Context Washing, AI Agents & Qualcomm's Big Bet | CRTV Episode 133

Episode 133 of ConstellationTV brought together analysts and industry leaders to tackle some of the most pressing questions in enterprise AI, from whether "context" is becoming meaningless marketing language to how Qualcomm is making a serious play for the data center.

The Context Washing Debate

Every major AI vendor now has a "context strategy." Snowflake, Databricks, SAP, Microsoft — they all claim it. But ask ten vendors what context actually means, and you'll get ten different answers: metadata, semantic layers, knowledge graphs, vector databases, memory, catalogs.

Constellation analysts Mike Ni, Holger Mueller, and Esteban Kolsky went head-to-head on whether context is a genuine architectural shift or the next AI buzzword. Mueller argued that while context is logically necessary, it's technically unfeasible at the scale vendors are promising; real-time context querying across complex enterprise environments requires infrastructure most organizations simply don't have yet. Kolsky agreed with the architecture challenge but argued that the solution lies in scoping the context to specific, bounded use cases rather than trying to solve it universally. Ni framed the takeaway: context is both marketing and necessary, but what the market currently has is not sufficient. Enterprises are context washing today — but this is where the next major AI control point will emerge.

OpenText: AI Agents as Network Managers

John Radko, SVP of Product Development at OpenText, joined to discuss how business networks are evolving. The core insight: AI isn't simply making transactions faster — it's beginning to actively manage the complex ecosystems that connect thousands of companies.

Radko described the near-term deployment of AI agents that help clients onboard partners faster, detect issues, and resolve them more quickly. Looking further ahead, OpenText envisions agents operating as the first line of defense — managing trading communities autonomously, as demonstrated by a recent deployment with a major automaker where an agent detected a file failure and automatically re-sent it once the issue was resolved.

His advice for enterprises: treat your business network as a platform and consolidate as much as possible into a single infrastructure. As agentic AI rolls out, it will leverage APIs and interfaces across your stack, but fragmented infrastructure will slow you down.

AI in Marketing: Less Stack, More Process

Liz Miller sat down with Tara DeZao of Pegasystems to discuss what AI is actually doing for marketing leaders right now. The headline: martech stacks are finally shrinking. After years of point-solution proliferation, enterprises are consolidating, and Pega's Customer Decision Hub is benefiting.

DeZao pointed to two standout capabilities: gap detection, where an AI agent proactively identifies underserved audiences and surfaces new opportunities mid-campaign; and centralized compliance, which lets marketers in regulated industries experiment more freely because guardrails are built in rather than bolted on after the fact.

The bigger takeaway: most AI failures in marketing aren't data problems or model problems. They're process problems. AI is most valuable when it's solving workflow challenges, not strategy questions.

Qualcomm's Data Center Bet

Holger Mueller closed the episode with his takeaways from Qualcomm's Investors Day. Qualcomm is making a significant move: from silicon and IP licensing toward platform-as-a-service, with data center infrastructure as the new revenue frontier.

The three-stage plan starts with connectivity offerings available now, moves to the Xi Accelerator for inference workloads, and extends to a CPU based on Orion infrastructure arriving in 2028. The headline acquisition: Modula, purchased for nearly $4 billion. Modula's compiler technology claims to run NVIDIA workloads more efficiently than NVIDIA itself does, which, if true, would have major implications for cloud vendors and enterprise AI infrastructure costs.

Physical AI rounds out the picture: Qualcomm has nearly swept automotive design wins and is expanding into industrial IoT and robotics. The ambition is clear: become the platform layer that powers AI wherever it runs.

Episode 133 makes one thing clear: enterprise AI is moving fast, but the real work isn't in picking the flashiest model or the loudest vendor; it's in getting the foundations right. Context, trust, governance, consolidation. The enterprises that nail those unsexy building blocks are the ones that will actually scale.