions: a written brief telling the agent who it is, how to structure answers (lead with the number, compare to the prior period, cite sources, use bullet points), and where its boundaries sit (only discuss company financial data, punt anything else to a specific email address). Paired with SDK content options like fixedAgentId to lock out the generic agent selector, and an initialPrompt that fires automatically so the assistant seems to open the conversation itself, the same revenue question now returns actual figures — $8.89M for Q2 2025, up 12.5% versus the prior year, broken down by month and region.
The last piece is a sendPrompt() method that lets your application's UI talk to the chat programmatically. Click a revenue metric card on the dashboard, and the app can fire off a pre-written question like "Explain the revenue trend this quarter and what is driving it" without the user typing anything. AWS's demo wires this into quick-action buttons for revenue, margins, and expenses sitting right in the chat panel header, turning the chat from a passive Q&A box into something that reacts to what you're already clicking on.
My take — AI-written commentary, not fact-checked reporting
This is the unglamorous but necessary work of turning a bolted-on AI widget into something people actually trust enough to use daily — nobody wants a chatbot that sounds like it read a glossary entry instead of their own quarterly numbers. The persona-instruction approach is really just prompt engineering with a nicer UI around it, and that's fine; the real value here is AWS admitting that a generic assistant answering with textbook definitions is basically useless in a real business dashboard. Companies that skip the tone customization step and just slap the chat panel on will end up with users ignoring it within a week.
Read more about this at: Amazon Web Services