Context Labs launched AI Distillery on 1 November and says the software lets businesses run artificial intelligence (AI) on their own operational data without handing control of that data to a model provider. The company grew out of research at the Massachusetts Institute of Technology (MIT) and offers the software through the Microsoft Marketplace, with a starter kit for the buyer’s industry in each enterprise licence. Pricing and customer names have not been disclosed.
The launch is timed for ADIPEC, the Abu Dhabi International Petroleum Exhibition and Conference, which runs in the United Arab Emirates from 2 to 5 November. At the show, Context Labs is demonstrating an AI agent that answers questions about plant operations and traces every figure back to the plant’s own records. Visitors can see the demonstration on the Arab Development Group stand and in the Dell AI Factory with NVIDIA area of the AI Zone.
Founder and chief executive Dan Harple argues that a company’s knowledge should outlast any single model. “Models come and go. An enterprise’s context should not,” he said. Harple added that the most advanced models from the large AI labs do not supply governance, trust or provenance, the reliable record of where data came from. He also said such models should not absorb the knowledge that makes a company competitive.
How AI Distillery Works
Context Labs describes the process as distil, package and prove. The software first pulls in database tables, documents and other records and maps them to an industry ontology, a shared vocabulary that defines how equipment, sensor readings and products relate to one another. According to the company’s product page, a person stays involved as the data is profiled, and every data point arrives with a record of its source, its timing and how it was processed.
Context Labs stores the mapped data in what it calls a Context Package, together with cryptographic provenance that keeps a tamper-evident record of each value’s origin. The package also holds the company’s governance rules, ownership policies and business metrics. Before the system goes live, the software checks a model’s answers against the company’s own records, and Context Labs says the package stays under the customer’s control and is not used to train third-party AI models.
Staff can put operational, commercial and compliance questions to Context Agents, which trace every figure in an answer to its source. A feature called the Semantic Context Tunnel also brings the company’s data into the AI assistants employees already use, and both tools follow the access controls the business sets.
Model Choice And Rival Offerings
Customers can run AI Distillery in their own data centre or in a cloud of their choice. AI assistants such as Claude and Copilot can query a Context Package through the Model Context Protocol (MCP) or Agent2Agent (A2A), two open standards for connecting AI systems to data and to one another. Context Labs says switching models is a configuration change rather than a rebuild. The company also says precise context means each query needs fewer tokens, the units of text a model processes and charges for.
Context Labs is pitching the product at companies whose AI projects stall because their data is poorly organised. Research firm Gartner predicted in February 2025 that organisations would abandon 60% of AI projects unsupported by AI-ready data through 2026. Other companies have released related products in recent weeks. Hyderabad cloud firm Digi edZe launched a website for private AI that runs on a client’s own infrastructure, and Boomi and Jade Global added audit trails to AI agents used in factories.
From Emissions Data To AI
Harple founded Context Labs in 2016 after a career building internet companies, including firms later bought by Netscape and Oracle. The company grew out of research he carried out at MIT as a Sloan Fellow, a mid-career programme at the university’s business school. Context Labs started as a project to verify information and counter misinformation before turning to emissions measurement around 2018.
The company’s earlier products include Immutably, which gathers and encrypts data from many sources and records it on a blockchain for third parties to verify. Its emissions product, Decarbonization as a Service, measures and verifies greenhouse gas output. An MIT profile of the company named EQT, Williams Companies and Coterra Energy among its clients. The profile also noted a Microsoft partnership that lets users of Azure, Microsoft’s cloud service, connect sensor data to the Context Labs platform.
The new product keeps the same industrial focus and is aimed at energy, chemicals, infrastructure, and mining and metals companies. Context Labs has offices in Amsterdam, Houston, Abu Dhabi and Cambridge, Massachusetts.
The launch takes Context Labs beyond emissions data and into the wider body of records industrial companies rely on, from plant readings to compliance files. The company’s case rests on ownership, with a business keeping its knowledge in a package under its own control while the AI model on top can change.
