Business process automation: types, stages and cost

Business process automation means handing a whole chain of routine steps over to software: a request moves from first contact to payment, tasks travel between departments on their own, data comes together without manual re-entry. What gets automated is whatever repeats, can be measured and eats up a lot of time.

Below we look at how automation differs from accounting systems, RPA and BPM, what classes of systems exist and what they actually cover, what stages a rollout consists of, what drives the price and why the system most often fails to take root.

What is business process automation?

In short: it is when software pushes the process forward by itself instead of waiting for an employee to move data across by hand.

The formal definition comes from Gartner's IT glossary: business process automation is the automation of complex business processes and functions beyond conventional data manipulation and record-keeping, usually through the use of advanced technologies. The same entry contains the wording that best explains the difference for an owner:

It focuses on "run the business" as opposed to "count the business" types of automation efforts — Gartner IT Glossary, Business Process Automation entry.

The difference is practical. An accounting package counts the business: it records what has already happened. Process automation runs the business: the request enters the queue by itself, the sales rep receives a task with the context already attached, the warehouse sees the reservation, the manager sees the deal stage without having to ask someone to send over a report. A company can keep flawless records and still lose requests, because there is a person with a spreadsheet standing between the systems.

How does automation differ from accounting systems, RPA and BPM?

In short: an accounting system records facts, BPM manages processes on paper, RPA runs robots on top of interfaces, and process automation executes the process end to end.

These terms get mixed up constantly, and that is why companies buy the wrong thing. The easiest way to separate them is by object and depth.

There is a separate misconception attached to BPM: that drawing a process diagram is enough to make it work. In practice it is not:

BPM systems do not implement processes exactly as they are described in the diagram. There is no universal solution that suits everyone — a breakdown of process management myths on Habr.

RPA stands apart: the robot works on top of existing windows, without touching the code of the systems. That gives a fast start on high-volume routine such as reconciliations and data transfers, but it also makes the solution fragile.

Which processes get automated first?

In short: the ones that repeat, can be measured and consume a noticeable number of person-hours.

Processes are usually split into three groups, and each has its own reason for automation.

1. Core. Sales, production, service delivery, shipping. They bring in the money, so they are automated for speed and revenue.
2. Supporting. Procurement, warehouse, hiring, accounting, IT. They keep the core running, so they are automated to cut costs.
3. Management. Planning, budgeting, control, reporting. They are automated for transparency and better decisions.

The criterion for picking a specific process is simple and verifiable. The process must be repeatable, meaning the operation is regularly the same. Measurable, meaning it has a time, a cost and an error count from before the rollout. And labour-intensive enough for the effect to show against the cost. A process that happens once a quarter and takes an hour is not worth automating at any price. A detailed look at choosing the first process is in our article on where to start with business automation.

What types of automation systems are there?

In short: the class of system is dictated by the class of process, and there is no universal solution.

The difference between a scripted bot and an agent that chooses its own steps is covered in detail in the article “AI agents for business”. How a consulting bot on a website is built, including its data requirements, is covered in the piece on an AI chatbot for a website. A separate topic is tying systems you have already bought into a single loop, and that is covered by the article on integrations.

Off-the-shelf or a custom system built around the process?

In short: off-the-shelf wins at the start and on standard processes, while a custom system is justified where the process is a competitive advantage.

The honest framing looks like this: most companies go through both stages. First they take something ready-made to launch quickly, then they move to their own once scale and specifics appear. Off-the-shelf is not the enemy, and custom development is not automatically better.

The trouble starts where the company's process does not fit someone else's logic:

No off-the-shelf solution on the market will be able to reproduce a company's unique business processes, and that means you are facing a difficult transformation not only at the point of automation but also in how work is organized inside the company — Uplab, a review of off-the-shelf solutions.

The choice then comes down to what you change: the system to fit the business, or the business to fit the system. If the process is exactly what sets the company apart from competitors, breaking it to fit a boxed product means losing that advantage. The second factor is ownership: with a custom system the code and the rights stay with the company, the contractor can be replaced, and the data sits wherever the business decides. A detailed comparison with the numbers is in our separate piece “Your own system or an off-the-shelf one”.

What stages does a rollout consist of?

In short: five stages, and the first two cannot be skipped, otherwise what gets automated is the existing mess.

1. Process audit (as-is). An inventory of processes, interviews with staff, measurement of the time and cost of operations. What you need to look at is the real process with its workarounds; the written procedure shows what is wanted, not what happens.
2. Target process design (to-be). A model of how the process should run after automation. This is where the process gets fixed, and only then is it locked into the system.
3. Choice of solution. Off-the-shelf, low-code or custom; cloud or your own server; Russian software or not. This fork is resolved after the first two stages, not before.
4. Rollout. Configuration or development, integrations, data migration, training, a phased launch. Launching everything at once almost always ends in a rollback.
5. Support and development. Monitoring, measurement of the effect achieved, refinements based on feedback.

The fifth stage is skipped most often, and it costs dearly: without measuring the effect there is no way to prove payback or to understand what to improve next. The practical side of the first step is covered in the article on automating routine work.

How much does automation cost and what drives the price?

In short: the license price is the tip of the iceberg, and what you need to count is total cost of ownership over three to five years.

Quoting specific price lists is pointless, because the spread depends on the process by an order of magnitude. It is more useful to understand the cost drivers:

• the number and complexity of scenarios;
• the amount of integration with CRM, ERP, 1C (the accounting and ERP platform most Russian companies run on), telephony, banks and marketplaces;
• data migration, its volume and quality;
• how deep the customization for non-standard logic goes;
• staff training and support;
• requirements on where data is hosted and how it is protected.

The point that regularly wrecks budgets: the cost of the rollout often exceeds the price of the licenses themselves. Configuring processes, building integrations, migrating data and training people cost more than the subscription, and the vendor's website shows only the subscription.

Hence the practical conclusion for an owner. Options should be compared on total cost of ownership over three to five years: licenses plus rollout plus integrations plus support plus tweaks plus training. Off-the-shelf has a low entry price, but the payments repeat and the customizations pile up. A custom system has a higher entry price, but the code becomes an asset of the company.

How do you calculate the effect and the payback?

In short: the basic formula is the benefit minus the cost divided by the cost, and the benefit is assembled from measurable metrics.

Calculating the effect "by feel" does not work, so the metrics are recorded before the rollout. Five groups do the job:

1. Payroll savings. Freed-up person-hours multiplied by the cost of an hour, plus the hire that did not have to happen.
2. Process speed. Cycle time from request to payment, or from order to shipment.
3. Fewer errors. The cost of rework, fines and losses from human error.
4. Cost per operation. The unit cost of a transaction before and after.
5. Indirect effects. Customer retention and the quality of data for management decisions.

Market reviews put the average payback period for automation projects somewhere between six months and a year and a half, but that is an estimate from vendor materials rather than a norm: a process with high frequency and expensive errors pays back noticeably sooner than a rare and cheap one.

Why does automation fail to take root?

In short: the cause is almost always outside the system — in the process, the people and the way the contractor relationship is set up.

The typical scenarios that repeat from company to company:

1. Chaos got automated. An inefficient process locked into code starts reproducing its problems faster. The only cure is the as-is and to-be stages.
2. The boxed product did not fit the process. What follows is either endless customization or breaking the business to fit someone else's logic.
3. Staff were not trained. People go back to spreadsheets, because that is more familiar and faster.
4. Noise automation. The system buries a manager under dozens of automatic tasks a day, and within a week they are closing them without looking.
5. Endless tweaking. A model where every month calls for a budget for yet another refinement while the business model stays the same. It turns into a subscription to dependence.
6. Data in someone else's cloud. Customers' personal data moves out to an external service, and control over it is lost.

The moment you have sent data to an external service, you no longer control it 100%. There is no telling where and how it will be stored, or who can see it — Iteco.Cloud on Federal Law 152-FZ and the cloud.

The last point is not only about trust. Under Federal Law 152-FZ, Russia's personal data protection law, the personal data of Russian citizens is stored on servers inside Russia, and the protection measures are set by Order No. 21 of FSTEC, Russia's technical regulator for information security, depending on the system's protection level. The operator has to be able to document where the data physically sits. How to build a loop that data does not leave is covered in the article “Neural networks without data leaks”. Why the most popular class of systems in particular stalls is covered separately in the piece “Why CRM fails to take root”.

What is changing in the Russian automation market?

In short: foreign vendors have left, the share of Russian systems is growing, and robots are turning into orchestrators with AI in the loop.

A few facts worth keeping in mind when choosing a system for the next three years.

ERP and the departure of foreign vendors. The Russian ERP market stood at 90 billion rubles in 2024 and will grow to 120 billion by 2030, at an average annual rate of around 8%. Over 2025 the share held by Russian ERP developers rose by 10 percentage points to 60%, and about 80% of that share belongs to 1C. This is reported by CNews, citing Alexey Stoyanov, head of the competence centre at the T1 holding. The same piece gives the scale of actual migrations: at Gazprom Neft, 1C is used at 34 plants, ten of which moved over from SAP.

Robotic process automation. Foreign RPA vendors have left the Russian market for good:

The international vendors (UiPath, Blue Prism, Automation Anywhere) have finally left the Russian market, and by now customers who care about compliance with data protection legislation and reliable support have moved entirely to solutions from domestic vendors — a review of Russian RPA platforms, CNews.

Meanwhile the technology itself is changing role. According to the study "Gromov's RPA Circle 2026", the robot is ceasing to be a script executor and becoming an orchestrator of a chain of decisions, bringing in AI where rules do not work and handing the task to a human when judgement is required.

Regulation. From 1 January 2025, the use of foreign software at significant facilities of critical information infrastructure is prohibited. For companies in the relevant industries, that settles the question of choosing between a Russian and a foreign solution.

Where do you start?

An order of operations you will not have to redo:

1. Pick one process. Repeatable, measurable and labour-intensive. A statement at the level of "let's automate the company" does not work; you need a specific "a request from first contact to invoice".
2. Describe how it runs today. With the workarounds and the delays between people, not according to the written procedure.
3. Calculate what it costs. Hours, errors, rework. Without that figure there will be nothing to measure the effect against.
4. Design the target process. First remove the unnecessary steps, then lock it into the system.
5. Choose a solution to fit the process. A standard process is covered by an off-the-shelf product; a process that is an advantage calls for a system of your own.
6. Budget for support. Monitoring, measurement of the effect, refinements based on feedback.

At IncubeAi we build systems around a company's process: ERP and CRM, AI agents and bots, integrations with 1C and external services. The code and the data stay on the business's side, the work runs under a contract, and support is available after handover. If the process is already clear, start with a conversation about it — discuss a project.

Sources

• Gartner IT Glossary, Business Process Automation (BPA) — the definition and the split between "run the business" and "count the business".
• CNews, “Instead of SAP: who is replacing ERP in Russia and how”, 9 January 2026 — an ERP market of 90 billion rubles, a 60% share for Russian developers, Gazprom Neft's migrations off SAP.
• A review of Russian RPA platforms 2025 on CNews Market — the departure of UiPath, Blue Prism and Automation Anywhere.
• An article on the Russian RPA market moving towards process orchestration, 17 July 2026 — the study "Gromov's RPA Circle 2026".
• Habr, “Why BPM systems are not about automation”, 2025 — on the gap between a process diagram and its execution.
• Uplab, “An off-the-shelf solution as an alternative to building your own” — on the limits of ready-made solutions.
• Iteco.Cloud, “Federal Law 152-FZ and cloud technologies” — on control over data in external services.