#337 Debdas Sen: Why AI Without ROI Will Die (Again)

What It Covers

Debdas Sen, CEO of TCG Digital, explains how his firm deploys hybrid AI combining proprietary knowledge graphs, enterprise data, and external LLMs to solve high-stakes industrial problems in energy and life sciences, arguing that AI without measurable ROI risks repeating the collapse seen after the 1990s hype cycle.

Key Questions Answered

• •ROI threshold as project filter: TCG Digital applies a 10x return benchmark when scoping client engagements — if a client spends $5M, the target outcome is $50M in recovered value. This forces problem selection toward large, core operational functions like manufacturing optimization and R&D acceleration rather than enabling functions like HR or finance.
• •Hybrid modeling over pure AI: In energy applications, combining chemical kinetic models with machine learning outperforms either approach alone. Pure neural networks ignore mass balance constraints that chemical engineers require, while first-principles models miss patterns in data. Enterprises that default entirely to one method consistently underperform on accuracy and stakeholder trust.
• •R&D cycle compression via virtual experimentation: Using multi-agent reasoning across internal proprietary data, curated knowledge graphs, and public LLMs, catalyst formulation candidates can be narrowed from millions of combinations down to five to fifteen testable options. This reduces the candidate selection phase from twelve months to one month — a 12x acceleration in early-stage R&D.
• •Trust architecture for agentic enterprise AI: Hallucination risk from external LLMs is managed by validating all outputs against internal enterprise data before any decision reaches management. Keeping final reasoning within the enterprise boundary — with external models contributing context, not conclusions — makes agentic systems acceptable to Fortune 100 clients with strict IP requirements.
• •Career positioning for high-stakes AI roles: The next wave of valuable AI practitioners will combine hardware-to-application stack knowledge with deep sector expertise. With roughly $400B invested in AI in 2025, enterprises will demand ROI accountability, meaning practitioners who understand specific business processes — not just model architecture — will drive the deployments that survive.