The Always-On Agent Is Here. The Enterprise Isn’t Ready, Yet.

OpenAI’s ChatGPT, followed quickly by Anthropic’s Claude, Google’s Gemini, and Microsoft Copilot, gave us a new way of interacting with machines. The experience was so unexpected, and so alarmingly easy, that nearly every product soon took on some flavor of generative chat. Users came to expect it everywhere. Our apps simply became easier to use across the board. And more fun, too, in the bargain. 

So, what of this new breed of services: Meta Muse, Instinct, OpenAI Dots, and others? Are they merely an extension of the chat interfaces we use today? After all, ChatGPT and Claude already use tools, keep memories, and move between voice and text with ever-increasing fluidity. Surely whatever these new apps bring to the table is just more of that same fluid chat experience, no? 

No. 

Personal agents are a different class of technology. For consumers, they’re remarkable. For the enterprise, they’re a risk no InfoSec team should accept. Until enterprise-grade versions arrive, they have no place inside corporate systems. 

Always On

The class of “applications” loosely called personal agents is built on the same fundamentals as chat apps: powerful LLMs to drive conversation and reasoning, memory that improves over time, and tools that act on the world in powerful ways. But personal agents have one characteristic that sets them completely apart. They are always on. 

A chat app makes it easy to work with tools and data. A personal agent works for you, individually, as a person, with a consistent view of you and your world that feels closer to a relationship. It gets to know you, expresses opinions, does your bidding at all hours, and, crucially, starts conversations instead of waiting for you to type. 

That proactivity is a fundamental departure from anything we’ve seen in even the most powerful chat interfaces. As users, we’ll need to adjust to the weight it slowly lifts from our shoulders. It’s amazing, and it’s exactly what makes these agents so risky at work. 

Why they don’t belong in the enterprise, yet 

Giving one of these agents access to your work email, Microsoft 365, or any SaaS or agentic business application would be a serious mistake, for two reasons. 

First, they’re too new to have shown the controls required under SOC 2, ISO 27001, ISO 42001, GDPR, or any other important governance standard. 

Second, they won’t guarantee that they won’t train on or retain your data, and it’s unrealistic to expect them to. An agent built to learn from you across multiple systems needs exactly that data to work. Once connected to the enterprise, these agents immediately become data sub-processors, and suddenly the InfoSec team has a lot of explaining to do. 

So, what does this mean for enterprises? Given the sensitivity of their data and what’s at stake, InfoSec teams should block consumer personal agents completely and prohibit them by policy. That should hold until enterprise-safe versions are available and security professionals have vetted them. There is no universe in which giving a powerful consumer agent with no data protections access to corporate systems is a good idea. 

However. 

The demand isn’t going anywhere 

BEmployees who use an always-on agent at home will expect the same at work, and that pressure won’t stay outside the firewall for long. The best AI technology vendors see this coming. They’ll respond with agents built for the enterprise from the ground up, which means: 

• Governed access to only the data they need 

• No training on customer data 

• Auditable actions 

• Compliance with the standards your security team already relies on 

Procurement is one place this is already happening. At Globality, our sourcing agent, Glo 2.0, is taking on these characteristics within those guardrails, adding user and corporate memory, proactivity, agent-to-agent communication, and always-on operational readiness. 

This new class of enterprise agents will serve as always-on, always-available teammates, initiating activity on their own while enabling every individual to reach their full potential. 

The personal agent is coming to the enterprise. The only question is whether it arrives safely.