Most small teams don't lose customers because they're bad at their jobs. They lose them because a WhatsApp message sat unread over the weekend, a website chat went nowhere at 9 p.m., and the person who took the phone call never wrote anything down. The channels are scattered, and so is the context.
AlmaTalk's answer is an omnichannel inbox: voice calls, WhatsApp, SMS and website chat land in one place, share the same notes, and can be assigned per thread. Here's how that actually works, and where the honest limits are.
One inbox, four ways in
An AI answering service is only useful if it covers the channels your customers actually use. AlmaTalk handles four:
- Voice — the AI receptionist answers phone calls 24/7, in English and French, and can book appointments or take a message during the call.
- WhatsApp — replies to messages the same way it handles a call, and the thread stays visible to your team.
- SMS — text-message conversations flow into the same inbox.
- Website chat — a chat widget on your site, answered by the same agent.
The point isn't that these exist separately — plenty of tools do one channel. The point is that they arrive in a single inbox, so nobody has to check five apps to see what a customer already asked.
One practical note: which channels are available depends on your plan. It's worth confirming that the channels you care about are included before you build your workflow around them.
Shared notes, so context follows the customer
When a caller phones on Monday and messages on WhatsApp on Thursday, you don't want your team starting from scratch. In AlmaTalk, threads carry shared notes your staff can read and add to, so the history travels with the customer rather than living in one person's head.
After each phone call, the AI also produces a post-call summary, and can capture caller details — name, topics discussed, follow-ups — from the transcript once the call ends. That summary sits alongside the conversation, so whoever picks it up next has the gist without listening to a recording.
Per-thread assignment
Not every message should go to the same person. AlmaTalk lets you assign conversations per thread, so a booking question can go to reception and a technical question can go to the specialist. Combined with the shared notes, this keeps handoffs clean: the next person sees what's already been said and what's outstanding.
When a human takes over
The AI handles the routine, but some conversations need a person. This is where AlmaTalk draws a deliberate — and honest — line.
A staff member can take over a live website-chat or WhatsApp conversation. If the AI is mid-thread and someone on your team wants to step in, they can, and the customer keeps talking to a real person from that point on. Text-based channels make this seamless: the medium is asynchronous enough that a handoff feels natural.
A live phone call is different. A person cannot silently seize an in-progress voice call — that's not how the telephony works, and we won't pretend otherwise. What the AI can do on a call is escalate or transfer to a person, passing along context so the customer isn't asked to repeat everything. So the honest summary is: live takeover for website chat and WhatsApp; warm transfer for phone.
What the AI does before a human is needed
The takeover matters less if the AI resolves the conversation on its own — which, for a lot of everyday requests, it does.
- Books appointments during the call and adds them to Google Calendar, Outlook or Cal.com, respecting your working hours. (One caveat worth knowing: an event added to your external calendar in the last little while may not be seen instantly, since inbound calendar sync polls periodically rather than in real time.)
- Takes a message to a named staff member from your own directory — and only to addresses in that directory, which is a deliberate security choice, not a limitation to work around.
- Confirms, cancels or reschedules an existing appointment by phone after verifying the caller's number.
- Recognizes returning callers by their number and can greet them by name in the language they used last time.
- Tags each call against your own service categories and urgency levels, after the call, from the transcript.
Several of these are opt-in and off by default — language and voice detection, auto-assigning service types to the right person, and so on — so you can turn on what fits and leave the rest alone. They can be configured; they aren't forced on you.
Getting started, and how billing works
You can have AlmaTalk build a draft agent from your own website, then a human on your team reviews it before it goes live — so nothing answers customers until you've read what it will say.
On cost, here's the straight version. Billing has two parts: a monthly or yearly plan that includes a set credit allowance, plus usage metered in credits (call time is measured in seconds, with a one-credit minimum per call). Paid plans have overage enabled by default, so usage beyond your allowance is billed as it happens. Everything is in Canadian dollars, and there's no setup fee. It's not a flat all-you-can-use plan — but it's predictable once you know your volume.
The takeaway
If your problem is "customers reach us five different ways and messages fall through the cracks," an omnichannel inbox with shared notes and per-thread assignment is the fix — one place to see everything, one history per customer, and a clear rule for when a human steps in. The honest boundaries (live takeover for text channels, warm transfer for phone, channels by plan, sync that polls rather than pings) are the kind of thing you'd rather know before you sign up than discover after.