Capabilities & Operations Dashboard#
Every feature that changes how the AI behaves or how your business is run has an on/off switch. Nothing new turns itself on and changes your customers' experience without you choosing it first.
The Capabilities panel#
Open Business Profile → Settings → Capabilities to see every switch in one place, grouped by what it controls:
- AI Agent capabilities — behaviors the agent itself performs (sending attachments, product cards, quick replies, placing orders, pausing when a human takes over, message batching, answer caching, and the newer additions below).
- Business policies — rules that apply across the whole business regardless of which agent is attached (same-day reorder warnings, order change cutoff times, comment-to-conversation conversion).
- Integrations — third-party connections (WhatsApp group broadcast, and everything in Integrations).
- Reliability — safety-net behaviors that only add resilience and never change what customers see when nothing goes wrong (LLM retry, complaint handling).
Every switch here maps to one specific behavior described somewhere in these docs — if you're not sure what a switch does, its label links back to the relevant page.
Newer capabilities#
A few capabilities were added after the original launch and are worth calling out specifically, since they're not part of every agent's default setup:
- Image → Product Matching — a customer sends a photo or screenshot of a product; the AI describes what it sees, searches your catalog, and replies with the matching product (or a configurable "sorry, not found" message if nothing matches closely enough). Tune how strict the match has to be — exact, strict, balanced, or loose — under the agent's settings. Off by default; turn it on once your product catalog has enough detail (titles/descriptions) for the AI to match against.
- Collection / Carousel Browsing — when a customer asks "কি কি আছে?" or similar, the agent can send a full product carousel for a category instead of describing items one at a time.
- Complaint logging — when a customer reports something wrong (missing, damaged, wrong item, late delivery), the agent logs it with a tracking ID (
CMP-000123) and notifies you, instead of trying to resolve it itself. On by default. - Payment mention flagging — if a customer mentions having paid (full or advance), the AI never confirms the order as paid on its own — it always flags it for you to verify manually first. This is a safety behavior and doesn't have an off switch.
- Same-day reorder warning — if a customer already has a confirmed order today, the agent asks before quietly placing a second one, instead of assuming it's intentional.
Operations dashboard#
Open the Operations tab on any Business Profile for a single, action-oriented view of everything that needs your attention:
- Error / Retry Queue — background tasks (AI replies, bulk image sends) that failed and are retrying automatically, or need a manual "Retry Now" / "Dismiss".
- Complaints — every logged complaint with its tracking ID, type, and status; mark resolved once handled.
- Payment Verification — customers who mentioned a payment, waiting for you to confirm before the order counts as paid.
- Escalated / Human Handoff — conversations the AI paused itself on because it needed a human.
- Image Match Tuning — recent photo/screenshot match attempts, useful for deciding whether to loosen or tighten the Image Matching Sensitivity.
Nothing on this page is a new data source — it's all built from things already happening elsewhere (orders, conversations, complaints) surfaced in one place so you don't have to go looking for them.
Why background tasks retry instead of just failing#
Some actions — sending a batch of product images — can hit a temporary failure (a slow API, a dropped connection). Instead of giving up immediately, MexvoSocial AI logs the failed attempt and retries it automatically with a short backoff, escalating to you by email only if every retry fails. You'll only ever hear about it if it truly needs your attention.