AI agents & automation
Multi-step workflows that carry out repetitive work instead of your team: reading an enquiry, pulling the relevant records, drafting the response, filing the result. Not a chatbot — a process that runs whether or not anyone is watching it.
- KD 1,450
- build, once
- KD 145
- a month to run
- 1 process
- to start, measured
What this actually replaces
The work that never appears in anyone’s job description but consumes their week. Copying figures between a supplier’s email and your accounting system. Checking whether an item is in stock before answering. Producing the same monthly report from the same four spreadsheets. Sorting fifty enquiries into three piles so the right person sees them.
These tasks share a shape: they happen often, they follow rules a person could write down, and nobody enjoys them. That shape is what makes them automatable. Judgement calls, exceptions and anything with legal weight are not on this list, and we will say so rather than sell you an agent that guesses.
How we build it
We start from the process as it is actually performed, not as the manual describes it. That usually means sitting with the person who does it and watching, because the real procedure always contains steps nobody documented.
Then one workflow, end to end, on your real data. Every step is logged, so when something goes wrong you can see which step produced the wrong output rather than guessing at a black box. Anything the system is not confident about is handed to a person instead of being invented — an agent that escalates its uncertain cases is worth more than one that answers everything.
We build against your worst inputs from the start: the scanned page, the message written half in Arabic and half in English, the supplier whose name is spelled three ways. Systems that only work on clean input do not survive contact with a real business.
What it runs on
Whatever you already use. Email, WhatsApp, a shared drive, an accounting package, a spreadsheet somebody maintains by hand. Moving you to new systems is a separate decision and should never be a precondition for automating a process.
Where we can, everything runs under your own accounts and credentials, so ending the engagement does not switch anything off.
Where it goes wrong
Three failures, in order of how often we see them. Automating a process nobody measured, so there is no way to tell whether it helped. Automating a broken process, which produces the same bad result faster. And building something with no owner inside the business, which quietly stops being used within a month.
All three are avoidable, and all three are cheaper to catch in the first week than the sixth month.
What you get
The workflow running in production, its documentation, and training for the people who will operate it. A measurement of what it saved against the baseline we agreed before starting. You own the code and the accounts.
What it costs
Two parts, kept separate so you can see exactly what you are paying for.
Build — KD 1,450, once. One process automated end to end: the workflow running in production, the baseline measured before we start, documentation, and training for whoever runs it. Paid in advance, before work starts.
Running — KD 145 a month. Monitoring, the fixes that real inputs force, and the small changes a live workflow always needs. A twelve-month agreement billed quarterly at KD 435 every three months, each quarter paid in advance, renewing automatically.
The monthly part is optional and some clients run the workflow themselves after handover — you own the code and the accounts either way. We say so plainly because an agent nobody is watching fails quietly, and that is worse than not automating at all.
A second or third process, once the first is proven, is quoted on its own. They are rarely the same size.
Most of this work starts on a single process — fixed scope, fixed fee, and a number at the end that tells you whether to scale it. Tell us which task wastes the most time and we will tell you honestly whether it is worth automating at all.