He speaks four languages, holds a Master's in Computer Science, wires circuit breakers in his free time, and will stay up until 3 a.m. methodically debugging why your AI agent is escalating too many calls. Meet Eduardo Perotti — RingLogix's new Conversational AI Analyst and, frankly, your new secret weapon.
Here's the short version: if you've been nervous about deploying a FlowbotAI agent because you're not sure how to structure the logic, build the prompts, or handle the edge cases that make or break real customer calls — Eduardo is exactly who you've been waiting for. His job is to sit between your business goals and your AI agent's brain and make sure the two actually speak the same language.
The longer version is even better. Let's get into it.
Eduardo is a Back-end and AI Software Engineer based in Belo Horizonte, Brazil, and a registered Computer Engineer at CREA-MG. Before joining RingLogix, he was at Blip, where he focused on backend architecture and conversational AI — mainly in WhatsApp and META's environment (Instagram, Facebook), but also across different use case scenarios and channels a chatbot could be used, including Health & Care, the Automobilist Industry, small and family companies, construction companies, and large enterprises looking to convert more users and leads to their business.
He's also currently finishing his Master's degree in Computer Science at UNIFEI, specializing in Ontology and Knowledge Graphs — which is essentially the science of teaching machines how to understand concepts and relationships at a philosophical, linguistic, and computing level.
That last part matters more than it sounds. When Eduardo designs a FlowbotAI agent, he's not just writing prompts. He's structuring knowledge — the same way a language evolves, your voice agent evolves with your business and your users.
We asked Eduardo what drew him to conversational AI in the first place. His answer:
"I've always been fascinated by how we can structure knowledge and logic to create seamless human-computer interactions. My academic research focuses heavily on domain ontologies and knowledge graphs, essentially teaching machines how to understand concepts and relationships into a philosophical, linguistic and science computing way. Applying that to conversational AI and multi-agent orchestration was a natural and exciting progression for me."
And why FlowbotAI specifically?
"It perfectly bridges my technical background with real-world business strategy. I love the technical challenge of backend development and logic design, but I also genuinely enjoy translating complex business requirements into intuitive AI workflows, effective prompting and thinking about standardization, business pattern/economic feasibility. This role allows me to sit right at the intersection of both, plus I'm able to know more about other people, other country culture, business goals and logic. I think we evolve by learning together with everyone, there is a lot of things I don't know about or don't do it a daily basis, so being able to interact with everyone's working environment and different scenarios is making me evolve a lot and improve my logic to new paths/outside of my comfort zone."
In plain terms? He describes his role this way:
"I am the bridge between a partner's business goals and the AI's brain. I take what a business wants to achieve on a phone call and map out the logic, rules, and conversational pathways so the AI acts exactly the way their best human employee would."
That's not a traditional "AI builder" role. Eduardo was clear about the distinction:
"A traditional builder might just write code to a spec. This role requires empathy, strategic foresight, and a deep understanding of human conversation. It's about structuring knowledge, much like the ontologies I study, so the AI can dynamically handle unpredictability rather than just following a rigid script. We are way more into the Linguistic/Philosophy area evolving with our partners, as the same a Language evolves, our voice agents evolve within our users."
If you've submitted (or are about to submit) a FlowbotAI use case, here's what Eduardo is actually paying attention to:
"I look for clarity on the primary objective, but more importantly, I look for the boundaries. What shouldn't the AI do? What are the specific business rules, terminology, and standard operating procedures that make their service unique? A unique/single AI won't solve all user's problems (as they are as generic and diverse), but we can generalize to some businesses rules or create specific agentic workflows for unique case scenarios."
The things that slow everything down?
"A lack of clarity around edge cases and escalation triggers, lack of data information in general but primary doesn't have goals and no-goals to achieve. A single AI agent won't resolve all problems, neither will it at the beginning, it must evolve with user's experience, interaction and data, the same way a company/human interaction changes over time, the agent must act as organic and alive as possible, receiving periodically updates. If we don't know how a business traditionally handles a frustrated caller or an out-of-stock item, the AI won't know either. Ambiguity is the enemy of smooth deployment."
His advice for partners before they bring a new use case to the table:
"Map out a real human call. Write down a transcript of a successful interaction and then write down a transcript of a call that went slightly off the rails. Understanding both extremes gives us the perfect blueprint."
Eduardo designs for the real world — not the demo room. On building prompts that can handle actual customer conversations:
"I design for the 'happy path,' but I build the safety nets for the 'unhappy paths.' Real conversations are messy, people interrupt, change their minds, or ask off-topic questions. Prompts need to be resilient, giving the AI instructions on how to gracefully pivot and guide the user back to the main objective."
On why edge cases aren't the exception — they're the whole game:
"Because human conversation is primarily made up of edge cases! People rarely follow a perfectly linear script. An agent's ability to handle an edge case is what separates a frustrating automated menu from an intelligent assistant."
On what "reliable" actually means for a deployed agent:
"A reliable AI Voice Agent should resolve the user/partner problems as the most efficient possible, making the ones who use them happy. A reliable AI should offer solutions even for the 'non-happy pathways', even if it doesn't resolve 100% the problem, it won't let the user unassisted, so the user feels their problem is being acknowledged."
And on what success looks like from the end customer's perspective:
"Frictionless resolution. The customer shouldn't feel like they are fighting a machine. They should feel heard, understood, and efficiently guided to the solution they called for. The final user just wants to solve their problem as much as possible, in the most efficient and less time-consuming way possible."
"Confident and supported. I want them to feel like they have a technical partner who deeply understands their business logic and goals and has expertly translated it into a reliable tool. I'm not a specialist in partners work environment/product, but I want that partners feel we can do a better job than working alone, a win-win relationship."
And his direct message to existing RingLogix partners:
"I'm thrilled to be here. Bring me your toughest use cases and most complex logic problems. I want to know more about you, your experience and know-how, so I can address together the best solution/resolution scenarios. 'Alone we go faster, but together we go further.'"
If you've been curious about FlowbotAI but not sure where to begin, Eduardo keeps it simple:
"Start small. Pick one repetitive, manual/'boring' task that may eat up your time. It can be a simple e-mail/routine check, or some small task, but that consumes your work time. Think of a way to automate this using voice agents, solving problems with a 'quick call/audio input'."
And remember — launch is never the finish line:
"Launch is just version 1.0. I'll be there to analyze real-world performance, debug edge cases, and continuously refine the prompt architecture as their business needs evolve. Launch is the initial step/start point but there are a lot of improvements, new ideas and continuous evolution of the agent as a business/user improves at a weekly/monthly basis."
A few things that might surprise you about Eduardo:
He has AuDHD, and he's upfront about it:
"I have AuDHD, and sometimes I may not understand very well some small talks occasions, but I am always proactive and ready to try to understand, to evolve my communication and know more about everyone. I'll always dig deep into the 'why' behind a project to ensure we are building proper agents that genuinely solve human problems."
His go-to productivity tool is a paper notebook. ("Absolutely nothing beats the satisfaction of physically crossing a completed task off the page.")
He's been playing League of Legends for over 10 years, loves deep anime lore (Evangelion, Fullmetal Alchemist, The Apothecary Diaries), builds PC hardware in his free time, and has a personal mission to try every flavor of energy drink available worldwide.
When you bring Eduardo your toughest FlowbotAI use case, you're getting someone who wires residential circuit breakers for fun and has a Master's thesis on ontologies. He will figure out why your agent is misbehaving — and he'll document it by hand while he's at it.
Ready to work with Eduardo on your next FlowbotAI deployment? Reach out to your RingLogix partner team and let's get started.
"My goal is to help partners use FlowbotAI to turn client/users' conversational friction/gaps into their competitive advantage." — Eduardo Perotti