Everything people ask before they start.

The Audit, always. Two weeks, fixed scope, ending in a build plan you could execute with anyone, including yourselves. We won’t quote a build before we understand what it’s worth.

One engineer and one operator with real decision rights, roughly a fifth of their time. Access to a representative slice of data, and someone who can say yes. That’s genuinely it.

Two weeks for the Audit, then typically six weeks to a system running on live work with one team. Weekly demos throughout, so nothing is a surprise at the end.

We say so, in writing, early. An Audit that concludes “fix the data pipeline first, then revisit” is a successful Audit. Deploying something you’ll switch off in a quarter is not.

Yes. We’re based in Dubai and most of our work is in the Gulf, but engagements run wherever your operation and your systems are.

You do. Source code, documentation and runbooks are yours, in your repositories, from the first commit. There is no proprietary runtime you have to keep licensing from us to operate your own system.

That’s the design goal. Handover includes runbooks, architecture decision records and paired on-call until your team is changing the system confidently without us in the channel.

Nothing breaks. The system runs in your environment on your infrastructure, and the people who know it are your own. Continuing with us should be a choice, not a dependency.

Yes. We deploy inside your cloud tenancy, on your own hardware on-premise, or fully air-gapped, using models available in your environment or self-hosted open weights. No training on your data.

Least privilege, credentials issued by you, time-bound and revoked at the end. Named engineers only, no shared service accounts, and everything encrypted in transit and at rest.

Every model call, tool use and write-back is recorded with actor, input reference, cost and outcome, retained to your policy so an auditor can reconstruct any decision the system made.

On anything that touches money, a customer, or a regulated record. We design those gates with the operator rather than bolting them on when the security review pushes back.

Yes. Every automated write has a documented rollback path and a kill switch your team controls. If it can’t be undone in one action, we don’t automate it.

Consultancies hand over recommendations. We hand over running systems, and we operate three businesses of our own on the same patterns. The advice comes from production experience rather than a methodology.

Hiring is the right answer eventually, and our engagements are designed to end in your team owning the system. Forward deployment gets you there faster, with people who have shipped this class of system before.

Whatever is available and appropriate in your environment: hosted frontier models through your own cloud, or self-hosted open weights on your hardware. Model choice is per task and changes as the frontier does; the rest of the system is yours regardless.

A scored eval set built from your own history, not a vendor benchmark, run on every change and handed to your team to own. Accuracy claims without an eval set are marketing.

That’s usually the project. SOAP endpoints, SFTP drops, fixed-width exports, permissions models built in 2009: we’ve shipped against all of it, and we scope it explicitly rather than discovering it in month three.

Fixed scope and fixed fee, quoted after the Audit and driven by pod size, environment constraints and integration surface. No hourly meter, and no licence on work you already own.

No. Forty-five minutes with two of our people, and you leave with an honest read on whether this is worth building, whatever you decide to do next.

Yes: monitoring, model and dependency updates, and the small changes a live system needs. It’s monthly and cancellable, because it should end when your team is ready to own it outright.

Three companies in the Gridwork Group that run on systems we built: an AI employee platform for operations, an AI-first listing platform for real estate agents, and a managed ecommerce service. They’re our proof, and the source of most patterns we deploy.

Two founders, Mohamed and Aaqil, are on every engagement, with more than ten engineers and operators across the group, and around thirty AI agents running production work alongside them.

Named engagements are walked through under NDA on the call, including the ones that didn’t work and why. We’d rather show you that than a logo wall.