Answer every customer and qualify every lead, and pay for the results you can count
The Front Door Agent answers customer questions and qualifies new enquiries on WhatsApp, email and web chat, in English and Arabic, connected to your order, booking or CRM system. Priced on resolved conversations and qualified leads, measured on your own data.
For businesses whose inbox is a queue
If five or more people spend their day answering the same questions on WhatsApp, email and chat, or if a handful of people field enquiries where one missed lead costs real money, this is for you. Online retailers, clinics, brokers and developers, schools and training centres, hotels and venues, and B2B firms with an inbound sales inbox. Below that size the cost case is thin, and we will say so in the first call.
- A written baseline of your inbox: volume by channel and intent, response times, what a conversation costs you today
- An agent that resolves the routine intents alone and hands the rest to a named person with the facts already gathered
- English and Arabic, tested on your real conversations before launch
- A shared dashboard that is the single record of what was resolved, escalated and reopened
- A per-outcome price with a monthly cap, billed only after the pilot on definitions you signed
- Your own team trained to update the knowledge base and read the numbers
Most AI projects cannot show a result. This one is built around the measurement.
Only 37% of companies attribute any profit impact to AI (McKinsey, August 2026), while 74% of UAE CEOs expect returns within one to three years (KPMG, November 2025). The gap is measurement, not technology. So the Front Door Agent starts with a baseline of your own numbers, runs a pilot with a holdout, and bills on definitions agreed in writing before anything goes live.
Five steps, from your baseline to a measured result
About two weeks to baseline and scope, three to four weeks to build, four weeks of pilot, then a ninety-day measured window.
- 1Weeks 1–2
Baseline and scope
We export eight to twelve weeks of your conversations and leads, classify them by intent, and record the numbers: volume by channel, median first response, share resolved without a person, staff cost per conversation, CSAT if you have it, lead contact and meeting rates. We agree in writing what counts as a resolution and a qualified lead, and which intents always go to a person.
OutcomeA signed baseline and a scope. If the numbers say the agent will not pay, we tell you here. - 2Weeks 3–6
Build
Knowledge base cleaned. Connections to your shop, CRM, helpdesk and booking tools. In-scope intents built with approval rules, hand-over to named owners, Arabic and English tested on your own past conversations, every action logged with its reason.
OutcomeAn agent that works on your systems, not a demo. - 3Weeks 7–10
Pilot with a holdout
Part of your traffic goes to the agent and a comparable part stays with your team, so the difference is the agent and not the season. Weekly report by intent: resolution rate, reopen rate, escalations, CSAT, response time.
OutcomeEvidence from your own inbox that a sceptic would accept. - 4Days 1–90 after pilot
Measured window
Full rollout on the in-scope intents. One shared dashboard is the record. Fifty conversations a week read by a person. Monthly review; new intents added only when the existing ones hold their numbers.
OutcomeA ninety-day result, reported with its limits, that you can hold us to. - 5Ongoing
Run and improve
Monitoring, the weekly quality sample, updates when your products or policies change, and your own team trained to maintain the knowledge base and read the dashboard.
OutcomeAn agent that keeps earning its fee, and a team that understands it.
How it is measured
A resolution is defined, not assumed
A conversation counts as resolved when the customer confirms the answer or asks for nothing more within 72 hours. Reopened conversations are not billed.
Cost per resolved conversation
Your baseline staff cost per conversation against the agent's cost per resolution, both on the dashboard.
Leads: contact, qualify, book
Contact within five minutes, qualified against your criteria, meetings booked into your calendar. Only what we control is counted.
Guardrails that can stop the rollout
Escalation rate, reopen rate, CSAT on AI conversations against human ones, complaints, and accuracy on the weekly sample.
A holdout, not a before-and-after
During the pilot a comparable share of traffic stays with your team, so seasonality and campaigns do not get credited to the agent.
Read by a person every week
Fifty conversations sampled and scored each week. The scores are on the dashboard next to the rates.
Fixed to build, per outcome to run, capped every month
Per-outcome rates sit between what you would pay HubSpot ($0.50 per resolved conversation) and Intercom ($0.99 per resolution) for their own agents, and the fixed parts sit inside published agency bands in the UAE. Model usage and WhatsApp template fees are passed through at cost and shown separately.
| Item | Price | What it covers |
|---|---|---|
| Baseline and scopeTwo weeks. Fixed. Credited against the build if you go ahead within 30 days. | AED 7,500 | Two weeks. Fixed. Credited against the build if you go ahead within 30 days. |
| BuildFixed price set in the scope, depending on channels and systems connected. | AED 25,000 – 45,000 | Fixed price set in the scope, depending on channels and systems connected. |
| Monthly floorHosting, monitoring, the weekly quality sample, and knowledge-base updates. | AED 2,500 | Hosting, monitoring, the weekly quality sample, and knowledge-base updates. |
| Per resolved conversationOn the signed definition, from the shared dashboard, after the pilot. Reopened conversations are credited. | AED 2 | On the signed definition, from the shared dashboard, after the pilot. Reopened conversations are credited. |
| Per qualified leadAgainst your written criteria. Meetings booked are reported, not charged. | AED 25 | Against your written criteria. Meetings booked are reported, not charged. |
| Monthly capSo your cost is predictable however busy the month. | Agreed in the scope | So your cost is predictable however busy the month. |
If you already pay for HubSpot's or Intercom's agent, we price the baseline, integration, intent design, quality sampling and training as the product, and their per-resolution fee passes straight through.
What the numbers look like for a Dubai online retailer
The company below is a composite built to show how the Front Door Agent is scoped, priced and measured. Volumes, staffing and results are assumptions, listed in the table and chosen to sit at or below published benchmarks. Real engagements start with a baseline measured on your own data, and results are published only once they have been measured.
A fashion and home retailer with 6,000 conversations a month
An online fashion and home retailer selling across the UAE through Shopify receives about 6,000 inbound conversations a month: roughly 60% on WhatsApp, 25% by email, 15% on web chat. Seven people answer them across two shifts. Most messages are about the same few things: where an order is, how to return or exchange, whether an item fits or is in stock, a delivery address change, a cash-on-delivery question, a complaint.
Median first response is two to three hours during the day; messages after 9pm wait until morning. Nobody knows what a conversation costs.
The assumptions, stated plainly
| Input | Value used | Basis |
|---|---|---|
| Inbound conversations per month | 6,000 | Assumption, chosen to represent a mid-sized UAE online retailer |
| Channel mix | WhatsApp 60%, email 25%, chat 15% | Assumption |
| Intent mix | Order status 35%, product and sizing 20%, returns and exchanges 15%, delivery changes 10%, payment and COD 8%, complaints 7%, other 5% | Assumption |
| Intents the agent handles alone | Order status, product and sizing, returns initiation, delivery changes (80% of volume). Payments, complaints and other always go to a person | Scope decision |
| People answering inbound | 7 | Assumption |
| Base salary, customer service representative, UAE | AED 3,285 a month | Verified: Indeed UAE, average of 2,300 reported salaries, updated 20 July 2026 |
| Loaded cost per person (salary, visa, insurance, allowances, overhead) | AED 6,500 a month | Assumption, roughly double base salary. Replaced with the client's own figure in a real baseline |
| Baseline staff cost per conversation | AED 7.58 | Arithmetic: 7 × 6,500 ÷ 6,000 |
| Agent's autonomous resolution rate by month three | 45% of all conversations | Assumption, set below Intercom's published average of 76% (vendor claim) and at the low end of named vendor cases of 45% to 65%. Sensitivity tested at 35% |
| Human productivity after rollout | Unchanged at about 860 conversations per person a month | Assumption, conservative: hand-overs arrive with the facts gathered |
| Model and infrastructure usage | AED 0.30 per conversation handled | Estimate, range AED 0.15 to 0.50 |
| WhatsApp cost for replies inside the 24-hour customer window | Free | Verified: Meta pricing page, rates effective 1 July 2026 |
| MornningStar fees | Baseline AED 7,500; build AED 35,000; monthly AED 2,500 plus AED 2 per resolved conversation | The pricing above, mid-range build |
What the ninety-day measurement would show under these assumptions
All figures in this table are estimates that follow from the assumptions above. None is a measured result.
| Measure | Baseline | Month three (estimate) |
|---|---|---|
| Conversations resolved by the agent | 0 | 2,700 (45% of 6,000) |
| Conversations reaching a person | 6,000 | 3,300, each with order and customer details attached |
| People needed for the human share | 7 | 4 (3,300 ÷ 860, rounded up) |
| Median first response, in-scope intents | 2 to 3 hours | Under one minute, all hours |
| Staff cost per conversation | AED 7.58 | AED 4.33 (4 × 6,500 ÷ 6,000) |
| MornningStar monthly fee | 0 | AED 7,900 (2,500 + 2,700 × 2) |
| Model usage | 0 | AED 1,800 (6,000 × 0.30) |
| Total monthly cost of the agent | 0 | AED 9,700 |
The money, and the condition attached
Three roles are no longer needed on the inbound queue. That is worth AED 19,500 a month in loaded cost, but only as cash if the retailer redeploys those people to work that was going undone, or does not backfill when someone leaves. Counting it in full, the monthly benefit is AED 19,500 against AED 9,700 of running cost, a net AED 9,800 a month, and the AED 42,500 of one-off fees is recovered in just over four months.
If only two roles are actually released, the net falls to AED 3,300 a month and payback stretches to about thirteen months. If the resolution rate lands at 35% instead of 45%, the agent resolves 2,100 conversations, five people are still needed, and the case rests on two roles, not three.
Those three lines are the point of the method. The client sees which assumption the return depends on before signing, and the ninety-day window measures that assumption rather than the brochure.
Sales from after-hours product questions answered instantly, which is a real effect for retailers but needs the pilot's holdout to measure. Fewer repeat contacts because answers are consistent. Any change in CSAT. Staff time freed for proactive work. These are measured in the pilot and reported; they are not in the arithmetic above because they cannot be assumed honestly.
Run the same arithmetic for a retailer with 2,400 conversations and three people: the agent resolves about 1,080, two people handle the rest, one role is released (AED 6,500 a month), and the running cost is about AED 5,400. Net AED 1,100 a month, and the one-off fees take more than three years to recover. That is why this service is aimed at businesses with five or more people on inbound, or with high-value enquiries where one extra qualified lead a week pays for the month. For a smaller retailer the honest advice is a lighter setup on HubSpot's or Intercom's own agent, with us doing the connection and the baseline, or one of our automation packages instead.
The inboxes it is built for
The mechanism is the same everywhere. It pays best where enquiry volume is high and each enquiry has a clear next step.
Online retail & D2C
Order status, returns and exchanges, sizing and stock questions, delivery changes, cash-on-delivery queries.
Clinics & dental practices
Appointment booking and changes, insurance and price questions, reminders, pre-visit instructions. Medical questions go to a person.
Real estate brokers & developers
Instant replies to portal and ad leads, qualification against budget and timing, viewings booked into the agent's calendar.
Schools, universities & training
Admissions enquiries, fee and schedule questions, document checklists, campus visit bookings.
Hotels, restaurants & venues
Reservations, group and event enquiries, opening hours and directions, special requests routed to the right team.
B2B services
Inbound sales inbox triaged and qualified, meetings booked, routine account questions answered from your systems.
Common questions
Is the worked example a real client?
What resolution rate should we expect?
What happens to the people who answer messages today?
Can it work in Arabic?
What will the agent never do alone?
What if we already use HubSpot or Intercom?
Work that is live, not slides
Client names are kept private at their request; the detail behind each project is something we walk you through on a call.
Ad tracking rebuilt so a beauty brand could sell across India
Two clean signals, one national market
Read case study →
A UAE apparel team analysed its own Shopify data and found its best sellers out of stock 39% of the year
141 days out of stock, found by the team itself
Read case study →
Orders on hold cleared before the customer notices
Holds cleared before the customer notices
Read case study →Tell us what slows your business down
Thirty minutes on a video call. We will tell you honestly whether AI can help, what it would take, and roughly what it would cost. No slides. No hard sell.