We build and run the boring layer. Company websites, the servers under them, the email and files a business lives in, and the locks on all of it. None of that is glamorous, and all of it is what stops working at the worst possible moment when nobody owns it.
So we own it. Every site we host is patched weekly after a staging run, backed up off-site every night, watched once a minute, and reported on every quarter. Every system we build comes with a written blueprint, and every tenant we run gets a runbook someone else could follow.
The number is agreed before the work starts, in writing, in riyals. It does not move unless the scope does.
Security is not a product you buy at the end. It is two-factor sign-in with no exception for the boss, admin rights on separate accounts, ports closed by default, backups that have actually been restored, and an alert that reaches a pager rather than an inbox.
Most of the systems we take over have none of that switched on. Switching it on is usually a week of settings and a written explanation, not a new platform. We would rather sell you the week.
When something does happen, there is a runbook, a person on call, and a written account afterwards of what changed so it does not happen twice.
AI comes third, on purpose. A model is worth building when it changes a number someone is accountable for — response time, cost per case, hours returned to a team — and when it can be measured against cases drawn from your own work rather than a public benchmark.
When that is true we build it properly: an evaluation set before the first line of code, a failure path before the feature list, and deployment inside your own network where the data has to stay. The founder has published peer-reviewed work on exactly this, and the same discipline applies to commissioned systems.
When it is not true, we say so. A scheduled report or a well-configured integration solves more business problems than a model does, and it costs a fraction to run.