One automated business process in Dubai costs AED 6,000 to 12,000 to build and AED 1,200 to 2,000 a month to run and monitor. A narrow flow that touches two systems and never moves money can land at AED 2,500 to 6,000. Anything that issues payment links, generates documents, or connects a system without a proper API sits at AED 12,000 to 20,000 and above. Our own productised version of a single process starts at AED 6,000 to set up plus AED 1,200 a month, and that number is published because most of this market keeps it behind a "request a quote" form.
I run the stack we sell. Our studio's enquiries are handled by a WhatsApp agent that qualifies, quotes, books the slot and sends the payment link, then writes the deal into the CRM with its source tag and raises the invoice when money lands. The figures below come from building that, breaking it, and rebuilding it, plus the client builds around it for service businesses in Dubai.
What counts as one process
A single automated process has one trigger, one outcome, and the steps in between, running across whatever systems it needs to touch. An enquiry arrives on WhatsApp, a deal is created in the CRM carrying its source tag, an owner is assigned, and a first reply goes out inside a minute. One trigger, one end state, four systems in the chain. That is one process, and it is the unit you should be quoting.
"Automate our sales department" is not a process. Neither is "automate operations" or "digitise the customer journey". Those are programs made of eight or twelve processes, and a vendor who quotes a program as if it were a process is either guessing or planning to send change orders later.
The test takes ten seconds. Write the flow on one line: trigger, steps, outcome. If you need the word "and" between two different end states, you have two processes. If you cannot name the trigger, you have a wish.
Programs are priced on a different scale entirely. A CRM rollout for a small Dubai service business runs AED 15,000 to 45,000 over three to six weeks and climbs past AED 150,000 for a fifty-user deployment, which is covered in our breakdown of CRM implementation cost and timeline. Automating one process inside an existing CRM is a fraction of that.
What the money actually buys
Take a typical AED 8,000 build and open it up. Roughly AED 1,500 goes to discovery and the spec: mapping the trigger, every step, every outcome, and what should happen when a step fails. This is the line owners try to cut, and cutting it is how a build gets rewritten in month two.
Around AED 3,000 goes to integrations. Each connection means authentication, field mapping, rate limits and a round of testing. The first connection is cheap. The third and fourth are where the money goes, because every added system introduces a new pair of failure modes with the systems already in the chain.
About AED 2,000 goes to exception handling, which is the line every cheap quote leaves out. The same customer messaging from two numbers. A phone field arriving as 050 in one system and +971 50 in another. A client replying "yes please" to a quote sent nine weeks ago. The happy path is a day of work. The exceptions are the build.
The last AED 1,500 splits between testing on real data and handover: a written flow document, access in your own accounts, and a kill switch. Then AED 1,200 a month keeps it alive, which is a separate discussion further down.
Why two quotes for the same automation differ by three times
Five things move the number, and none of them are visible in the brief you send out.
The count of systems in the chain matters more than the count of steps. Two systems is a connection. Five systems is a small distributed system with five ways to disagree about the same fact.
Whether those systems have a real API is the hard gate. A cloud tool with documented endpoints and webhooks is buildable. A desktop package that only imports and exports files can be faked with scheduled uploads, and the price doubles while the reliability halves.
Edge case volume is the third. A booking flow for a studio with one location and fixed hours is clean. A clinic with five practitioners, split shifts and insurance rules carries a hundred branches, and each branch is a decision someone has to specify before it can be built.
The accuracy bar is fourth. When the output is an internal report, ninety-five percent right is fine. When the output is a payment link with an amount on it, every path needs a human gate or a hard validation, and that discipline costs real hours.
Language is fifth and the most underestimated. Running the same process in English and Arabic means a second set of message templates, name and number parsing that handles both scripts, and a routing rule for which language a customer gets, on a channel where people switch mid-conversation.
What one process costs in Dubai, by scope
Bands below are what we see quoted around Dubai for a single process, and what we charge for our own.
- Two systems, no money moving, for example a web form writing a tagged record into the CRM: AED 2,500 to 6,000 to build, AED 300 to 800 a month.
- A standard service-business process across three or four systems with follow-ups and source tagging: AED 6,000 to 12,000 to build, AED 1,200 to 2,000 a month.
- Money or documents move, for example quote to payment link to invoice to reconciliation: AED 12,000 to 20,000 to build, AED 1,500 to 2,500 a month.
- A legacy system with no API, file drops, or screen-level work anywhere in the chain: AED 15,000 and up, with a monthly figure that reflects how often it breaks.
Our productised single process is priced at the entry of the second band. The scope, the failure alerting and what the monthly fee covers are laid out on the process automation page, including the parts we refuse to leave as surprises.
The payback sum, run honestly
Two lines of arithmetic. Nothing else is needed at this stage.
Loaded hourly rate = monthly salary divided by 176 working hours. A coordinator on AED 6,000 to 9,000 costs AED 34 to 51 an hour. An operations manager on AED 15,000 costs about AED 85.
Payback in months = build cost divided by (monthly hours saved times hourly rate, plus monthly leak recovered, minus the monthly run fee).
Here is the sum on a real shape of business. A coordinator spends ninety minutes a day copying leads between WhatsApp, a spreadsheet and the CRM, and chasing quotes that went quiet. That is roughly 30 hours a month, or AED 1,350 at AED 45 an hour, or AED 16,200 a year. The automation costs AED 8,000 to build plus AED 14,400 a year to run, so year one is AED 22,400 and every year after is AED 14,400.
On labour alone that is minus AED 6,200 in year one and plus AED 1,800 in year two. I would rather put that on the page than sell against it. Automating a single process almost never pays back inside a year on hours saved, and any vendor whose ROI slide says otherwise is counting hours nobody was actually spending.
The second term is where it pays. Our working figure for a typical Dubai SMB is around AED 4,100 a month lost to unanswered and slowly handled leads, which is money that never appears in a payroll line. Recover a third of that and the sum reads AED 16,200 in labour plus AED 16,400 in recovered leak against AED 22,400 of first-year cost, which turns minus 6,200 into plus 10,200. That is the honest version: the labour saving pays for the run fee, and the leak pays for the build.
Run the sum on your own hours before anyone quotes you. If you want the hours and the leak measured rather than estimated, that is what a growth audit at AED 3,000 produces, and the process map it delivers is yours whoever builds the thing afterwards.
When automation does not pay
Skip it when the process changes every month. A spec that gets rewritten in week six costs more in rebuilds than it ever saves, and seasonal service businesses do this to themselves constantly.
Skip it when the volume is small. Under about two hours of manual work a month, the labour side is AED 90 and there is nothing to recover. Fix the process on paper first and revisit at volume.
Garbage input is the third case. Handwritten notes, voice messages, WhatsApp photos of a delivery slip. Getting that into a structure a machine can read costs more than the automation saves, and the honest first project is fixing how the data arrives.
Then there is the bottleneck that is a human decision rather than a data transfer. If a quote waits three days because the owner has not decided the price, no integration on earth fixes that.
Last, skip it when the underlying process is broken. Automating a broken process gives you a broken process running at speed, at 3am, with nobody watching. The ordering question is covered separately in our guide on which process to automate first.
Why automations break, and what monitoring covers
Nothing here is exotic. A vendor deprecates an API version. Somebody renames a field on a form. A date arrives as 03/04 instead of 2026-04-03. An OAuth token expires quietly on a Thursday. A staff member adds a new dropdown value in the CRM that no rule knows about. Each of these is a one-line change that stops a flow, and none of them announce themselves.
The dangerous failure is the silent one. On our own stack we found two sync bugs between payment confirmation and CRM stage, and until proper alerting existed, roughly AED 15,000 of paid bookings had been sitting in the CRM marked unpaid for weeks. The dashboard was green the entire time. That is the argument for the monthly fee in one sentence.
The second thing that catches people is memory. A client goes quiet, comes back nine weeks later and writes "ok let's do it". If conversation state has expired, the system re-qualifies them from zero and the customer feels it immediately. Deciding how long a deal stays warm and what the system remembers is a design decision, and it belongs in the spec rather than in a support ticket.
So a monthly fee that means anything covers run-failure alerts going to a human, handling API and format changes as they land, a reconciliation check that compares what the automation says happened against what the systems actually recorded, tuning as volumes move, and a kill switch you can pull yourself. A flow built without alerting only looks cheaper, right up until the week it fails quietly.
If the process you want to automate is the first reply and qualification on WhatsApp, that one is productised as a WhatsApp AI sales agent rather than a custom build, which is usually faster and cheaper than wiring it from parts.
