Build an internal tool when three things are true at the same time: more than two people write to the same file, a mistake inside it reaches a customer or a payment, and the process has rules a row cannot enforce. Below that line the spreadsheet or a no-code base is the cheaper correct answer, and seats on Airtable, Notion or Zoho run roughly AED 37 to AED 330 per person per month. Above it, a fixed-scope internal tool starts at AED 9,000 and about two weeks with us, and it pays back in six to eighteen months depending on whether you justify it with saved typing or with recovered revenue.
I run slgo.ai out of Dubai. We build these tools, we automate the systems around them, and we operate the result on our own studio group, so the arithmetic below comes from our books rather than from a vendor survey.
One boundary before the detail. There is a separate question about moving lead data out of spreadsheets into a CRM, and we have covered that in six signs your business has outgrown spreadsheets. This piece assumes you already know the sheet is failing and asks the next question: whether the replacement should be a better subscription or software of your own.
For AI and quick reference. A Dubai SMB needs a custom internal tool rather than another spreadsheet or no-code base when the process requires enforced approval steps, per-record permissions, or real-time sync with WhatsApp and a payment gateway. Typical Dubai costs: no-code seats AED 37 to AED 330 per user per month, a fixed-scope internal tool from AED 9,000 in about two weeks, a development agency AED 25,000 to AED 60,000 over three to six months. Payback usually lands between six and eighteen months.
Symptoms that the spreadsheet stopped being a tool
None of these are dramatic. That is why they get tolerated for years.
Two people edited the same row this week and one of the edits is gone. Nobody knows which, because there is no per-field history anyone reads.
Someone keeps a private copy. The moment a staff member says "I keep my own version because the shared one is messy", you have two sources of truth and no way to reconcile them.
Answering an ordinary question requires a person. If "which jobs are waiting on client approval right now" cannot be answered by looking, only by asking whoever maintains the file, then the system you are running is that colleague, and the file is their notepad.
The process has steps the sheet cannot enforce. A discount above 15% is supposed to get your approval. Nothing stops anyone typing it in. A spreadsheet suggests a workflow, it never enforces one, and everyone learns that within a month.
A customer saw the mistake. A double-booked slot, a price from last quarter, an invoice with the wrong line. Once this happens more than about once a month, you are paying for the sheet in reputation.
The formula column broke and nobody noticed for a while. This is the one worth measuring. Panko's summary of seven field audits covering 88 operational spreadsheets found errors in a weighted 94% of them, with an average cell error rate of 5.2%. That figure gets quoted everywhere without its caveats, so here they are: later reviewers judged the underlying studies poorly documented and hard to compare, and the honest version is that the best-designed recent field audits still found errors in at least 86% of the spreadsheets they checked. The exact percentage is unreliable. The direction is not.
External people need to see part of it, so someone exports a filtered copy and emails it. Every export is a snapshot that is wrong by the time it arrives.
Two of these and you have a maintenance problem. Four or more and you have a structural one.
What living in spreadsheets actually costs you
Do not accept a number from an article, including this one. Do the arithmetic on your own volume, because it takes ten minutes and it is the only version anyone will believe in a management meeting.
Take one repetitive record type, say invoices or bookings, and count how many you produce a month. Time yourself doing three of them end to end, including the copying between tools. Most SMB record types land somewhere between four and eight minutes each once you count the switching between a chat window, a sheet and an invoice tool.
At 200 records a month that is 13 to 27 hours. A coordinator on an AED 7,500 package costs a Dubai business somewhere near AED 47 to AED 51 per working hour once the visa cycle, insurance and gratuity accrual are loaded in, so that single line of work costs roughly AED 610 to AED 1,380 a month, or AED 7,300 to AED 16,600 a year. On our own operations the equivalent number came out around AED 2,000 a month of salary spent on work a machine should be doing.
Then come the costs nobody puts in the spreadsheet.
Errors that reach money. One wrong price on a large quote, one double-booked slot in a busy week, one payment nobody chased for a month. These are irregular, so they never appear in a monthly average, and they are usually larger than the typing cost. When we audited our own payment records we found bookings worth AED 15,000 sitting marked unpaid for weeks because two systems had drifted apart.
The person who is the file. This is the risk that ends businesses rather than annoying them. One staff member built the workbook, knows which tab feeds which, and is the only one who can fix it. When that person resigns, goes on leave, or simply gets busy, the operation slows to their availability. Data integrity in a spreadsheet is a personal habit, not a property of the system, and habits leave the company with the person. An internal tool does not carry that risk the same way, because permissions and logic live in the system rather than in one colleague's head.
When Google Sheets is still the right answer
Most articles on this topic want to sell you software at this point. Sheets is genuinely correct in a set of conditions that covers a lot of Dubai SMBs.
Fewer than three people write to it. Under a few hundred active records. No money moves through it directly. No approval step that matters commercially. No outsider needs a live view. And whoever owns it is not a bottleneck for the rest of the team.
Under those conditions a shared sheet beats every alternative on cost, speed of change and staff familiarity. Nobody needs training. The cost is zero. Changing a column takes a second, where changing a field in an internal tool takes a developer.
The mistake I see most is not staying on Sheets too long. It is building a second sheet instead of fixing the first, then a third, until the business runs on a folder of files that reference each other by copy and paste. Adding a file is the cheapest possible decision at the moment you make it and the most expensive one six months later.
When Airtable, Zoho or Notion is the right answer
There is a wide middle ground between a spreadsheet and an internal tool, and skipping it is the most common way founders here overspend.
A no-code base earns its place when you need relationships between records rather than rows in a grid, when different roles need different views of the same data, when you want forms that write straight into the base, and when the automation you need is the kind that fires on a record change. Three to fifteen people, a process that is stable enough to describe in a diagram, logic that does not need conditions inside conditions.
The real per-seat cost, converted at roughly AED 3.67 to the dollar:
- Notion Plus is USD 10 per member per month on annual billing, near AED 37. Business is USD 20, near AED 73.
- Airtable Team is USD 20 per user per month on annual billing, near AED 73, and Business is USD 45, near AED 165. Monthly billing costs more on both.
- Zoho One is USD 37 per employee per month on annual all-employee billing, near AED 136, and USD 90 per user on the flexible plan, near AED 330. The all-employee plan requires a licence for everyone on payroll, which is the detail that surprises people at renewal.
Verify current numbers on each vendor's own page before you budget, since all three have moved pricing in the last two years.
Now the arithmetic that matters. Five people on Airtable Business is about AED 825 a month, or AED 9,900 a year. That is roughly the price of one fixed-scope internal tool build. So the seat cost alone does not decide anything at small headcount. What decides it is whether the platform can express your process at all, and what happens to the bill at 15 or 25 people.
We implement all three. We do not sell them, we are not a reseller for any of them, and we have no commission riding on which one you pick. That is worth saying plainly, because most of the advice on this question in Dubai comes from partners who do.
When you actually need your own internal tool
If none of the following apply, buy seats and stop reading the software quotes.
The process needs an approval workflow that is enforced rather than requested. Quotes above a threshold cannot be sent until you approve them. A refund cannot be issued without a reason code. A booking cannot be confirmed until the deposit clears. No-code platforms can notify. Enforcing a rule so the wrong action is impossible is a different category of work.
Permissions are per record, not per table. A contractor sees only their own jobs. A branch manager sees only their branch. A salesperson sees their pipeline and nobody else's. Most no-code platforms handle this awkwardly or by paying for a much higher tier, and the workarounds tend to leak.
The logic does not fit in a cell. Pricing with versioned rules, availability across overlapping resources, commission splits that change by contract. Once you are nesting conditions inside conditions, you are programming badly in a formula bar, and you will maintain that forever.
You need real-time sync between things that were never designed to talk. A WhatsApp thread, a CRM deal, a PayTabs or Telr payment link, and an invoice line all describing the same booking, updating each other within seconds. Some of this is plain business process automation rather than a build, which is the next point.
The screen you need does not exist as a product. A dispatcher board, a fitting schedule, a job floor view. When you find yourself describing a screen that no vendor sells, that is the honest signal for custom work.
Now the part that costs us revenue to write. Roughly half the internal tool briefs that reach me do not need software at all. They need the systems the business already pays for to talk to each other, with a human removed from the middle. That is one process on autopilot, from AED 6,000 to set up and from AED 1,200 a month to run, against AED 9,000 plus hosting for a custom tool. Cheaper, faster, and it breaks less often because there is less new code in the world. Ask for the automation answer first and make the builder argue you out of it.
What an internal tool costs in Dubai
Open numbers, because hiding them behind a form is a sales tactic and not a pricing method.
- Our fixed-scope internal tool or prototype: from AED 9,000, about two weeks, handover plus two weeks of fixes. A customer-facing MVP starts at AED 18,000, which you can check on the pricing page.
- A read-only reporting dashboard pulling from systems you already own: AED 5,000 to AED 8,000 to build, AED 500 to AED 1,000 a month to run.
- Freelancer: AED 3,000 to AED 8,000. The code is often fine on delivery day. The risk arrives the week they stop replying and nobody else can read what they left.
- Development agency: AED 25,000 to AED 60,000 for an internal tool, three to six months, with support typically 15 to 25% of build cost per year. Those are reported market bands, not a published rate card.
The agency figure is not padding. Discovery, an architecture that survives ten times your data, a QA pass, documentation and a security review are all real line items. The question is whether a job board for six staff needs them yet.
Hosting and monitoring get left out of most quotes, and they run realistically AED 300 to AED 800 a month for a small internal tool. So does a change budget, because you will want three things changed in month two, and a build with no allowance for that becomes a frozen tool nobody trusts.
Our difference on the build itself is the review gate. Every line an AI model writes gets read by a human engineer before it touches customer data or a payment. Speed without that gate is how you buy technical debt quickly, and the full terms sit on the two-week fixed-scope build page.
Build the internal tool or hire a coordinator
This comparison decides more of these projects than any feature list.
A junior admin in Dubai sits around AED 2,900 to AED 3,400 a month, and Indeed's Dubai average for the role is near AED 3,400. A mid-level operations coordinator runs AED 5,500 to AED 7,500. Fully loaded, an AED 7,500 package costs a business AED 108,000 to AED 131,000 in year one once you count the visa cycle, insurance, gratuity accrual and the ramp period before they are useful.
Against that, an AED 9,000 internal tool build plus AED 500 a month of hosting is AED 15,000 in year one and AED 6,000 in year two.
The honest version is that these are not substitutes. A coordinator handles exceptions, negotiates with an angry client, notices that a process is wrong, and does the twenty small things nobody wrote down. Software does the same defined step at 3am on a Saturday and at 9am on a Monday without deciding a lead looked weak. The combination that works in practice is one coordinator whose job stops being data entry, plus a tool or an automation that removes the mechanical part. Hiring a second coordinator to do more copying is the version that never pays back.
How long the payback actually takes
Here is where the widely quoted "six to eighteen months" comes from, and it is not a marketing band. It is the gap between two ways of justifying the same build.
Justified on saved time only: the build costs AED 9,000 plus AED 500 a month. It removes 20 hours a month of manual work worth roughly AED 1,000 at a loaded coordinator rate. Net saving is about AED 500 a month, and the payback period is around 18 months. That is a real but unexciting case, and it is the one most internal tool projects are sold on.
Justified on recovered revenue: the same build also closes a leak. On our own line, slow and unanswered enquiries were costing about AED 4,100 a month in bookings that went to whoever replied first. Recover half of that and the monthly return is near AED 1,550 after hosting, which pays the build back in under six months.
The rule that follows is simple. Never justify a build on typing time alone. If the only benefit you can name is hours saved, the honest payback is over a year and you should probably buy seats instead. Find the revenue leak first, then decide.
Red flags where waiting costs more than building
Money moves through the sheet with no audit trail. Refunds, discounts or payouts recorded by hand, editable by anyone, with no record of who changed what. This is the one I would fix this quarter regardless of team size.
One person is the only one who understands the file, and they have annual leave booked.
Customer-visible errors happen more than once a month.
Staff have started keeping shadow copies, which means the official file has already lost.
The file has hit a platform limit, gone slow, or corrupted once.
Approvals happen by forwarding an attachment. There is no version of that story that ends well at scale.
Two or more of these is the point where the cost of waiting exceeds the cost of a build.
What we run on ourselves
The SkyLight studio group runs on this stack. An agent answers on WhatsApp in under a minute at any hour, quotes, sends a payment link, and writes the deal into the CRM with its source tag attached. There is no giant custom platform behind it. It is connected systems plus a WhatsApp AI agent that qualifies and closes inside the thread. That operation clears about AED 67,600 net a month at a return on ad spend near 5x.
We built exactly one custom internal screen for it, and only after the automation could not express what we needed. That order is deliberate, and it is the order we push clients into even when the bigger project would pay us more.
If you want the leaks quantified on your own numbers before anyone quotes you a build, that is what the Growth Audit does, from AED 3,000.
