AIToday
Large Language ModelsAI Business & IndustryPractical AIPublished: Sep 24, 2026, 19:00 JST

VMware Tanzu's Nick Kuhn: Treat AI agents like any app

VMware Tanzu's Nick Kuhn: Treat AI agents like any app

3 Key Points

  1. What happened

    Nick Kuhn of VMware Tanzu Platform said his team now offers a Tanzu Agent Build Pack that pushes an agent described in Agents.md to the same platform as Java apps.

  2. Why it matters

    Kuhn argues enterprises want agents deployed like apps so they can reuse years of platform engineering for consistency, auditing, and security. That is likely to reduce risk and cost of agent adoption.

  3. What to watch

    The approach hinges on whether agents' shared memory and state can be handled as cleanly as traditional app storage, since Kuhn says that is the biggest challenge. He also notes customers want agents running close to apps and data to avoid latency from being 30 hops away.

WHO IT HITSEnterprise platform teams and developers who manage private clouds or air-gapped data centers will get a familiar path to production for AI agents. Teams running agents alongside traditional apps may need to address state management and latency.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Context & Analysis

Nick Kuhn, a tech marketing lead at Broadcom's VMware Tanzu who previously spent fourteen years inside a large enterprise, says the core challenge is that enterprises are highly regulated, often with no internet access or even true air gaps where hardware is carried physically into data centers. He notes that vendors often assume internet access, which fails in these environments. Prior to agents, platforms like Cloud Foundry and Tanzu helped developers push code that the platform would turn into a secure container with ingress, certificates, and health monitoring handled automatically. Kuhn says the same principle now applies to agents: customers want the same repeatable pattern so they can audit and secure deployments without recertifying everything. The main technical gap he highlights is that agents were originally built with file system access and markdown files as memory, which breaks in ephemeral cloud environments where state must be saved elsewhere. He also says customers want agents and LLMs co-located with apps and data to avoid latency from many network hops, and that some SaaS providers are not reliable enough for traditional enterprises. The outcome hinges on whether state management and reliability can be solved in a way that matches the expectations enterprises have for software.

FAQ
What is a buildpack in this context?
Nick Kuhn explained that a buildpack is a set of best-practice commands used to create a container for a specific type of application. The Tanzu Agent Build Pack applies that same model to agents, so pushing an agent works like pushing a Java app.
Why do enterprises want agents close to their apps and data?
Kuhn said that if an agent or LLM is far away from the microservices calling it, latency adds up from many network hops. He also noted that SaaS providers may not meet traditional enterprise reliability expectations.
What is the biggest difference between deploying apps and agents?
Kuhn said traditional apps decouple storage and state cleanly, so they can scale to thousands of instances. Agents were originally built with file system access and MD files as memory, which becomes a problem in cloud environments where things spin up and down and are ephemeral.

Get the latest Large Language Models news every morning

For example, today's edition would include:

  • Meta's Muse Tops Apple's App Store; Bank, Insurer Stocks TumbleYahoo Finance AI · 49m ago
  • Schmidhuber joins Sakana AI to lead RSI LabITmedia AI+ · 3h ago
  • TypeSafe AI launches Jev, claiming up to 200x faster inference without hallucinationsITmedia AI+ · 3h ago

AI-summarized, only the topics you pick: one digest a day via Email, LINE, or Slack.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleOpenAI releases GPT-6 Astra, warns of cyber capabilities