Birago Jones

Birago Jones runs Pienso, and Pienso sells a strange kind of software. It does not write your AI model for you. It does not rent you someone else's. It hands you a set of tools and a blank canvas and says: you know your data, now build the thing yourself. In a market obsessed with automating the human out of the loop, Jones spent a decade arguing the opposite.

The company he co-founded with Karthik Dinakar keeps a blunt motto near its center: customers should own their models, and they mean it literally. No donating your data. No renting the intelligence back at a markup. The models trained on Pienso belong to the enterprise that trained them, and they stay in-house, away from third-party APIs. For a US government client, or for Sky, the largest broadcaster in the UK, that distinction is not a slogan. It is the whole point.

Today Pienso lives at the intersection Jones has circled his entire career: the seam where human expertise meets machine learning. His bet is that the analyst who reads customer calls all day understands those calls better than any engineer ever will, and that giving her the ability to build a model, without writing a line of code, is worth more than another clever algorithm.

Back up to 2010. Jones and Dinakar are graduate students at the MIT Media Lab, and they take on a class project that sounds noble and looks simple: build a tool to help content moderation teams at places like Twitter and YouTube flag cyberbullying. They build it. The model works, technically, exactly as designed.

Then it meets the real world and quietly falls apart. The model cannot recognize how teenagers actually talk. Harmful content dressed in teenage slang slides right past it. The math was fine. The vocabulary was wrong. And the gap between the engineers who built the model and the kids it was meant to protect turned out to be the actual problem.

Here is the detail that matters. Jones found the flaw by reading the data patterns his co-founder had missed. Not by rewriting the algorithm, by looking at what the data was actually saying. The pair took their approach to Cambridge-area teenagers and let the kids help train the model directly. Accuracy went up. A thesis was born: the person who understands the data should build the model.

Six years later, in 2016, that thesis became a company. Pienso spun out of MIT with seed money from Eniac Ventures and the MIT E14 Fund and a mission that has barely wavered since: the leading machine learning platform for people who do not program.

Ask Jones where enterprise AI goes next and he points at the quiet expertise sitting inside every company. The senior analyst who can smell a fraudulent claim. The support lead who knows which complaints turn into churn. That knowledge usually walks out the door at retirement. Jones wants to capture it, digitize it, and turn it into a model the company owns forever.

Pienso's interface is built so a business analyst, not a data scientist, can process data, train a model, test it, and push a live version, all without touching code. The models are imprinted with the user's expertise and fine-tuned to answer their specific questions. To keep the compute honest, Pienso works with chip makers like NVIDIA and Graphcore, and it charges only for the models a customer actually deploys.

At Sky, that philosophy translated into something concrete: roughly half a million customer calls processed a day, feeding reported savings of more than £7 million. Not a demo. A production line, run by the people closest to the problem.