Workflow-specific AI systems for marketing, WhatsApp CX, quotation automation, document search, and focused pilot projects. We do not sell general-purpose AI platforms.

A mini-pilot can start in 2 to 3 weeks. A full pilot runs 6 to 8 weeks. Both depend on data readiness, integrations, and scope.

A draft scope, proposed KPIs, required data and access, a timeline, risk notes, and an indicative SAR range. From there you decide whether to proceed.

Pick the workflow closest to revenue or risk. Then pick the buyer role that owns it. Then read the matching service page.

Services. Each engagement is a scoped pilot with locked deliverables. We do not sell SaaS subscriptions.

Mini-pilot: SAR 20,000 to 30,000. Full pilot: SAR 60,000 to 80,000. Both scoped after a 30-minute discovery call.

Yes. Once the SOW is signed, the pilot is fixed-price. Out-of-scope work becomes a separate phase with its own scope and price.

Milestone-based by default: discovery, build complete, UAT pass, Acceptance Pack sign-off. Custom terms available for enterprise procurement.

Third-party API, hosting, and LLM provider costs are billed to the customer and defined in the SOW. We do not mark these up.

You do. Full IP transfer. No vendor lock-in. Source code, documentation, and configuration belong to the customer.

Scope, KPIs (baseline / target / accepted), runbook, audit trail samples, UAT results, and a scale recommendation memo.

We iterate within scope until the agreed KPIs are met. If the failure is due to out-of-scope expectations, it becomes a new phase.

Not inside the baseline pilot. A second workflow is a change request or a next phase with its own scope.

30-day support window is included after Acceptance Pack sign-off. Ongoing managed service is a separate offering with its own scope.

Yes when feasible. Architecture decisions are made per pilot based on data residency, integration cost, and operational risk.

PDPL-aware by design. We use minimization, masking, RBAC, audit logs, and retention rules. We do not provide legal certification; clients run their own legal review.

In the customer environment whenever possible. When a managed service is needed, data residency is set per provider capability and documented in the SOW.

No. Training on customer data is disabled at the provider level by default. This is documented per pilot in the SOW.

Yes. Every action that affects data or outputs is logged with actor, time, and reason. Logs are accessible to your security team during and after the pilot.

The architecture is provider-agnostic by default. Switching providers is a configuration change, not a rebuild.

Yes. Saudi-natural Arabic is first-class. We codify dialect rules and reject drafts that drift to non-Saudi MSA or other dialects.

Mixed-language input is handled naturally for WhatsApp CX and document search. Output language is matched to the user prompt by default.

Saudi-natural by default. Other GCC dialects are configurable per pilot when the brand voice rules support them.

Citations, golden-set evaluation, retrieval score thresholds, and explicit refusal when sources are weak. We do not measure usefulness without measuring honesty.

The default during the pilot: the system reads from approved sources and produces outputs, but does not write to your systems until UAT and a security gate are passed.

No. The default is to read from approved sources and produce governed outputs. We do not replace ERP, CRM, or core systems.

Baseline before, target during, accepted after UAT. All three documented in the Acceptance Pack with the measurement method.

The system flags the case as an exception and escalates to a human. Silent failures are blocked at design time, not patched later.

Yes. Knowledge transfer and a runbook are part of the Acceptance Pack. A 30-day support window covers operational questions after sign-off.

Yes, with the terms defined in the SOW. Typical clauses cover early termination at milestones with proportional fees and full IP transfer of completed work.

Workflow-specific AI systems for marketing, WhatsApp CX, quotation automation, document search, and focused pilot projects. We do not sell general-purpose AI platforms.

A mini-pilot can start in 2 to 3 weeks. A full pilot runs 6 to 8 weeks. Both depend on data readiness, integrations, and scope.

A draft scope, proposed KPIs, required data and access, a timeline, risk notes, and an indicative SAR range. From there you decide whether to proceed.

Pick the workflow closest to revenue or risk. Then pick the buyer role that owns it. Then read the matching service page.

Services. Each engagement is a scoped pilot with locked deliverables. We do not sell SaaS subscriptions.

Mini-pilot: SAR 20,000 to 30,000. Full pilot: SAR 60,000 to 80,000. Both scoped after a 30-minute discovery call.

Yes. Once the SOW is signed, the pilot is fixed-price. Out-of-scope work becomes a separate phase with its own scope and price.

Milestone-based by default: discovery, build complete, UAT pass, Acceptance Pack sign-off. Custom terms available for enterprise procurement.

Third-party API, hosting, and LLM provider costs are billed to the customer and defined in the SOW. We do not mark these up.

You do. Full IP transfer. No vendor lock-in. Source code, documentation, and configuration belong to the customer.

Scope, KPIs (baseline / target / accepted), runbook, audit trail samples, UAT results, and a scale recommendation memo.

We iterate within scope until the agreed KPIs are met. If the failure is due to out-of-scope expectations, it becomes a new phase.

Not inside the baseline pilot. A second workflow is a change request or a next phase with its own scope.

30-day support window is included after Acceptance Pack sign-off. Ongoing managed service is a separate offering with its own scope.

Yes when feasible. Architecture decisions are made per pilot based on data residency, integration cost, and operational risk.

PDPL-aware by design. We use minimization, masking, RBAC, audit logs, and retention rules. We do not provide legal certification; clients run their own legal review.

In the customer environment whenever possible. When a managed service is needed, data residency is set per provider capability and documented in the SOW.

No. Training on customer data is disabled at the provider level by default. This is documented per pilot in the SOW.

Yes. Every action that affects data or outputs is logged with actor, time, and reason. Logs are accessible to your security team during and after the pilot.

The architecture is provider-agnostic by default. Switching providers is a configuration change, not a rebuild.

Yes. Saudi-natural Arabic is first-class. We codify dialect rules and reject drafts that drift to non-Saudi MSA or other dialects.

Mixed-language input is handled naturally for WhatsApp CX and document search. Output language is matched to the user prompt by default.

Saudi-natural by default. Other GCC dialects are configurable per pilot when the brand voice rules support them.

Citations, golden-set evaluation, retrieval score thresholds, and explicit refusal when sources are weak. We do not measure usefulness without measuring honesty.

The default during the pilot: the system reads from approved sources and produces outputs, but does not write to your systems until UAT and a security gate are passed.

No. The default is to read from approved sources and produce governed outputs. We do not replace ERP, CRM, or core systems.

Baseline before, target during, accepted after UAT. All three documented in the Acceptance Pack with the measurement method.

The system flags the case as an exception and escalates to a human. Silent failures are blocked at design time, not patched later.

Yes. Knowledge transfer and a runbook are part of the Acceptance Pack. A 30-day support window covers operational questions after sign-off.

Yes, with the terms defined in the SOW. Typical clauses cover early termination at milestones with proportional fees and full IP transfer of completed work.