Gupy is Brazil's largest HRTech. Through organic growth and acquisitions, the company expanded from a recruiting platform into a full ecosystem spanning hiring, onboarding, training, performance, and career development. Today, around 700 employees (they call themselves Guppers) work across the country. And with that growth came an internal challenge: making all of that institutional knowledge accessible to every employee, at the moment they needed it.
Gupy went from 120 to 700 employees in what felt like a blink, largely through acquisitions that brought new people, processes, and institutional knowledge. But knowledge management never kept pace. The company used Google Sites for their internal knowledge base: functional, but rigid and hard to make intuitive.
Gupy never did knowledge management well. We were never that mega-organized company that had everything mapped out. We grew very fast, mainly because of the companies we acquired.
— Duda Lima, Internal Communications & Culture, Gupy
But the real bottleneck wasn't the site. It was Slack. To centralize employee questions, the People team created a channel called "Gente Escuta" ("People Listen"). In practice, nearly 40 people had to keep one eye on that channel at all times. Questions ranged from vacation policies to performance reviews to onboarding processes. Each required a different specialist. No single person could answer everything. And because the channel was public, some Guppers were too embarrassed to ask questions they feared might seem basic.
My Slack was completely overloaded with questions I couldn't even answer. People assumed I held all the knowledge. We wanted to depersonalize this. Guta is the source of knowledge, not Duda.
— Duda Lima
Guta isn't new. She's a character that Gupy's People team created long before any AI was involved: the face of internal communications, with a name, appearance, and personality chosen by the Guppers themselves. Before the project, Guta was a Slack profile that Duda manually operated behind the scenes. It worked at small scale, but it was one person doing the work of what should be a system.
The vision was to make Guta real. An always-on AI assistant that lives in Slack, understands every internal policy and cultural reference, and responds in Guta's established voice. Not a generic chatbot. A virtual team member that Guppers already knew and trusted, now actually capable of answering their questions.
Bringing Guta to life required solving two problems at once. The knowledge base needed to be modernized so it could serve as a reliable source. And the AI assistant needed to be built to pull from that source and deliver answers in Slack. The two are deeply connected: Guta is only as good as the knowledge she can access.
The first step was migrating Gupy's knowledge base from Google Sites to deco CMS. A developer on the deco team built the initial site structure and custom sections tailored to Gupy's needs: pages organized by employee journey stages, SSO login restricted to Gupy emails, and reusable components designed so the People team could create new pages and update content on their own.
On Google Sites we were locked into their format. Here, the sky was the limit. We could lay out knowledge, policies, and processes in a much more elegant, fun, and appealing way. People actually want to consume the content now.
— Duda Lima
After the initial build, the People team took over day-to-day operations. For content updates and smaller adjustments, they use DecoPilot, deco's AI-powered editing agent that lets anyone make changes through natural language. Duda now describes herself as "close to a specialist" and regularly adjusts content and sections on her own without involving the engineering team.
In the beginning, I depended a lot on Daniel. He'd say, 'Talk to DecoPilot.' Now me and the assistant are best friends. I can resolve half my issues without involving the team at all.
— Duda Lima
Guta's agent, prompt, and workflows all live inside deco CMS alongside the knowledge hub. The architecture connects three systems: a vector database stores the hub's content and stays in sync through scheduled deco workflows that detect page changes and update automatically; the company's shared drive provides supplementary documents like policy PDFs and guides; and a custom open-source Slack integration delivers the experience, allowing Guppers to message Guta in private conversations, just like messaging a colleague.
When a Gupper asks a question, the agent searches the knowledge base, retrieves supplementary context if needed, and composes a response with links to source pages and related topics. If someone asks about their performance review, Guta might also surface information about professional development plans, because those topics are naturally connected. Responses stream in real time, so users see the answer forming rather than waiting in silence.
Guta needed to speak the way Guppers speak: approachable, aware of internal vocabulary, comfortable with memes and wordplay. What made this possible is how simple it is for the People team to update both sides of the system. They edit content on the knowledge hub through deco CMS, and they refine Guta's prompt and personality in the same platform. Early responses were too dry. The team iterated on the prompt, tested exhaustively, and gradually shaped Guta into something richer. No engineering tickets required.
Culture is the big key. You can see an amazing project at Amazon and want to replicate it, but Amazon's culture is completely different from ours. Guta works because she internalized our culture. The Guppers chose her name, her face, her tone of voice. They're all parents of Guta.
— Duda Lima
Guta's answers have gotten richer through iterative refinement. The team has learned unexpected skills managing prompts and knowledge retrieval, and the volume of direct messages to the People team has dropped noticeably. Employees are actively engaging with both the bot and the knowledge hub, and some have even created Guta-themed artwork using internal AI tools.
For the People team, the shift is tangible: Guta went from a weekly communications messenger to an always-available knowledge partner. The team that used to spend its time answering repetitive questions now focuses on strategic work.
The plan is to expand Guta's coverage into new knowledge domains and deeper integration with internal systems. The team has already identified other use cases across the employee experience where the same approach can create value. For a company that manages the entire employee lifecycle as its core product, the way they manage their own employees' experience matters. Guta is the first step in treating internal knowledge with the same care Gupy brings to their customers' HR processes.
If it's up to me, we have many more projects together. We saw how much the platform can help us go beyond what we've been doing. I have so many ideas to explore.
— Duda Lima
Whether it's internal knowledge management, customer support, or operational workflows, we're helping companies build AI agents that work in production. Come see real demos and discuss your use case at VTEX Day 2026.
Subscribe to our newsletter and get the latest updates, tips, and exclusive content delivered straight to your inbox.
