Business Automation in Bishkek

What can be automated

Automation suits work that repeats every day and follows clear rules. In companies in Kyrgyzstan, that is most often:

• Documents. Contracts, quotes, acts and reports are assembled from data already in the system — no copying from last time's file.
• Customer conversations. Sales on WhatsApp and Telegram around the clock: the bot answers the usual questions, takes the request and hands it to a manager.
• Accounting and 1C. Bank statements are parsed automatically and loaded into 1C instead of being re-typed by hand.
• Requests and record-keeping. Orders, warehouse, vehicles, working hours — in one system, not in spreadsheets and chats.
• Reports for the owner. The numbers gather themselves from every system instead of being pulled together by hand at the end of the month.
• Routine work with text. Where a person reads one document and fills in another from it, AI agents take over — with the result checked before it moves on.

Signs a process is ready for automation

Companies usually come to automation not because AI is in fashion, but because manual work starts holding back growth. These are the signs we see most often:

• The same data is entered several times. A request is written down in a messenger, then in a spreadsheet, then in 1C — with errors in each place.
• The numbers don't match. Stock in the warehouse, in the spreadsheet and in 1C are all different, and working out which is right is done by hand.
• The owner's report takes days to compile. By the time it is ready, the decision has already been made without it.
• Customers wait for an answer. Messages sent in the evening or at the weekend go unanswered until morning, and the request goes to whoever replied first.
• A process depends on one person. They fall ill or go on holiday — and the work stops, because nobody else knows how it is done.
• Company growth means headcount growth, one to one. Twice the orders means twice the people moving them from one place to another.

If you recognise two or three of these, a process review will show which site to start with and how much is being lost there.

Automation and AI agents: what's the difference

Classic automation follows fixed rules: a statement arrives — allocate the payments to accounts; an order arrives — write off the stock. It is reliable and cheap, and most tasks are solved exactly this way.

AI agents are needed where the rules can't be written down in advance: reading a contract and pulling out its terms, understanding what a customer is asking, compiling a report from documents of different kinds. The agent does the rough work, and the result is checked — automatically or by a person — before it moves on.

In practice we combine both approaches: rules where the process is stable, AI where the input is text, photos or live speech. More on this in our piece on what sets an AI agent apart from a chatbot (in Russian).

Where automation starts

We don't rewrite everything at once. The work goes step by step, and at each step you can see what has been done.

1. Review. We study how the work runs today: who does what, where data is lost, what costs the most.
2. Design. We build the data model and the process map and show you what the solution will look like before it is written.
3. Pilot. We launch on one site and test against real work.
4. Rollout. We deploy company-wide, migrate the data and train the team.
5. Support. Changes, monitoring and growing the system with the business.

Tasks we have already handled

We name clients only with their permission. Where there is no permission — only the industry and the task.

ALNUR — wholesale distribution

Before. Orders from shops came by phone and messengers, stock in the warehouse didn't match what had been promised to customers, prices were recalculated by hand, and the owner learned about sales from reports at the end of the period.

What we did. For ALNUR we built a system based on optoERP, and since August 2026 the company has been running its real work in it — products, batches, stock, orders, customers and prices:

• B2B storefront. A client shop orders from its phone at its own price list and sees its order history and outstanding balance.
• Order pipeline. Each order has a status from placement to delivery; stock is reserved immediately, and goods are written off from the batch with the nearest expiry date.
• Warehouse. Batch receiving, pick lists, write-offs and stocktaking with discrepancies.
• Prices. Markup rules, a change history and a queue in which the owner approves automatic recalculation before customers see the prices.
• Delivery. The courier sees their jobs on the phone, confirms delivery with a photo and records cash received.
• Sales reps. Visits to assigned outlets with a GPS check-in, a walkthrough checklist with photos and an order on the shop's behalf.
• Products from a photo. A new product is added from a photo: AI suggests the product card, category and brand, and anything uncertain goes to a person for review.
• Owner dashboard. Orders in real time, role-based in-app notifications and push notifications to the phone.

DSKOM — construction equipment, production and shipping

Task. Timesheets, waybills, repairs, shipments of asphalt, concrete, crushed stone and soil, and incoming raw materials were all kept separately by site. The summary report was compiled by hand, payroll was calculated from data that had already drifted apart, and attachments moved between machines and sites without any record.

Solution. For DSKOM — a system based on parkoERP, with all of this in one place:

• Check-in by face and location. An employee checks in and out from their phone; the system matches their face and confirms they are at their site. The timesheet fills itself in without a timekeeper.
• Payroll. Calculated automatically from the timesheet, overtime, night shifts, fines and bonuses; the accountant checks and closes the month instead of rebuilding it.
• Equipment. A register, maintenance and repairs, waybills and acts with print forms.
• Attachments. Every bucket and hydraulic breaker has a QR sticker: you can see which machine it is on, who has it and where, and stocktaking is done from a phone.
• Shipping and raw materials. The weighbridge operator enters trips and deliveries, raw material stock is calculated automatically, and items running low come first in the list.
• Reports with AI analysis. Reports in Excel and PDF at one click; AI highlights deviations — rising overtime, idle equipment, unusual raw material consumption.

Accounting firm

Before. Client bank statements were retyped into 1C by hand — line by line, for every company.

What we did. Parsing of statements from PDF and Excel with a check before output, and a one-click export for import into "1C: Accounting for Kyrgyzstan". There is also a Telegram bot option: the accountant sends a statement and gets back a ready file. More on the 1C integration page.

Occupational safety service

Before. Requests and violations lived in chats and on paper; acts and reports were drawn up by hand.

What we did. We launched Telegram bots and built a client portal with dashboards, violation tracking, reports and acts in PDF.

What this can look like in your industry

Below are typical scenarios, not projects for specific clients. This is how automation is usually set up in the Bishkek companies we talk to.

• Clinic or salon. A patient writes on WhatsApp at night — the bot replies, shows free slots and books them in. In the morning the administrator sees ready bookings instead of a list of unread messages. The day before the visit, the client gets a reminder.
• Service company. A request from a messenger or the website lands in the system by itself and is assigned to a technician by area and workload. The technician closes it from their phone with a photo, and the act and invoice for the client are generated automatically.
• Shop or chain. Sales and stock from every outlet come together in one report every morning. The system suggests when it is time to reorder and prepares the order for the supplier.
• Construction or design firm. Standard documents — work execution plans, process charts, acts — are assembled by AI agents from the source data and the regulatory base; the engineer checks and signs them rather than writing from scratch.

Where automation doesn't pay off

Not everything that can be automated should be. If a process happens once a month, runs differently every time, or depends on one particular person's judgement, a system will cost more than the manual work. We name these spots plainly during the review — so you don't pay for automation for its own sake.

How to tell whether a rollout has paid off is covered in our piece on how to measure the effect of automation: three numbers we record before work begins.

What we need from you

• Time with the people who do this work today. Not only the manager: the person who runs the process every day knows best how it really works.
• Sample documents and data. Exports, spreadsheets, standard documents — we test the solution on them, not on made-up examples.
• One person on your side who makes decisions about the pilot.
• Access to the systems that need connecting — when the work gets there, and only to the extent the task requires.

What the cost depends on

We quote an exact price after the process review. It depends on:

• How many processes and sites we automate — a single process or a chain from request to report.
• Which systems we connect. 1C, banks, CRM, messengers, spreadsheets: each connection means setting up the exchange and testing it.
• How much data needs migrating — reference lists, stock, history.
• Whether AI agents are needed and in which languages they work: Russian, Kyrgyz.
• Where the solution runs — in the cloud or on your server.
• Support after launch — a one-time handover, or ongoing support and development of the system.

The process review is the first step of every project: after it, the scope of work, the price and where automation will pay off are all clear.

Where we work

The AI Lab office is in Bishkek, at 110 Toktogul Street. We work with companies across Kyrgyzstan; outside Bishkek we work remotely. AI Lab is a brand of Devline LLC (ОсОО «Девлайн»).