How do enterprises move from AI pilots to real, scalable outcomes? Box Director of Product Management Matt Terrell sits down with Shyam Suresh, Segment Head of Business Strategy at TCS, to unpack what it actually takes to build an AI-ready organization. They cover two pillars every leading AI adopter shares, AI-first culture and AI-first architecture, and why most enterprises are still stuck at the pilot stage. Matt and Shyam explore why AI agents fail not because of their capability, but because they lack the business context to act intelligently. They discuss how Box serves as the content intelligence layer that gives agents the context they need, and how TCS is helping enterprise clients navigate this transformation across financial services and beyond. Shyam opens by identifying the two traits that separate AI leaders from laggards: an AI-first culture, where AI is embedded in every process, every team, and every customer interaction, and an AI-first architecture built on a scalable infrastructure layer, a governed data layer, a flexible tooling layer, and an orchestrated application layer. The conversation digs into why adoption so often stalls at the individual level. Matt shares a candid observation: people given a SaaS license aren't automatically prepared to use it. Shyam's answer is to stop forcing AI into workflows and instead start with business outcomes, improving NPS, reducing risk, accelerating end-to-end productivity, so that demand for AI comes from employees, not mandates. Shyam shares how TCS transformed its own 600,000-person organization by running a company-wide hackathon, from senior leadership to new hires, to build an AI-first problem-solving mindset before deploying tools at scale. On architecture, Shyam walks through a five-layer AI stack: infrastructure, data, tooling (LLMs), application, and insight. He makes a critical point: the real failure mode for agentic systems isn't missing data, it's missing context. Agents can't build entity relationships on the fly without cost overruns and errors. That's where the data layer, and specifically unstructured data, becomes the enterprise moat. Matt connects this directly to Box's role: Box is the "librarian in the Library of Congress", an intelligent content layer that welcomes AI agents, understands what they're looking for, navigates the ontology of your unstructured data, and directs them to the right files. Rather than a needle-in-a-haystack RAG problem, Box pre-synthesizes context so agents can act with precision. The pair also discuss model neutrality, why you don't need a frontier model for every task, and Satya Nadella's concept of Token Capital Management (TCM): managing AI token costs as deliberately as human capital. They close with a look at the evolving operating model: federated innovation with centralized governance, teams where 2 of 10 members are already AI agents, and the shift from seat-based to agent-based pricing. Chapters: 00:00 — Introduction: Matt Terrell and Shyam Suresh 01:38 — The two pillars of AI-first organizations 03:02 — Why AI adoption stalls at the individual level 04:17 — Shifting from technology to business outcomes 05:13 — How TCS transformed 600,000 employees with a hackathon 06:31 — AI-first architecture: the five-layer stack 08:05 — Why agents fail: the context problem 11:24 — Box as the content intelligence layer 13:43 — The "librarian" metaphor: Box and unstructured data 15:01 — Model neutrality and right-sizing AI 15:53 — Token Capital Management (TCM) 17:46 — The evolving operating model: federated innovation, centralized governance 20:08 — AI agents as team members: 2 of 10 are already agents 21:19 — The shift to agent-based pricing 23:40 — Financial services use cases: front office and back office 25:37 — Box and TCS: automating insurance claims processing 29:45 — Closing thoughts: what it means to be AI-ready FAQs: Q: What are the two key traits of organizations leading AI adoption? A: Leading organizations share two traits: an AI-first culture — where AI is embedded in every process and team — and an AI-first architecture built on a scalable infrastructure, governed data, flexible tooling, and orchestrated application layers. Q: Why do AI agents fail in enterprise environments? A: Agents fail not because of capability, but because they lack business context. Without a governed content layer, agents build entity relationships on the fly — causing cost overruns and unreliable outputs. Q: What role does Box play in an AI-first architecture? A: Box is the content intelligence layer — the "librarian in the Library of Congress." It helps AI agents navigate unstructured enterprise data and access the right files within existing governance and compliance controls. Q: What is Token Capital Management (TCM)? A: A term coined by Satya Nadella for managing AI token costs efficiently — being intentional about which tasks use frontier models versus smaller, more cost-effective ones.

 126      2