For repetitive, rules-based work that needs to happen fast and around the clock, AI agents usually win. For judgement-heavy work, people do. Most teams end up with both: agents take the busywork, people take the decisions.
Staying as you are is reasonable if volumes are low and your team has real spare capacity.
Start with one agent on your most repetitive task and keep your team on everything that needs a human. Measure the hours saved for 30 days, then decide what to do next.