Go behind the scenes of the partnership that transformed complex document processing, eliminating operational errors and ensuring scalability with the Multi-LLM architecture.
The Critical Challenge of Foreign Trade
Sertrading, one of the largest foreign trade companies in Brazil and recently acquired by BTG Pactual, handles a massive volume of business-critical documents on a daily basis: commercial invoices, bank slips, bills of lading, invoices and documents from international suppliers.
The problem was not just the sheer volume of documents, but mainly the high variation in layout, multiple languages and heterogeneous formats that these files present. This scenario required time-consuming manual analysis, which was subject to human error and had a direct impact on operational efficiency, internal team expertise and the ability to scale critical processes.
To eliminate rework, reduce operational friction and speed up response times in business-sensitive processes, Sertrading decided to invest in the MPL Doc.IA intelligent agent, a document automation solution based on generative AI and Multi-LLM architecture.
Why did Sertrading choose MPL?
Before making the final decision, Sertrading conducted a thorough evaluation of the different automation and computer vision platforms available on the market.
Alberto Basile, CIO of Sertrading, explains the criteria that led to the choice of MPL:
“Before hiring the MPL Doc.IA agent, did you evaluate other solutions on the market? What criteria were decisive in choosing MPL?”
Alberto Basile, CIO of Sertrading:
“Yes. We evaluated other automation and computer vision platforms, but MPL stood out for three main reasons:
- Real accuracy in complex scenarios, especially foreign trade documents, which often have irregular layouts. Other solutions tested achieved around 70% to 85% accuracy, while in this model we averaged over 95% and, in many cases, 100% accuracy.
- Multi-LLM architecture, which gives us the flexibility to choose the ideal model for each type of document, balancing cost and performance intelligently.
- Speed of implementation and the ease of configuring new documents without relying on the technical team or long, complex projects.
In addition, we noticed the maturity and closeness of the MPL team throughout the journey, which was strategic in ensuring a project with low risk and quick results.”
These three pillars: superior accuracy, technological flexibility and agile deployment, were key to Sertrading's trust in MPL as a strategic partner for digital transformation.
The Project's High Accuracy Result Exceeded Expectations
The success of the project is testimony to MPL's technical excellence and the maturity of the Doc.IA solution. The agent is an advanced platform capable of performing automatic document classification, OCR (optical character recognition) and contextualized data extraction with high accuracy, even on complex, multi-lingual documents with great variation in layout.
Project Phase 1 Performance
The accuracy achieved in the first phase of the project impressed the Sertrading team and proved that the solution stood out from the alternatives evaluated:
| Document Type | Accuracy |
| Boletos |
100% |
| Bill of Lading (CE) |
99% |
| Bill of Lading Item (CE Item) |
99% |
| Invoices |
98% |
These results made it possible to eliminate a large part of the operational errors that previously required manual rework, additional conferences and delays in critical processes, freeing up the team for analytical, strategic and higher value-added activities.
Innovation with Predictability: The Multi-LLM Architecture
The technological differential implemented by MPL was the Multi-LLM (Multi-Large Language Model) architecture, which uses generative AI models provided by Oracle Cloud Infrastructure (OCI) Generative AI.
This approach allows Sertrading to automatically select the language model best suited to the complexity of each type of document, optimizing accuracy and reducing operating costs by avoiding the use of more robust (and more expensive) models in scenarios where lighter models are sufficient.
The Practical Benefits of Multi-LLM Architecture
Alberto Basile highlights the concrete benefits of this approach:
“The Multi-LLM approach completely changes the game. We're not stuck with a single model that's more expensive or slower. For simple documents, we can use lighter models at a reduced cost. For the most critical documents - such as CE Mercante and Commercial Invoice - we can opt for more robust models.
This has brought us two clear benefits:
- Direct reduction in operating costs, because we don't pay for ‘excess computing power’ where it's not needed;
- Increased performance, as each type of document is handled by the most appropriate template.
For us, this guarantees financial predictability and scalability of document automation.”
This architectural flexibility is essential for companies that want to scale automation with generative AI in a sustainable way, controlling costs without sacrificing quality.
The MPL Experience: Technical and Transparent Partnership
When asked about his experience with the MPL team throughout the pre-sales, implementation and post-go-live support phases, Alberto Basile emphasized the quality of the partnership at all stages of the project:
Alberto Basile, CIO of Sertrading:
“The experience was excellent at every stage. The MPL team is extremely technical, approachable and transparent. During the pre-sale, it was already clear that they understood our documents and our operation deeply.
In implementation, the highlight was the speed and quality of delivery, as well as the ability to adapt the agent quickly to the most critical formats.v
Post-go-live, we had continuous support, with close monitoring of indicators and fine-tuning of performance.
This type of partnership makes a difference because it reduces risk and speeds up results.”
This testimony reinforces the importance of choosing a partner who not only masters the technology, but also has a deep understanding of the business processes and acts in a consultative manner throughout the project cycle.
Strategic Results and Scalability
By eliminating operational errors and automating repetitive reading and typing tasks, the Sertrading team was able to redirect effort to analytical activities, strategic decisions and continuous process improvement, instead of spending time on low value-added operational tasks.
In addition to the immediate gains in productivity and quality, Doc.IA's potential for scale is huge. Sertrading is already identifying concrete opportunities to expand the solution to other areas and processes, such as:
- Billing
- Supplier registration
- Validation of logistics documents
- Compliance and auditing
- Financial processes and accounts payable
- Customer service
Scalability, according to the Sertrading team, lies not only in the platform's technological capacity, but in MPL's ability to transform various document flows into faster, more precise processes that are fully integrated into the company's digital ecosystem.
Next Steps: Take This Transformation to Your Operation
Companies like Sertrading are already making real, measurable gains with MPL's intelligent document automation agent. If your operation faces similar challenges - such as high document volume, layout variation, multiple languages, error-prone manual processes or difficulty scaling critical operations - you can replicate this success.
Contact MPL and schedule a demonstration of Doc.IA to assess the potential impact on your operational reality and understand how Multi-LLM architecture can bring superior accuracy with cost predictability.
About MPL
MPL specializes in digital transformation with a focus on Oracle JD Edwards and emerging technologies such as generative AI, low-code, intelligent automation and cloud computing (Oracle Cloud Infrastructure). With more than 35 years in the market and global operations via the Redfaire International consortium, MPL is positioned as a strategic partner for companies seeking innovation with security, predictability and measurable results.
1TP7IntelligentAutomation #IAGenerative #TigitalTransformation #CommerceExternal #MPL #DocIA #OracleCloud #MultiLLM #OCR #PedocumentProcessing
Leave a Reply