The Center for Data Innovation spoke with Birago Jones, co-founder and chief executive officer of Pienso, a Virginia-based company that offers a platform for organizations to create and fine-tune AI models without technical expertise or coding skills. Jones spoke about how a student project at MIT inspired him to help start Pienso and his vision for a future that empowers anyone—not just machine learning engineers—to create AI models to enhance work and deliver data-driven insights.
Martin Makaryan: What is the Pienso platform and how does it work?
Birago Jones: Pienso is an AI platform that empowers subject matter experts in various enterprises, such as business analysts, to create and fine-tune AI models without coding skills. Pienso’s platform is essentially a user-friendly, no-code interface. The central hub of the platform, our dashboard, includes a series of web apps that can process user data, create a model, and test that model. Customers can install Pienso in their preferred environment—on their servers or in the cloud—and upload their raw data, turn them into structured datasets, and annotate or reorganize them. Once the data is ready, they can either create their small models using the tools we provide or fine-tune open-source LLMs that we have available using their datasets. Customers can test as many models as they would like before deciding which one to deploy live: the platform has tools, such as data analysis and visualizations, to guide customers through the process, but ultimately, they must make the call when it comes to what works best for them.
Today, our platform is helping people build LLMs to detect misinformation, identify human trafficking cases, track weapons sales, and more without writing any code. Most of our customers consist of hybrid teams: subject matter experts who leverage a nuanced understanding of their data to use our platform, and data scientists who assist with connecting Pienso to an enterprise’s internal data sources. Users can then perform various tasks once the models are ready, such as summarizing text, categorizing it, and generating insights from the data. Customers can also import these insights to visualization tools like Tableau or Google Data Studio or send them back to the original data sources like Salesforce. We also recently launched PromptFactory, which allows customers to test English-language prompts before deploying them live.
Makaryan: What inspired you to co-found Pienso?
Jones: Pienso’s story dates to my time studying at MIT. My partner, who years later helped me co-found Pienso, and I were working on a group project to combat cyberbullying of teenagers. We decided to create a machine learning algorithm that could automatically detect harmful social media content. We discovered that existing natural language processing models, which at that time used vocabulary from formal sources like mainstream newspapers, did not reflect the teenager vernacular. This experience taught us that creating models requires good data that truly represents the target group. In this case, we needed teens to help us annotate data to create an effective dataset.
And this is why our philosophy at Pienso reflects the core belief that to create an accurate, useful, and effective LLM, we need the input of the subject matter experts who are dealing with relevant issues and topics in their fields on a daily basis. We wanted to put subject matter experts, whether a call center worker or a financial analyst, at the forefront of model creation, which gives them the opportunity to incorporate their understanding into the models and their outputs. And the best way to do that, in our opinion, was to create a platform that allows essentially anyone—not just machine learning experts or data scientists who have coding skills—to create their model.
Makaryan: What makes Pienso’s platform unique?
Jones: We have combined our old research from MIT with new innovations we have created after founding the company. We work with various chip providers like NVIDIA and Graphcore, a British semiconductor company that provides fast computing for AI models. A key aspect of our technology that sets us apart in the industry is that it does not require enterprises to “donate” their data to rent back a model. Our motto is that our customers should own their models, and we mean this quite literally. Since they can install this on their premises or fully in the cloud, only our customers have access to their data, which also gives them the ability to easily update their models if their data changes over time. We do not use any third-party application programming interfaces (APIs) so customers know that their data is secure. We also avoid volume-based pricing, instead charging customers only for the models that they decide to deploy.
Makaryan: Is there a customer success story that shows Pienso’s impact on an organization?
Jones: One of our premier customers is Sky UK, a major telecommunications company. They run over 500,000 calls per day from their call center through models that they created on our platform. The platform resides in their environment, and their analysts trained it themselves without needing programming skills. This system allows Sky UK to identify in real-time which customers are likely to churn, which ones need financial assistance, and what problems customers are facing that their online self-help tools were not able to resolve. They have moved beyond using Net Promoter Score (NPS) as their primary success metric, as our platform allows for much more nuanced analysis of conversations that call center workers have with customers.
As a result, Sky UK saved over £10 million last year and improved their customer service: the U.K. Office of Communications rated it as the company with the best customer service in Europe.
Makaryan: What is your vision for Pienso’s future?
Jones: We envision a future where creating models based on subject expertise becomes part of an employee’s regular duties to enhance their performance and streamline their work regardless of their field. This is because duty-specific knowledge represents valuable intellectual property for the company and for the employee. The tribal knowledge of workers within an organization, when digitized into these models, becomes a significant asset. We have worked with various organizations, including the U.S. government in the intelligence space, and I believe that everyone has some kind of expertise. Combining that expertise with enterprise data not only allows to create impactful AI models but to also bridge the gap between AI and non-technical experts. Whether it is a call center, intelligence analysis, or any other field, these specialized models represent a valuable way to capture and utilize expert knowledge. We hope Pienso can be a valuable platform to help a variety of enterprises navigate the transition into an AI-powered business model.
