AI What Is AI Implementation for Businesses in Uzbekistan?
AI implementation is not adding a chatbot. It is embedding intelligent agents into real business workflows - lead qualif...
Friday, 16:40. The warehouse has enough packaging material for two more days. Meanwhile, the procurement manager is staring at seven quotes received through Telegram, email, and Excel. One supplier listed the price without VAT, another failed to specify the exchange rate, and a third separated the delivery fee from the quoted amount.
A well-designed AI assistant collects these quotes, matches product names and units of measurement, and converts prices into UZS. It then adds VAT and transportation costs and accounts for customs duties. Payment terms and delivery dates are displayed separately as well. The system ranks the options by final landed cost and delivery time, flags missing information, and explains the reasoning behind its recommendation. The manager checks the source lines and then approves the decision.
AI should identify the purchase that offers the best overall value for the business, not merely the lowest unit price. A system that compares prices alone may work quickly, but one poor choice can wipe out all the time it saved.
The problem is not the number of quotes. It is the way suppliers present their terms. One supplier gives a price per item, while another quotes per box. One delivers the goods to a warehouse in Tashkent, while another hands them over at the border. An Excel formula cannot identify these differences in the text on its own.
According to the Statistics Agency, Uzbekistan’s foreign trade turnover reached 41 billion dollars in six months. As procurement volumes grow, manually checking currency and logistics terms becomes expensive. The most dangerous mistake often does not appear in a formula. It happens when a manager misinterprets a single sentence in a quote.
A good system does not declare a winner based on the price column. It calculates the true cost of the purchase to the business.
01
Standardizes the price, VAT, and discount using the same calculation basis.
02
Adds transportation and insurance. It also includes customs duties in the total amount.
03
Distinguishes between the promised lead time and the date when the goods will actually reach the warehouse.
04
Calculates how advance or deferred payment will affect cash flow.
05
Flags the quote for additional review if a certificate or product code is missing.
Suppose a retail chain in Tashkent needs 1 000 packages. Supplier A quotes 94 million soums. After transportation is added, the total reaches 102 million soums, with delivery in five days. Option B appears to cost 89 million soums, but delivery brings the total to 103 million soums. The shipment will arrive in 14 days.
Supplier C will deliver the goods to the warehouse in seven days for 96 million soums but requires a 50 percent advance payment. If the warehouse has enough stock for eight days, AI recommends option C and shows how the choice will affect cash flow. The 89 million soum quote looks cheap on paper. In practice, it is neither the least expensive nor the fastest option.
Warning
Do not tell AI to "choose the winner" without requiring an explanation. Every recommendation should be verified against the calculations and the relevant line in the original document. A human approves the final decision.

A spreadsheet alone cannot resolve mixed currencies, units, and hidden supplier terms.
Start by selecting one frequently purchased product group. Trying to automate all procurement at once will prolong the pilot and make it harder to trace the source of errors.
Quotes in Uzbekistan may arrive in Uzbek, Russian, or English. AI can process documents in all three languages within the same workflow. The point that needs verification is different: do "kit," "box," "set," or a manufacturer’s code refer to exactly the same product?
For import purchases, a description alone is not enough. Certificate and labeling requirements must be checked, and the documents required for customs clearance must be compared. If the system finds an ambiguity, it should not assign a value on its own. Instead, it prepares a precise question to send to the supplier through Telegram or email.
Finding the cheapest line item is easy. To make the right purchase, you also need to read the terms behind the price.
Celion engineering practice

The lowest quoted price does not always mean the lowest landed cost.
Do not begin an AI project with a vague requirement such as "we need a smart system." Choose a measurable task. If a procurement specialist compares 15 requests per week and saves 45 minutes on each one, that frees up more than 11 hours every week. That time can be spent negotiating with suppliers or evaluating alternative sources.
Packaging, office equipment, or regularly purchased spare parts are suitable choices for the first pilot. Three KPIs are enough for one month. Measure the time spent on comparisons and the hidden costs identified. Also record the percentage of cases in which the manager accepts the AI recommendation. If the pilot produces no result, check the criteria before blaming the model.
The following rules are enough to build a system that works. You can add complexity later.

Term normalization lets AI compare equivalent products accurately.
Can AI also read quotes received through Telegram?
Yes. Text, PDFs, Excel files, and photos of documents can be received through a Telegram bot or a dedicated service account. The system extracts the required fields and places them in a comparison interface. There is no need to connect an employee’s private conversations. Procurement should use a separate, controlled channel.
Does the system work with 1C, Odoo, or ERP?
Yes, provided the system has an API or another data exchange mechanism. The AI assistant can retrieve product records and warehouse inventory and compare them with previous purchase prices. The approved result is then returned to the ERP. If integration is unavailable, the pilot can begin with an Excel export to validate the benefits.
What happens if AI reads a number in the quote incorrectly?
Fields with a low confidence score should automatically be sent for review. The manager sees the relevant line from the original document next to the result and approves any correction. Additional control rules can be applied to large amounts, unusual exchange rates, or sharp discrepancies. Automatically placing an order without an explanation is risky.
Does the AI assistant choose the supplier on its own?
Not in a standard workflow. The system calculates the options, highlights risks, and prepares a recommendation. Final approval remains with the procurement manager or another responsible leader. Later, repetitive low-value purchases governed by strict rules may be approved automatically. Spending limits and an audit log must be defined in advance.
How quickly can a pilot like this be launched?
If there is one product group, two or three input channels, and clear evaluation rules, an initial pilot can usually be prepared within a few weeks. The timeline depends on document quality, ERP integration, and approval stages. A precise estimate can be provided after reviewing sample workflows and documenting the criteria.
The Celion team will review your actual commercial quotes and prepare a pilot scenario, integration plan, and measurable KPIs for an AI assistant. Contact us. Together, we will decide which product group to start with and how to measure the results.
Callback
Leave your details and we will contact you shortly.
Become a Partner
Leave your contact details and we will get back to you shortly.
Order Service
Leave your phone number and we will contact you shortly.