Document production time cut from 16 hours to 10 minutes
Accuracy target of 90% exceeded on first proof of concept
50,000 pages of proprietary data now powering AI-generated documents
Angelus used Gemini to automate regulatory documents across 90 global markets, freeing specialists to focus on innovation.
Angelus used Gemini to automate regulatory documents across 90 global markets, freeing specialists to focus on innovation.
Eliminating manual documentation bottlenecks to free specialists
Angelus is a Brazilian medical device company with 34 years of history and a presence in more than 90 countries. Founded by a dentist whose first product was based on a patent, Angelus competes on differentiation, developing bioceramic products and fiberglass post lines that require clinical proof, intellectual property protection, and regulatory compliance across every market where they are sold.
That last requirement was where the bottleneck lived. A single complex regulatory document could take a specialist 16 or more hours to produce manually. Private label clients in markets like France, Germany, and Italy required documents translated into up to 27 languages and formatted to each country's regulatory template. When a document came back rejected for a formatting error, the company paid for a second review in euros or dollars.
"We had a product blocked in inventory for six months because of a naming discrepancy between what was approved by the regulatory agency and what was on the label," says Cesar Bellinati, director of research and development. "That kind of inconsistency has a very significant cost."
The question was never whether to adopt AI. The question was which problem we wanted to solve with it. When we found the right problem and chose Google as our platform, everything else followed from there.
Cesar Bellinati
Director of Research and Development, Angelus
With a 28-person R&D team spending most of their time on documentation rather than development, Bellinati saw the problem clearly. "We looked around and the laboratory was empty," he explains. "Everyone was behind a computer doing process after process.”
When they discovered that controlling the data sources fed to an AI model could eliminate hallucinations, the path forward became clear. Angelus connected with IPNET by Vivo, a Google partner, who helped design the solution. As a company that competes through patents and proprietary data, data isolation was non-negotiable. Google Cloud provided the ability to process confidential regulatory and engineering data in a fully controlled environment.
"We wanted to choose leaders in AI solutions," adds Luiz Gustavo Branco, IT manager. "That gives us the support to grow in a sustainable way and to scale."
Turning 50,000 pages of knowledge into an AI compliance engine
With Google Cloud in place, IPNET by Vivo built Angelus a document intelligence platform powered by Gemini Enterprise. The system connects directly to Angelus's document management system and project task manager, ensuring the AI always accesses the most current version of every document.
Having a partner that gives us this security is fundamental. But more than that, Google is a company with decades of history. That tangibility of the solution, that confidence in the platform, makes all the difference for us.
Cesar Bellinati
Director of Research and Development, Angelus
When a researcher receives a task, the agent retrieves the relevant technical and regulatory data from across Angelus's 50,000-page document repository, interprets the requirements, and generates the content. That connection to live documents eliminates one of the most common sources of non-conformity in regulatory audits: a researcher unknowingly working from an outdated version.
What makes the system work in highly regulated environments is a deliberate separation between content generation and document formatting. "For us, content is sine qua non," says Bellinati. "We need absolute certainty that everything placed in the document comes from our own proprietary data universe and that nothing is being invented."
To prevent formatting errors, the solution writes directly into Angelus's official Golden Sample templates, the exact files that regulatory agencies expect to receive. The AI fills the right content into the right placeholders without altering the structure. Angelus selected 39 document types for automation across the R&D cycle.
The proof of concept validated everything. The team deliberately chose their most complex document type, a full gap analysis comparing project outputs against regulatory requirements across multiple sources simultaneously, without telling Bellinati which one they had picked. The target accuracy was 90%.
"We realized the accuracy exceeded 100%. because the model identified errors in documents that had been produced manually,” notes Branco. “Those were documents we thought were correct."
That result convinced the board and led directly to a Gemini Enterprise contract for the entire company, expanding the solution beyond the R&D team to every department at Angelus.
Taking the laboratory from empty to the center of innovation again
The numbers tell the first part of the story. A document that once took 16 or more hours of specialist time to produce manually now takes 10 minutes. The accuracy of the AI-generated output exceeded the 90% target and surpassed that of the manually produced documents it was tested against, because the system identified errors humans had missed. For a company that estimated more than 10% of its revenue had been lost over the previous decade to documentation inconsistencies, those results carried immediate financial weight.
But for Bellinati, the most significant outcome was cultural. "The biggest gain was changing the mindset," he reflects. "Without board sponsorship, we would have stayed with good intentions but gone nowhere. Now other directors are asking what we are doing and wanting to bring AI into their own areas."
From the beginning, Angelus was clear with the R&D team about what this project meant. "We never said we are going to reduce the team," Bellinati emphasizes. "We are going to give you augmented intelligence. We are going to do more."
The next step is building a fleet of specialized agents, each trained in a specific area of Angelus's expertise, covering regulatory affairs, market authorizations, intellectual property, and quality management systems. Because Angelus spent the last decade building methodological knowledge in structured formats, templates, procedures, and specialist expertise, training those agents on proprietary data is now both feasible and strategic.
Our intention is to transform AI from a tool into a platform that supports innovation. Innovation for us means putting differentiation in the hands of our clients. That is what this partnership with Google makes possible.
Cesar Bellinati
Director of Research and Development, Angelus
For other companies in highly regulated industries considering a similar move, their advice is the same: start with the problem, not the technology. Choose a process that is repetitive, measurable, and has historical data behind it. "And make a POC first," Bellinati notes. "Because AI is new not just for companies. It is new for AI developers too."
Angelus is a Brazilian medical device company specializing in endodontics, with 34 years of history and products sold in more than 90 countries across bioceramic and fiberglass post lines.
Industry: Healthcare and Life Sciences
Location: Brazil
Products: Gemini Enterprise, Google Cloud
About Google Cloud partner – IPNET
IPNET by Vivo is a Google Cloud partner with over 20 years of experience, named Latin America Partner of the Year multiple times, helping organizations migrate and innovate on Google Cloud.
