Self-service plus smart escalation is the model: bots and suggested actions handle routine requests, humans take the complex ones, and the whole interaction stays in one rich thread. One large financial-services provider reported 42% fewer service calls, a 94% read rate on urgent notifications, and 300% faster resolution after moving support to RCS.
Looking ahead, RCS 4.0’s Messaging-Initiated Video Calls open a path where a stuck text conversation escalates straight to video with an agent, a natural next step for hard-to-resolve issues.
Deflection has to be designed rather than hoped for, and the design starts by sorting intents into three buckets. Automate the ones where the answer is a lookup and the customer only wants a fact: order status, balance, appointment time, store hours. Assist the ones where a form is the real work, rescheduling, address change, returns, using suggested replies so the customer taps through structured options instead of typing. Route straight to a human anything involving money in dispute, an account security concern, or a distressed customer, and do it on the first message rather than after a bot has failed twice. The third bucket is where automated service earns its reputation or destroys it.
The escalation path is the part most often left until last. A person who wants a human should be able to get one in a single tap, from a chip that is visible without scrolling, and the handoff has to carry the transcript so nobody is asked to repeat themselves. Agents need the thread in the same console they use for other channels, or the response times that make messaging attractive will not survive contact with a real queue.
Two-way messaging also creates an expectation that broadcast never did. A thread looks like a conversation, so people reply at midnight and expect an answer. Set that expectation inside the thread with stated hours and an honest response window, and make sure something answers out of hours even if it only acknowledges and books a callback. Silence in a channel that looks conversational reads as being ignored, which is worse than the same delay on email.
Staffing changes shape rather than size at first. Handle time per contact falls because context arrives with the message, but contact volume usually rises, since messaging is easier than calling and surfaces requests that previously went unmade. Plan for that, and measure resolution rate and reopen rate alongside call deflection. A deflection number that looks excellent while reopens climb means the calls came back through a different door.
- RCS service = self-service via suggested actions + AI, with human (and soon video) escalation in one thread.
- Example: 42% fewer calls, 94% read rate, 300% faster resolution at a global financial-services provider (anonymized).
- UP 4.0 Messaging-Initiated Video Calls (2026) enable text-to-video escalation as it rolls out.
- Sort intents into automate, assist and route-to-human, and send sensitive cases to a person on the first message.
- Make the human handoff one visible tap and carry the transcript across.
- Expect contact volume to rise even as handle time falls; watch reopen rate, not just deflection.