AI
AI implementation in Uzbekistan
Updated 20 September 2026
Majento develops AI products and integrates artificial intelligence into business workflows. We are based in Tashkent and work with companies in Uzbekistan and Central Asia. We start with a specific task and deploy the system in the client's environment.
Workflows we can automate
- Executive summaries: collect information from documents, CRM and messages, with sources and discrepancies visible.
- Customer support: find answers in a knowledge base, check order status and hand difficult questions to a person.
- Marketing operations: prepare product content, campaign plans and performance reports.
- Meeting follow-up: capture decisions, create tasks and track priorities.
We build custom services and integrate CRM, CDP, ERP, e-commerce and other systems.
Our AI implementation process
- Review the workflow with employees. Identify time lost, available data and responsibility for the outcome.
- Choose a pilot. Record the baseline, quality criteria and actions requiring human approval.
- Design the system. Specify integrations, model selection, permissions, deployment and data boundaries.
- Build and test. Evaluate answers against checked examples, test failures and measure processing costs.
- Deploy and support adoption. Train the team, provide instructions and agree support arrangements.
The engagement can end after assessment. If conventional automation fits the task better, we include that option in our recommendations.
AI strategy deliverables
A process map, suitable use cases, pilot priorities and data requirements. For each use case, we define the expected benefit and how to test it. Consulting is available separately from development.
Data and private deployment
Options include cloud, on-premises and hybrid deployment. We configure permissions, data masking, an AI proxy and cost tracking. Where needed, we deploy inference and a company knowledge base. Model selection and fine-tuning depend on the task and available data. Storage requirements are agreed with the client's IT, security and legal teams.
What we need to assess a project
Describe one workflow, the systems involved and the change you want to achieve. Initial examples should exclude personal data and trade secrets. We propose pilot scope after the review. Timing and cost depend on integrations, constraints and acceptance criteria.
Does every process need AI
No. Rules, templates and conventional integrations often solve the task at lower cost. AI can help with unstructured text, voice and changing context. We decide after testing representative examples.
Can an agent act without an employee
Each action has separate permissions. Answering from a knowledge base and making a payment require different controls. Consequential actions stay subject to approval by the responsible person.