Why every agent starts in read-only mode
Agents in Orvyn can't change anything until you switch them on. Here is why we built it that way and how to use the trial period well.
Lena Novak1 min read
When you connect a new agent, it starts in read-only mode. It can read your inbox, your accounting system and your helpdesk, but it cannot send, pay or update anything.
Trust is earned with evidence
Teams don't hand over work because a demo looked good. They hand it over after they've seen the agent make the same decisions they would have made, on their own data, many times in a row.
In read-only mode the agent runs next to your team. For every task it records what it would have done and why. You compare its decisions with your team's and see exactly where they differ.
What to look for in the first week
- Matches: how often the agent reached the same decision as your team.
- Escalations: whether it asks for help in the cases you would expect.
- Explanations: whether its reasons are clear enough to check quickly.
Most teams switch an agent on when it matches their decisions in more than 95% of cases and every difference has a reason they agree with.
Switching on is not all or nothing
You can turn on one rule at a time. Many teams start by letting the agent handle the simplest 60% of the work and keep the rest in read-only mode for another few weeks.
