“Let’s create a world where every interaction with governments is assisted by digital agents”
Let's not.
Those pseudo-AI agents never understand what it is you're trying to explain, because they can't. They respond to a fixed set of rules and if you don't fit, you end up literally cut off (the "agent" hangs up on you).
Back in 2019, I had my fiber connection installed (finally). I'd had it for one month when a passing tractor ripped it off the post - it wasn't broken, it still worked, but it was on the ground (living in the countryside has its drawbacks). The problem was, the connection still worked. I phoned the nearest police station to ask what I had to do, and the officer politely told me that it was a dangerous situation and that, if anyone got hurt, I was liable to end up responsible.
So I called my provider's support line. The automated agent drilled my down through choices 2, 1, 2, 4, and then decided to test my connection and hung up. I knew that wasn't going to work, so I waited for the test to finish and I called again. Same rigmarole. I was starting to get angry by that point when the phone rang and an actual human being from the support line asked me what the problem was. I still had to explain twice that no, my connection was fine, but the physical wire was on the street. When he finally got the message, he promised that I would have a ticket and he would see what he could do.
Three weeks later, nothing had been done. Then a passing motorcycle ripped the line in two. This time, no more connection. Call support again and, lo and behold, there is a real problem (for them, that is). Two days later, a new line was installed and, after explaining to the technician, he installed it a meter higher on the pole.
And that happened with actual humans answering the phone. I shudder to think how a pseudo-AI would handle the issue.