The future of AI is not just agents; is deep customization.
Instead of simple recommender systems that correlate user behavior to identify patterns and apply them to custom workflows, large language models (LLM) and AI agents can directly analyze users to create deeply personalized experiences.
This kind of aggressive personalization is increasingly demanded by users – and the savviest businesses that provide it (and soon) will win.
Here’s the goal: “Don’t try to randomize or guess who I am. I’m telling you, that’s what I’m interested in,” Lijuan Qin, Zoom AI’s product manager, explains in a new article. Beyond the pilot podcast.
How Zoom integrates customization
Zoom is one company adapting to this trend: Its generative assistant, AI Companion, goes beyond basic summaries, smart notes, and post-meeting action items to track engagement and user engagement.
Users can customize meeting summaries based on their specific interests and create targeted templates for follow-up emails to different people (whether it’s a salesperson or an account executive). An AI assistant can then auto-fill these documents after the call. Meanwhile, a custom dictionary in Zoom AI Studio can process unique enterprise terminology and vocabulary for more relevant AI results, and a deep research mode can quickly provide comprehensive analyzes based on “internal expertise and external insights”.
Control is key here; Qin explained that a human agent can be “very specific (and) nail-biting.” They have “very clear controls” on follow-up actions, such as: Can the agent automatically send emails to specific recipients? Or will it trigger a verification step when it recognizes that the transcripts contain sensitive information (as dictated by the user)?
Aware that AI can sometimes go off the rails, human users can monitor agent behavior in Zoom, enable and disable features, and control access to data. This can help prevent inaccurate or off-target outputs.
“Most importantly, we don’t think AI is smart enough to fix everything,” Qin said.
Getting the context right
Sam Witteveen, co-founder of Red Dragon AI and host of Beyond the Pilot, explains in the podcast that this new age of agent AI is essentially a “land grab for context.”
“Of course, knowing your users is a big thing, right? Knowing what apps they live in, what daily tasks they do?” he said. “Companies are realizing that the more information they have about you, the better the (AI) memory, the better they can personalize.”
Witteveen says that Claude Cowork is a program that “really shines”; OpenClaw is different. The models are good enough that they can begin to make decisions for users and respond to directions such as: "You know a lot about me. You have all this context. Go and build skills that will help me do a better job."
“With something like OpenClaw, you can customize it however you want, right? You can have a conversation with it, you can tell it, ‘Hey, at 4 o’clock, I want you to do this,'” Witteveen said.
However, token usage and security should always be considered, he advised. OpenClaw has been faced with a security problemsince its launch. This has prompted many enterprises to eliminate or completely ban the use of autonomous agents; however, these deletions must be done correctly so that IT leaders do not accidentally delete the entire enterprise stack.
Meanwhile, in terms of token budget, customization can increase costs. “You have to think about the metrics you’re tracking,” Witteveen said. “It’s very different from product to product, but metrics around these things will be key."
Check out the podcast to hear more about:
-
Why aren’t companies testing their AI capabilities now? "it can be toast"
-
How Zoom built an AI companion that tracks not just action items, but disagreements in your meetings
-
The question of why to build and buy has become more relevant for enterprise software
-
Why "skills" enterprise may be more important to the future of AI than MCP
You can also listen and subscribe Off the pilot about Spotify, apple or wherever you get your podcasts from.
