Business automation is no longer a topic reserved for large corporations with an IT department budget. In the past month, the query "business automation" in the Kazakhstani segment has garnered over a hundred impressions, and along with related terms like "business process automation," "automation with AI," "automation and digitalization," the count reaches several hundred. This is not an abstract interest: behind it are small business owners who, in 2026, faced a new Tax Code, a VAT rate of 16%, and an increasing burden on primary documentation — and realized that manual handling is no longer feasible.
In our practice at West Star Ltd, we see the same scenario every week: a company wants to "automate everything at once," spends months choosing the perfect system, and ends up launching nothing. Business automation works the opposite way — it starts with one narrow process that takes the most time and is most prone to errors. In this article, we'll explore where to realistically start, what gets automated first, how much time it takes, and where the hidden pitfalls are that are rarely mentioned in promotional articles.
Where Business Automation Begins
The first step is not purchasing software, but an honest inventory of processes. Take a week and note where work time goes: how many times an employee manually transfers data from one table to another, how many requests are lost in correspondence, how often the same number is entered into three different systems. These repetitive, predictable operations are candidates for automation. If a process is performed differently each time and requires human decision-making, it's too early to automate it.
A good priority criterion is the formula "frequency × cost of error." A process that repeats dozens of times a day and costs money when errors occur (incorrect stock balance, lost request, overdue invoice) provides the maximum return from automation. We usually advise starting with one such process, bringing it to a working state in two to three weeks, showing the team the result, and only then moving on to the next. This approach to business process automation reduces risk: if something goes wrong, you lose weeks, not half a year and the entire budget.
The second principle is that automation should rely on data you already have. For most Kazakhstani companies, the core of accounting remains 1C, and this is a huge advantage: it already contains nomenclature, balances, counterparties, documents. Through integration with 1C via the OData protocol, this data can be fed into chatbots, websites, analytics, and AI services without manual exports. Starting automation from scratch, ignoring the accounting system, is a typical mistake that leads to data duplication and perpetual desynchronization.
The third principle, often forgotten, is that automation should be measurable from the start. Before launching a bot or integration, fix the current numbers: how many requests per day, how much time is spent on a document, how many errors per month. Without this starting point, you won't be able to prove to yourself and the team that business automation has truly paid off — meaning the next step will rely on belief rather than facts. In our projects, we always agree on a metric before starting: it turns the vague "it's more convenient" into a concrete "application processing time reduced from twelve minutes to one and a half."
What Really Gets Automated First
Let's break down specific scenarios that yield quick results and don't require restructuring the entire company.
Processing incoming requests. A client writes in WhatsApp or Telegram, and an employee manually transfers the order to 1C or CRM. This is a classic candidate: a bot accepts the request, checks the stock, records the contact, and creates a document in the accounting system automatically. The first line stops "burning out" on repetitive questions about availability, price, and order status.
Primary documents and tax reporting. With the transition to the new Tax Code and 16% VAT, the burden on accounting has increased: ESF, SNT, accompanying invoices, virtual warehouse. Automation here is not a replacement for the accountant, but a removal of mechanical work: generating documents from 1C data, deadline control, checking details. The threshold for mandatory VAT registration in 2026 is reduced to 10,000 MRP (about 43.25 million tenge at 4325 tenge per MRP), and more small businesses fall under it — meaning the volume of primary documentation that someone must handle without errors is growing.
Synchronization of sales channels. Products and prices managed in 1C should reach the website, Kaspi, and marketplaces without manual labor. Manual price list export once a week leads to outdated prices and sales of items not in stock. Automatic synchronization closes the gap between accounting and the showcase, and the more channels a company has, the more costly each desynchronization becomes. For retail and wholesale trade, this is often the first process worth automating because the cost of error is immediately visible — in the form of a canceled order or an unhappy customer.
HR and internal processes. Onboarding a new employee, request approvals, time tracking, HR document flow — all of this is also subject to automation, especially when employee data is already maintained in 1C. The return here is not as immediate as in sales, but it accumulates: the company stops losing documents and time on manual forwarding, and employees get a clear and consistent process instead of correspondence in different chats.
Internal notifications and control. A manager shouldn't have to log into five systems to find out the day's revenue or see that an order is stuck at the payment stage. Automatic summaries in a messenger, notifications of critical events, dashboards with real data — this is cheap automation with high returns for management. Such a scenario almost doesn't require changing team habits: the data is already in 1C, you just need to set up its regular delivery to where the manager reads it — usually the same Telegram or email.
A separate category that has become practical in the last two years is automation with AI. AI agents are already confidently handling classification of inquiries, data extraction from documents, draft responses, and initial support. But it's important to understand the boundary: AI is good where probabilistic error is acceptable and human verification is present, and bad where 100% accuracy without control is needed. We wrote in detail about how we build such pipelines in our analysis of content and process automation — the same logic applies to any repetitive workflow.
Limitations and Pitfalls
An honest conversation about automation is impossible without its weak points. Here's what you need to know before starting.
-
Automating a bad process gives a quick bad result. If a process is convoluted and relies on verbal agreements, it needs to be organized first, and only then automated. A program only accelerates what exists — along with chaos.
-
Time savings don't always translate into money savings. If a bot frees up two hours a day for an employee, but those hours don't get redirected, there won't be a direct financial return. The effect needs to be calculated in advance and in money, not in the feeling of "it's more convenient."
-
Integration with 1C requires order within 1C itself. Crooked nomenclature, duplicate counterparties, non-working balances — all of this will surface during the first export via OData. Data needs to be cleaned, and this is separate work that is often underestimated.
-
AI makes mistakes, and this needs to be factored into the architecture. A model can confidently produce an incorrect number or misunderstand a request. For critical operations — money, taxes, legal documents — human control and a clear rollback are mandatory, otherwise, automation becomes a source of costly errors.
-
Dependence on a contractor and "black boxes." If a solution is written in such a way that only the creator can understand it, you become dependent. Request documentation, access to code, and a clear handover — this is part of healthy automation, not a whim.
-
Team resistance. People fear that automation is about layoffs. If you don't explain that it removes routine, not jobs, employees will quietly sabotage the implementation. Technically everything may work, but practically — not.
None of these points are reasons to abandon automation. They are reasons to approach it engineeringly: with calculation, step by step, and with result verification at each stage.
Frequently Asked Questions
How much does business automation cost at the start?
The range is wide, but the first working scenario — for example, receiving requests from a messenger with recording in 1C — usually fits within the budget of one to two work weeks of development. We intentionally advise starting small: this way you test the return on real money before investing in a comprehensive solution. If you need an estimate for a specific process, it's easiest to get it by describing the task through the form on the contacts page.
Do I need to change 1C or another accounting system?
Most often, no. Modern automation is built on top of existing 1C via OData or web services — data remains in the accounting system, and the bot, website, or AI service simply reads and writes it. Changing the configuration makes sense only when the accounting itself is incorrect, but this is a separate task, not directly related to automation.
Where to start if there are many processes and everything seems important?
Choose one process using the formula "frequency × cost of error" — the one that is most often performed manually and costs the most when it fails. Bring it to a working state, measure the effect, and only then move on to the next. Attempting to automate everything at once is the most common reason projects don't reach launch.
Will AI replace employees during automation?
In the near future — no, but it will redistribute work. AI confidently takes on routine: sorting inquiries, drafts, data extraction, first-line support. Decisions, responsibility, and exception handling remain with humans. The most sustainable model for 2026 is not "AI instead of people," but "human plus AI," where automation frees up time for tasks that machines are not yet capable of handling.
The practical takeaway depends on where you sit. A specialist should start by inventorying their own routine and proposing one measurable scenario to management — this is the best way to demonstrate the value of automation without grand promises. A manager should calculate the effect in money and demand phased implementation with verifiable results, not a large "turnkey" project with vague ROI. An owner should view automation as a way to reduce the business's dependence on specific people and manual operations: in the context of 16% VAT, growing primary documentation, and a labor shortage, this is no longer about convenience, but about the company's resilience.