AI agent risk

1. Why this page exists

virtual24 gives you a team of AI agents, and AI agents are genuinely useful and genuinely fallible. The terms of service state the legal position in a paragraph; this page is the longer, plainer version, written the way we would explain it across a table.

2. What agents are good at

Agents excel at work with a clear outcome and a lot of surface: research that means reading forty pages, first drafts in your voice, analysis of a spreadsheet you did not want to open, the multi-step routine that eats an afternoon. They are fast, they do not get bored, and they hold many threads at once. Delegated well, they return finished pieces, not fragments.

3. Where they fail

AI output can be incorrect, incomplete, or out of date, and it fails confidently: a wrong number sits in the same neat sentence as a right one. Agents can misread an ambiguous instruction, fill a gap with a plausible invention, or carry an early mistake through every step that follows. The polish of the output says nothing about its accuracy. That is the single most important thing to know.

4. Your part

You are responsible for reviewing what your agents produce before you rely on it or send it to anyone. Treat agent output the way you would treat a capable new hire's first draft: usually strong, never unexamined. Nothing the service produces is professional advice - legal, financial, medical or otherwise.

5. Access and scope

Agents act only with the access you grant them. The credentials in your vault, the connectors you link, the tools you give each agent: those are the boundaries of what your team can touch. An agent with no access to a system cannot act on it, which means the scope of what can go wrong is a decision you make, deliberately, per agent.

6. Working with them well

Three habits cover most of it. Review before send: anything leaving your business passes a human eye first. Least privilege: give each agent the narrowest credentials that let it do its job, and no more. Human judgment on consequential decisions: agents prepare the decision, you make it, especially where money, people or reputation are on the line.

In plain words: let the team do the hours, keep the signature for yourself.

7. The legal position

virtual24 is not liable for decisions made on the basis of AI-generated content or for losses arising from it, as set out in the terms of service.