Glossary

    The 31 terms that actually come up when you automate a front desk in the Gulf — what each one means, and why it matters operationally rather than as jargon.

    Written to be useful when you are comparing vendors: several of these are the terms a sales conversation leans on without defining. Where a term has a page of its own, the definition links to it.

    WhatsApp and messaging

    WhatsApp Business API

    Meta's programmatic interface that lets software send and receive WhatsApp messages on a business number. It is not the WhatsApp Business app: it has no phone interface, requires an approved business account, and charges per conversation. CARE runs on it, which is why a business number already tied to a personal WhatsApp account has to be migrated first. Read more →

    Meta Tech Provider

    A company with its own Meta app that builds on Meta's business platforms, including the WhatsApp Business Platform. Thamra Group is one, which is how CARE runs on the official API rather than an unofficial workaround. Worth knowing what it does not mean: it is not Meta endorsing or certifying any particular product, and it is a different thing from the badged Meta Business Partner programme. Ask a vendor which of the two they hold. Read more →

    Meta-approved template

    A message whose exact wording Meta has reviewed in advance, required for any message a business sends outside an open 24-hour conversation window. Reminders, recall and follow-ups all travel as templates. You cannot improvise the text, which is why a reminder reads consistently and why a new campaign needs approval before it can go out. Read more →

    24-hour window

    The period after a customer's message during which a business may reply freely, in its own words. Once it closes, only an approved template can be sent. It is the single rule that shapes most WhatsApp automation: answering inside the window is unrestricted, reaching out after it is not.

    Opt-in

    A customer's explicit permission to receive messages from a business on WhatsApp. Required before any reminder, recall or promotional message. Without it, sending is both a policy violation and the fastest route to having a number blocked, so CARE only messages customers who have given it.

    Voice note transcription

    Converting a recorded WhatsApp voice message into text the system can act on. It matters disproportionately in the Gulf, where customers routinely record a request instead of typing it, and it is harder than standard speech-to-text because the recording is in dialect rather than Modern Standard Arabic. Read more →

    Arabizi

    Arabic written in Latin letters, often with digits standing in for sounds that have no Latin equivalent — "3" for ع, "7" for ح. Common in Gulf messaging, especially among younger customers. A system that only parses Arabic script silently fails on it, so handling Arabizi is a practical requirement rather than a nicety.

    Code-switching

    Moving between two languages inside one conversation, or even one sentence — an Arabic message with English service names in it, or a customer who opens in English and continues in Khaleeji. Normal in Qatar's mixed customer base, and a common failure point for systems that pick a language once and commit to it.

    Arabic and dialect

    Khaleeji Arabic

    The Arabic actually spoken across the six GCC states — Qatar, Saudi Arabia, the UAE, Kuwait, Bahrain and Oman. It differs from Modern Standard Arabic in vocabulary, grammar and register, and differs between states: a Qatari writes «أبغى» where an Emirati writes «أبا» and a Saudi writes «ودّي», all meaning "I want". Read more →

    Modern Standard Arabic (MSA)

    The formal, written register used in news, official documents and schooling across the Arab world. Nobody speaks it at home. A reply composed in MSA is grammatically correct and reads as foreign to a Gulf customer, which is why MSA output is the clearest signal that a chatbot is an English-first system translating on the way out. Read more →

    Translation layer

    An architecture where a system reasons in English and machine-translates its output into Arabic at the last step. It produces fluent MSA and loses dialect, idiom and the words customers actually use for services. The giveaway is a reply that is correct, formal, and reads nothing like how anyone in the Gulf writes. Read more →

    Dialect detection

    Identifying which variety a customer is writing in from their own message, then replying in the same one for the rest of the conversation. The practical test is not whether a system can produce Qatari phrasing on request, but whether it recognises a Kuwaiti opening message without being told. Read more →

    Tashkeel (diacritics)

    The vowel marks written above and below Arabic letters. Usually omitted in everyday writing and common in formal text. They matter technically because the same word with and without them is a different string, so naive text matching treats «تدخّل» and «تدخل» as unrelated.

    Front-desk operations

    No-show

    A booked appointment the customer neither attends nor cancels, so the slot is lost with no chance to refill it. Distinct from a late cancellation, which at least leaves time to offer the slot to someone else. No-shows are the single largest recoverable loss in an appointment-led business. Read more →

    No-show risk score

    An estimate of how likely a specific upcoming appointment is to be missed, from signals like whether the customer confirmed, their past attendance, and the time of day. Its purpose is to decide where to spend attention: a high-risk slot gets a reminder, a very-high-risk one gets offered to the waitlist. Read more →

    Recall

    Contacting a past customer when a service is due again — a six-month dental check, a colour appointment six weeks on, an annual physical. It is distinct from promotion: the customer has already bought, and the message is a reminder of an interval rather than an offer. In most appointment businesses it is the largest unworked revenue line. Read more →

    Handover

    Passing a live conversation from the AI to a member of staff, with the thread history intact so the customer does not repeat themselves. It happens either because the customer asks for a person or because the question falls outside what the assistant was given. A system without clean handover eventually traps someone in a loop.

    Escalation

    Interrupting the normal flow because something needs a human now — a described emergency, an angry customer, a complaint. It differs from handover in urgency: handover moves a conversation along, escalation stops it and alerts someone. Escalation triggers are an operational convenience, not a safety system.

    Go-live

    The moment the assistant starts answering real customers on your number. It is a specific, dated event rather than a gradual rollout, because it is what the refund window is measured from. It follows setup and manual activation, so it does not happen the moment you finish uploading documents. Read more →

    Knowledge base

    The set of documents an assistant is allowed to answer from: price list, service menu, opening hours, staff roster, booking rules, common questions. It is the boundary of what the system knows. A question outside it should produce a handover, not an invented answer. Read more →

    Booking buffer

    Padding deliberately left between appointments for cleaning, turnaround or overrun. It is a booking rule rather than a service duration, so a system that only knows how long a service takes will overbook a day that looked free.

    Service duration

    How long a specific service actually occupies a chair, room or clinician. It is why bookings have to be service-aware: a thirty-minute cut and a two-hour colour cannot be offered the same slot, and a system that treats every appointment as one length will promise times it cannot keep.

    Waitlist

    Customers who wanted a slot that was already taken and agreed to be contacted if it frees up. It converts a cancellation from a loss into a filled appointment, and it is the difference between "we're fully booked" closing a conversation and opening one.

    Owner dashboard

    The view where a business owner reads every conversation as it happens, plus a daily summary of bookings, cancellations, what customers asked about and anything needing attention. Its purpose is supervision: an assistant nobody can audit is an assistant nobody should trust.

    Staff takeover

    A member of staff stepping into a conversation the assistant is handling and taking control of it, at any point and without ending the thread. Distinct from handover in who initiates: handover is the assistant passing the conversation up, takeover is a person reaching in.

    Commercial and legal

    Annual licence

    The commercial model CARE is sold on: one payment covering twelve months of use, renewed by a decision each year rather than by an automatic charge. There is no monthly bill and no card on file. It is not a perpetual licence — when the term ends without renewal, the assistant stops answering. Read more →

    Manual activation

    Switching the assistant on by hand once a bank transfer is confirmed, rather than automatically on payment. It is a consequence of taking bank transfer instead of cards, and it means go-live is scheduled rather than instant — which also rules out any self-serve trial. Read more →

    Refund window

    The period after go-live during which a customer can ask for most of their payment back. For CARE it is thirty days, returning 85% and retaining 15% against onboarding work already done. On a renewal the same window runs from the renewal date rather than from the original go-live. Read more →

    Data residency

    Which country's infrastructure data is physically stored on. Worth asking any vendor directly, because "regional" on a website often describes where the sales office sits rather than where the servers do. CARE runs on third-party cloud and AI providers that include infrastructure outside the GCC, and commits instead to export and permanent deletion on request. Read more →

    PDPPL

    Qatar's Personal Data Privacy Protection Law, Law No. 13 of 2016, the country's general data protection statute. Note what a vendor claims about it: stating that the law exists and applies is ordinary; asserting certified compliance with it is a specific claim that should come with evidence. Read more →

    Uptime commitment

    A contractual promise about what fraction of the time a service will be available, usually with compensation if it is missed. Distinct from a response-time promise, which is about support rather than availability. CARE offers neither, and says so rather than publishing a number it cannot hold to. Read more →

    Still unclear on something?

    The 32-question FAQ covers the same ground as answers rather than definitions, and the setup guide shows where each of these appears in practice.