How would you measure the success of TCS WisdomNext, TCS's GenAI aggregation platform, for enterprise clients?

Introduction

TCS WisdomNext, launched in June 2024, is TCS's generative AI aggregation and orchestration platform. It lets enterprises access multiple large language models through one interface, compare them with evaluator bots, reuse industry blueprints and control cost, instead of building separate integrations for each model vendor. Measuring success for a platform like this is tricky: it is sold to enterprises, used by developers and business teams, and sits on top of models TCS does not own. I'll define product context, goals, a North Star metric, supporting metrics, guardrails, trade-offs and counter metrics, and then suggest initiatives.

Step 1

Product Context

WisdomNext targets large enterprises in banking, insurance, retail, manufacturing and life sciences, often existing TCS clients.

Key stakeholders:

• Client CIOs and chief data/AI officers: Want a governed way to scale GenAI without vendor lock-in.
• Developers and AI engineers at the client: Build applications using the platform's APIs, prompt tools and blueprints.
• Business users: Use the resulting applications, such as claims summarisation or customer-service copilots.
• TCS delivery teams: Implement use cases on WisdomNext as part of larger engagements.
• Model providers: OpenAI, Google, Anthropic, Meta (open models) and others whose models are orchestrated.

User flow:

1. Client onboards WisdomNext (TCS cloud tenancy or client's cloud).
2. Teams select a blueprint (for example, "insurance claims summariser").
3. Evaluator bots compare candidate models on the client's sample data for quality, latency and cost.
4. The team deploys the application with guardrails (PII masking, content filters).
5. The platform routes production traffic, monitors cost and quality, and suggests model switches.

Lifecycle stage: early growth. The platform is in market and deployed with large clients, and the priority is moving from pilots to production at scale.