TCG Digital makes AI do the shift work

Consider the problem of choosing a pool of blood plasma. The units have different ages, volumes, expiry dates and antibody concentrations. There are manufacturing requirements to satisfy. A choice that looks sensible in isolation may be wasteful across the whole inventory. At ADMA Biologics, this had been a heavily manual exercise. In early 2024, the company reported producing its first ASCENIV batch using ADMAlytics to automate plasma pooling. Behind that software was TCG Digital’s mcube platform. Here was enterprise AI with a rather concrete assignment: help choose what goes into the next batch.

Start with the pool

The interesting detail in TCG Digital’s ADMA case study is the sequence. Teams first held workshops across departments to identify operational needs. They built a shared data foundation. For pooling, they applied mixed-integer programming, a mathematical method for choosing combinations under constraints. Other applications addressed donor information, production indicators, yield forecasts and document search.

For an executive shopping for AI, that sequence is useful. The model arrives after the problem has acquired a shape. Expiry dates and inventory constraints give the software something definite to work on. The operator can then judge whether the recommendation makes sense.

“a standardization that could not be achieved through manual pool construction.”

Robert Brooks · ADMA Biologics
On the plasma-pooling implementation

ADMA’s subsequent annual filing describes reduced variability and personnel hours, alongside better production visibility. It also describes wider benefits as expectations. The distinction matters: a batch produced using the software is an observed event; future earnings improvements remain a forecast.

First, teach the data to agree

TCG Digital is the AI and digital business of The Chatterjee Group. It sells mcube alongside the people needed to connect it to an enterprise. Its services include data engineering, cloud transformation, application development, IoT and cybersecurity. That combination places it between enterprise software vendors and consulting firms.

The platform’s architecture follows a comprehensible progression. mcube.data brings information together. Its ontology layer supplies shared business meaning. mcube.ai provides models and agent capabilities. An ontology, in this setting, records what things are and how they relate: which asset belongs to which process, for example. That helps a system interpret a record in the context of the operation.

The current mcube 5.5 release adds an Agentic Plane, described as a governed workspace for approved models, agents and tools. Its stated controls cover access, policies, usage, cost and risk. September 2026 demonstrations showed asset-failure prediction and scientific reporting. These are demonstrations of the platform’s direction, rather than proof that every customer has autonomous operations.

The airline in its family tree

There is an unexpectedly old airline ticket in this story. In 1999, United Airlines and TCG formed SkyTECH Solutions in Calcutta, now Kolkata. Contemporary reporting described a programming shortage and an initial team of 50. United held 49 percent of the venture. TCG Digital’s aviation business traces its roots to that company.

Seen through that history, today’s focus on coordination feels less surprising. Airlines have long needed separate systems and teams to exchange information before something expensive happens. A refinery has its own version of that challenge. The operating vocabulary changes; the need to connect a signal with an action persists.

TCG Digital’s market argument rests on this combination of integration and industry knowledge. A customer can assemble data infrastructure and models separately, or hire a systems integrator. TCG offers a platform plus implementation expertise, with partners who understand the physical process. The useful comparison for a buyer is how much work remains between a software demonstration and a functioning workflow.

A refinery gives the idea weight

In November 2020, TCG Digital and Lummus Technology formed Lummus Digital. Lummus brought process-technology expertise; TCG Digital brought its data and AI capabilities. Both sit within the wider Chatterjee Group ecosystem, giving this collaboration an industrial family connection.

In January 2026, TCG Digital announced that a mcube-powered LC-Max Digital Suite had gone live at HPCL’s Visakh refinery. Delivered by Lummus Digital, it combines plant-data ingestion, monitoring, predictive intelligence and economic visualization. The target is a residue-upgrading facility, where low-value material becomes higher-value distillates.

A separate collaboration announced with Linde Engineering in December 2024 proposed linking mcube optimization with Linde’s AI Plant Control. The ambition was to identify profitable operating modes and move plants toward them automatically. Such arrangements reveal the expertise the software needs around it: process knowledge, control systems and operational responsibility.

The price is a project

Enterprise buyers purchase a working arrangement of software, integration and specialist effort. UC Procurement’s October 2025 agreement makes that arrangement tangible. It covers data services, aggregation and analysis, with pre-negotiated service fees and staffing. It also includes specified data-protection provisions. The announcement describes more than 15 years of work with UCLA.

That is a revealing form of distribution. A procurement agreement helps institutions buy defined work under established terms. It does not mean every campus has deployed the same AI application. TCG’s commercial reach depends on converting those opportunities into projects that solve local problems.

A decision you can measure

The company is extending the same platform into new settings. In July 2026, it announced licensing intellectual property and knowhow from Memorial Sloan Kettering Cancer Center to develop and commercialize PlanQ, supporting radiation-oncology workflows. The announcement concerns a development and commercialization initiative, with mcube as its foundation.

The lesson a reader can borrow is practical: choose a decision with a recognizable cost, connect the evidence it needs, and measure what changes. Reliable data, explicit rules and an accountable operator are prerequisites. Without them, faster recommendations merely accelerate uncertainty. The plasma pool offers a useful starting point: a bounded problem, a real production process and an outcome someone has to stand behind.