LinkedIn Digest: three public posts on running agents in a big company
A McKinsey architecture blueprint, Walmart’s mix of vendor models and its own, and a ServiceNow president on managing agents like a team. Short quotes, links and my commentary.
Three public LinkedIn posts from the last four days ask the same thing: once a company has more than one agent, who designs the whole? I quote a few words from each, link the original and say what I think. Nothing here comes from anyone’s private feed.
The short version
Each item is a short quote from a public LinkedIn post, with the author and a link. The commentary is mine. I read the posts as published on October 3 and October 6, 2026. I did not watch the CNBC-TV18 interview or read the McKinsey blueprint, so I only report what the posts say.
Bruno Aziza: the second agent is where architecture shows up
“Everyone’s shipping an agent. Almost nobody has an actual architecture underneath it.”
Bruno Aziza, October 3, 2026, on a proposed blueprint from QuantumBlack, AI by McKinsey. Original post
My take: The useful claim is the order of work. A first agent proves an idea. A second one tests whether the first was built as something others can plug into, and teams that skip that question rebuild later. It is also a claim a consultant would like to be true, so ask what the blueprint says about identity, permissions and logging before you borrow its diagram.
Not verified: I did not read the McKinsey blueprint, only Aziza’s description of it. The post gives no evidence for teams hitting a wall at the second or third agent. That is his observation.
CNBC-TV18: Walmart says both, vendor models and its own
“Frontier AI or build your own? Walmart’s answer: Both!”
CNBC-TV18, October 6, 2026, on a conversation with Walmart CTO Suresh Kumar. Original post
My take: The post says Walmart uses OpenAI and Anthropic models and also builds its own platform, Element, with an intelligence layer that routes each task to the right model. The design choice worth noting is the router in the middle: it lets a company swap vendors without rewriting every application. What the post cannot tell you is how tasks are chosen, what it costs or how often the router is wrong.
Not verified: I did not watch the interview. The post is a summary by the broadcaster, not Walmart’s own statement, and it gives no numbers.
CNBC-TV18: a ServiceNow president on managing agents like a team
“I manage around 8 or 10 AI agents.” And from a second post: “The software part is 20%. For an enterprise to maintain the software is 80%.”
CNBC-TV18, October 6, 2026, quoting Amit Zavery of ServiceNow. Original post · Second post
My take: The two lines fit together. If one person supervises eight or ten agents, the hard part is not writing the code but owning what runs afterwards: updates, access, fixing it when it breaks. The 80 percent figure is an executive’s estimate in an interview, not a measurement, and ServiceNow sells enterprise software, so read it with that in mind.
Not verified: I did not watch the interview, so I cannot say what “manage” means in practice or how the second quote was framed.
Sources and limits
I opened each public post on October 6, 2026. Engagement counts are not used. I did not contact the authors, and I did not use anything from a private feed.
- https://www.linkedin.com/posts/brunoaziza_mckinseys-enterprise-agentic-platform-architecture-activity-7512149678281416704-23KF
- https://www.linkedin.com/posts/cnbc-tv18_walmart-aistrategy-openai-activity-7513199282368466946-yb3N
- https://www.linkedin.com/posts/cnbc-tv18_globaldialogues-amitzavery-servicenow-activity-7513264645701423106-aOrx
- https://www.linkedin.com/posts/cnbc-tv18_servicenow-ai-amitzavery-activity-7513267089625882624-Z5pN