01
Support repeats the same lookups
Your team moves between systems to answer familiar requests.
AI / AI agents
AI assistants that use approved tools, work within defined permissions and bring a person into the loop when a task needs judgement or approval.
Scope, timeline and pricing agreed after reviewing your needs.
A task with visible boundaries
- 01Understand
- 02Lookup
- 03Approve
- 04Act
Illustrative concept, not a live system or measured result.
AI agents
Tool integrationsApproval rulesEvaluation casesRun visibility
Who this is for
For one repeated, useful task with a clear owner and clear limits on what the agent may read or change.
02
Your chatbot cannot take a step
A conversation ends before the customer’s actual task can progress.
03
Useful automation needs boundaries
You want to test agent actions without giving open-ended access.
What you get
Tool contracts, input validation, approval rules, audit events and an evaluation set, agreed before the agent touches a production system.
A bounded use case
Select a task with clear inputs, rules, success criteria and an accountable owner.
Approved tool connections
Connect the required system operations with scoped credentials and validated inputs.
An observable task runner
Make the task state, tool activity and outcome visible to authorised reviewers.
A task with visible boundaries
UnderstandIdentify the request and required context.
Look upRead the permitted records through approved tools.
Check permissionCheck the action rules and approval requirements.
Tool integrationsApproval rulesEvaluation casesRun visibility
Action limits and approvals
Define the actions that require confirmation, escalation or a full stop.
Scenarios in agreed languages
Evaluate realistic requests and ambiguous wording in the agreed languages, including mixed-language cases where in scope.
Evaluation and handover
Test completion, incorrect actions and human handover before widening access.
How we work
Select a bounded task, connect approved tools, test success, errors and handover, then pilot with an agreed group before expanding.
Select the task
Review candidate requests and pick a contained workflow with a clear owner.
A small, useful starting point.
- Task definition
- Success criteria
- Permission boundaries
Connect carefully
Build the required tools and define which actions need confirmation or escalation.
Capabilities with limits.
- Tool contracts
- Approval points
- Audit events
Test the exceptions
Use representative and difficult cases to review behaviour before a controlled pilot.
Failures are part of the test.
- Task evaluations
- Misuse scenarios
- Human handover
Pilot and review
Introduce the agent to an agreed group and review errors, usage and operating cost.
Evidence before expansion.
- Pilot feedback
- Action review
- Expansion decision
Explore the possibilities
Illustrative concepts from our demo portfolio. They show possible experiences, not commissioned client projects, platform-specific implementations or measured results.
Context before action
The knowledge-assistant concept shows a possible place for source context. Tool execution would need separate integration, permissions and testing.
A task with a visible next step
The workflow concept illustrates how a supported task and its handover could be made visible to an operations team.
PLANS & INCLUDED BENEFITS
Discovery produces a specification and estimate, not a deployed agent. The pilot runs one use case on one channel in one language; the connected plan covers up to two channels and two integrations, not unlimited automation. Evaluation scenario counts and the no-high-risk-decisions rule apply to every plan.
Defined scope
Agent discovery
Validate one agent use case before commissioning a build.
Starting from
4,000SAR
/ project · excl. VATOne-time project fee. Support beyond the included period is quoted separately.
Included in your plan
- One workshop and up to 3 stakeholder interviews
- One workflow, tool and permission map
- Evaluation plan with 20 example tasks
- Written specification and build estimate; no deployed agent
Feasibility review session included
Included in the starting price. Scope and limits are listed below.
Recommended
Agent pilot
A limited working agent for one channel and one language.
Starting from
18,000SAR
/ project · excl. VATOne-time project fee. Support beyond the included period is quoted separately.
Included in your plan
- One use case with up to 5 supported intents
- One documented system integration
- Human approval and handoff controls
- Evaluation against 30 agreed scenarios
30-day defect fixes included
Included in the starting price. Scope and limits are listed below.
Defined scope
Connected agent
A bilingual agent with a broader set of controlled actions.
Starting from
38,000SAR
/ project · excl. VATOne-time project fee. Support beyond the included period is quoted separately.
Included in your plan
- Arabic and English across up to 2 channels
- Up to 10 intents and 2 documented integrations
- Permission checks, audit events and human handoff
- Evaluation against 60 agreed scenarios
Team handover session included
Included in the starting price. Scope and limits are listed below.
Starting prices in SAR, excluding VAT where applicable. Final scope and price are confirmed in your proposal. Model/API usage, hosting, software subscriptions and major data cleanup are separate. Scope assumes client-provided access to documented APIs and approved data. No foundation-model training or autonomous high-risk decisions. Handover is one 60-minute session; included defect fixes cover the delivered scope for 30 days.
Before we start