The multi-channel customer inbox: SMS, WhatsApp, voice, email in one record

A customer of a small business in 2026 does not stick to one channel. They text from their car. They WhatsApp on the weekend. They call when they need to talk through something. They email when they have a long question. They DM the business's Instagram when they saw the latest post. By the end of a normal customer relationship, there is a record on five different surfaces, each one its own little inbox.

A small-business owner running this with normal tools ends up checking five inboxes a day, none of them coherently. The customer who texted Tuesday and emailed Friday looks like two different leads on different surfaces. The team coordinates badly because nobody knows the full state of any one customer.

The multi-channel customer inbox fixes the surface problem. Every channel writes into one customer record. Every team member sees the same picture. Every conversation is one continuous thread regardless of which channel each turn happened on. This post is what that looks like and how it changes the operations.

What "one inbox, all channels" actually looks like

A customer record in the multi-channel inbox has a chronological merged view of every interaction.

Mrs. Henderson - customer record
=============================
 
Channels: SMS (404-555-0193), WhatsApp (same number),
          Email (henderson@gmail.com)
Joined: Aug 2024
Lifetime value: $2,140 across 7 visits
 
Latest:
2026-05-06 14:22  call (5min) - capacitor question;
                  booked Tom for Thursday 11am
2026-05-04 09:15  SMS - "any chance you could come earlier?"
2026-05-04 09:16  SMS (auto-reply, agent) - "let me check
                  the day's roster, will get back in 10"
...
 
Open thread context: thermostat, mentioned in March visit,
flagged for follow-up; not addressed yet.

The merged view solves three problems at once. The team member opening the record knows the full state. The agent picking up a new call has the context. The customer never has to re-introduce themselves.

The architecture, in plain English

Three pieces have to be true.

Identity stitching. When a message arrives on any channel, the agent maps it to a customer record by phone number, email address, or other identifying signal. New customers get a new record; returning customers extend the existing one.

Cross-channel memory. The agent's memory of the customer (their preferences, their history, their flagged issues) is shared across channels. A note made during a phone call shows up in the next SMS thread.

Per-channel render. When the agent replies, it picks the channel the customer prefers (or the one they just contacted on) and renders the reply with that channel's conventions (markdown for WhatsApp, plain text for SMS, signed for email).

The user-facing simplicity is enabled by these three pieces working together. Drop any one and the inbox becomes incoherent.

What this changes about the operation

Three concrete shifts.

Handoffs are clean. Mrs. Henderson called. Tom takes over the conversation from the agent. Tom sees the entire history of Mrs. Henderson, on every channel, in chronological order. He does not re-ask her questions she already answered.

Reporting is honest. "How many open threads do we have?" The number is the number of customers with unresolved issues, not the number of threads in five different inboxes that may or may not be the same person.

Marketing is sharper. "Send everyone who has a thermostat-replaced-but-flagged-for-follow-up note a check-in text." The query runs against the customer records, not against per-channel send tools. The targeting is real.

What about per-channel preferences

Customers have channel preferences. Some prefer text; some prefer email; some only respond on WhatsApp. The agent learns these per-customer:

Mrs. Henderson - prefers SMS for short questions,
                  prefers calls for anything urgent.
                  Almost never opens email.
 
Mr. Chen      - prefers email for detailed quotes,
                 calls when something is broken.
 
Mrs. Patel    - WhatsApp only. Spanish-speaking.

The agent honours these preferences. When you ask the agent to "send Mrs. Patel an estimate," it goes via WhatsApp in Spanish, not via email in English. The preference is captured from observed behaviour and is editable in plain English.

Where multi-channel inbox products typically fail

Two failure modes most existing tools have.

Lossy channel coverage. Many "omnichannel inbox" products have deep SMS support and shallow voice support, or vice versa. The integration depth determines whether voice transcripts are searchable, whether email threads merge cleanly, whether a WhatsApp voice memo is findable. Lopsided coverage looks fine on a dashboard demo and falls over in real use.

Stale identity stitching. A customer who calls from a different number gets treated as a new customer. The history fragments. The fix requires manual merging, which most tools support badly. Ajeris's stitching uses multiple signals (phone, email, voice patterns, name + context); ambiguity is flagged for human resolution.

What we still need to ship

Two open items.

Cross-business inbox for franchise operators. A multi-location restaurant operator wants one inbox view across all locations. We have it for single-business agents; the multi-business view is on the roadmap.

Custom social channels (TikTok DMs, LinkedIn). Each platform has its own API surface. We add new channels as they earn it. Most small businesses do not have meaningful inbound on TikTok DMs or LinkedIn, so the priority is calibrated.

Try it

The first time you ask 'what's pending with Mrs. Henderson' and get a complete cross-channel summary, the system has converted.