Caiden SMS AI

The problem

Small businesses miss follow-ups because voice, SMS, and email live in different systems with no shared context. A patient who called yesterday and texted today gets a stranger both times. A customer who booked a viewing via voice and then texts to reschedule triggers a manual coordination scramble. Every disconnected channel is a place the business loses continuity, and every context reset is a place trust erodes. Most SMS bots make this worse — they respond in isolation, treat each thread as a new session, and can't reference the caller's prior interactions. Autoresponders and reminder blast tools solve narrow problems but don't create the unified experience that turns messaging into a real customer channel.

Key challenges

SMS is asynchronous and users don't expect immediate responses — but they also don't expect to be forgotten. The bot has to remember the thread, remember the voice call from two days ago, and reconstruct enough context to be useful without asking the customer to re-explain themselves. Handoff to a human agent needs to preserve the full multi-channel history, or the human ends up re-asking the same questions the AI already handled. And carrier compliance (A2P 10DLC registration, opt-in tracking, STOP handling) is a nontrivial engineering surface.

What we built

Caiden SMS is a channel extension of CAIDEN — text conversations that share memory with the voice AI. Every customer interaction lands in the same conversation store, keyed by phone number, indexed for retrieval, and available to whichever channel the customer uses next. Voice call summaries feed the SMS context; SMS threads feed the voice call context. The SMS pipeline handles carrier compliance (10DLC, opt-in/out, delivery receipts), threads bidirectionally with shared session state, and hands off to human staff with full multi-channel transcript. Appointment confirmations, refill requests, insurance follow-ups, and reminders all flow through the same context — a customer texting "reschedule" doesn't need to re-explain what they're rescheduling.

Our approach

1. 1

   Unified conversation store across channels

   One session store, keyed by phone number, with per-turn entries tagged by channel (voice, SMS). Retrieval on new inbound message loads the last N relevant turns regardless of channel. This is the design that makes cross-channel context feel effortless.

2. 2

   Compliance-first SMS pipeline

   A2P 10DLC registration, opt-in tracking, STOP/HELP handling, delivery receipts, and per-tenant sender identity built into the SMS gateway from the start — not bolted on after the first compliance incident.

3. 3

   Async-friendly response timing

   SMS bots that respond in milliseconds feel robotic. The system enforces natural response cadence and respects human-typing pauses, which improves engagement without adding meaningful latency to actual resolution.

4. 4

   Handoff preserves full multi-channel history

   When a human agent takes over, they see the entire cross-channel thread — voice transcripts, prior SMS, and current context — in one view. The customer never has to re-explain.

Key architectural decisions

Results

• Unified SMS + voice context per customer, retained across sessions
• Automated appointment confirmations and reminders
• Faster response to inbound texts — measured in seconds during business hours
• Full audit trail per customer across voice and SMS channels
• 10DLC-compliant carrier delivery with high deliverability
• Handoff to human preserves cross-channel history
• Multi-tenant deployment across CAIDEN's healthcare and real-estate clients

Impact

Adding SMS didn't create a new product; it made the existing product better. Customers experience CAIDEN as one system that remembers them, regardless of how they choose to reach out. That continuity is what shifts messaging from a nuisance channel to a real business tool — and it's why clients renew.

Tech stack