About the client The client is the first fintech specialized in import solutions for businesses in Brazil and Latin America that offers financial solutions in credit, financing, and currency exchange, along with technology solutions that simplify, streamline, and unify services on a single platform. The company stands out through three fundamental pillars: Technology : A team of programmers specialized in innovative solutions. Financial Expertise : Financial knowledge to identify the best products and maximize results. Import Focus : Founded in international trade with a long history in one of Brazil’s leading trading companies. The challenge The company operates through Sales Development Representatives (SDRs) who create business opportunities by contacting clients through various channels, including direct phone calls with leads. The business challenge was to extract quality metrics from these calls and ensure optimal use of the sales pitch , improving SDRs’ operational quality and increasing business opportunity conversion. Previously, all call analysis to identify service improvement points , pitch adjustments, and service quality assessments were conducted manually by the manager, who would listen to each call’s audio and perform evaluations. To help generate automated reports for analyzing these calls, e-Core offered support with a custom artificial intelligence solution using AWS resources . The solution The solution begins by using the SDRs’ phone call recordings with leads. The first step was to create a processing pipeline to convert the recordings into text. With the transcribed audio, we used Generative AI to evaluate the call. We developed a prompt to assess the dialogue between the SDR and a lead, analyzing aspects such as pitch adherence, communication skills, and presentation of the company’s product and services. The final analysis result provides constructive feedback focused on areas for improvement, development, and motivation for the SDR. We used a Generative AI model to automatically transform call transcriptions into structured data with Amazon Bedrock, extracting and organizing essential fields, and generating constructive feedback to improve SDR performance. The resulting file is stored in an AWS S3 bucket and sent to the manager’s area on the company platform via an AWS SQS queue. Below is the architecture of the custom solution developed by e-Core.
Artificial Intelligence (AI) is a transformative force in today’s digital landscape, yet terms like Machine Learning (ML), Deep Learning (DL), and Generative AI (GenAI) are often used interchangeably, leading to confusion. This article aims to demystify these concepts, highlighting their differences and significance in the current industry. Investments in AI are becoming increasingly substantial ( “Stanford AI Index report” published at World Economic Forum , KPMG , Unite.AI , All About AI ). However, many people still lack a clear understanding of the differences between AI, ML, DL, and GenAI. This gap often leads to unrealistic expectations and misdirected investments. So, how do these technologies differ, and how can we apply them effectively? Our goal is to provide a clear understanding of these concepts, presenting the perspective of authors Ian Goodfellow, Yoshua Bengio, Geoffrey Hinton, and Andrew Ng. They have extensively addressed these concepts and their applicability in solving specific problems, which will be discussed throughout this article. What is Artificial Intelligence (AI)? AI is a broad field within computer science that aims to create systems capable of performing tasks that typically require human intelligence. It encompasses various technologies, from rule-based systems to advanced algorithms that can learn and adapt. Implementing AI can be complex and expensive, requiring high-quality data and sophisticated algorithms. AI is classified into two categories: Narrow AI, designed to perform specific tasks, and General AI, which can perform any intellectual task a human can. Within AI, Machine Learning (ML), Deep Learning (DL), and Generative AI (GenAI) are specialized subsets, each designed for different objectives and varying in complexity. These technologies, while interconnected, serve distinct roles in advancing AI’s capabilities.
The functionalities of generative AI have gained popularity with ChatGPT from OpenAI, sparking a series of concerns and projections for the coming years. One of the most critical concerns for an efficient AI strategy is the quality of data used to train these models . Data does not appear by chance, so ensuring access to reliable sources is essential to harness the full potential of this technology. To understand the importance of this point, we can examine the evolution of our search for information, from paper to digital. In the book “Talk to Me,” which explores the evolution of voice computing, author James Vlahos extensively discusses the development of search mechanisms. Decades ago, we sifted through hundreds of encyclopedia entries for information. With the advent of the internet, we began reviewing dozens of content pieces, a process further streamlined by the emergence of search engines. With the advancement of smartphones, we now often see only the top results on a Google search. The emergence of voice assistants a few years ago and the now-amplified potential of GenAI bring us to “position zero” in search results: we ask for information, and it is delivered to us without much knowledge of the source’s reliability or whether there was any breach of intellectual property in generating the requested content. Moreover, open solutions can be utilized by anyone. There are excellent use cases, such as assistants for code development and brainstorming ideas, but limitations still exist in terms of organizational differentiation. Hence, companies are building personalized GenAI solutions using their own databases. This autonomy ensures quality and, most importantly, creates differentiation. As Swami Sivasubramanian, Vice President of Database, Analytics, and Machine Learning at AWS, said: “Your data is the differentiator and the key ingredient in creating remarkable products, exceptional customer experiences, or enhanced business operations.” Indeed, a considerable number of companies have GenAI on their agendas due to the trend. However, many lack a robust and well-prepared data strategy to support their initiatives. Unveiling the path to AI maturity through data The Gartner AI Maturity Model comprises 5 levels, as illustrated in the following image: