I've been watching the shift from agent-as-feature to agent-as-interface in SaaS products, and it's a bigger change than most people give it credit for. It fundamentally alters how users give instructions, how the product understands intent, how work gets executed, and, crucially, how much control people expect along the way.
A colleague in sales shared that their agent runs live in the background during calls - no dashboard to check, no recordings to scrub. It surfaces the important stuff in the moment. He described it as less like a tool and more like a sharp co-pilot who never misses a thing. That sounds compelling for narrow, high-stakes workflows.
But another colleague pushed back: if the agent isn't faster than clicking through menus manually, it's dead in the water. For repeat workflows, a good dashboard still beats a chat box. And a third person pointed out they'd only use an agent-first product if it shows its steps, asks before risky actions, and hands back control when needed. Low-key, they don't want everything hidden behind a dialogue box and just hoping the agent understood the job.
That last point rings true. Agent-first needs guardrails - transparency, speed, and the option to take the wheel. Are we ready for that kind of product experience? I'm not convinced most SaaS teams have figured out the balance yet.
