Process automation in the company – where to start

Process automation in the company – where to start

Getting data manually from a form to CRM, copying invoices to a spreadsheet, sending the same emails after each order. Each of these actions takes a few minutes, but multiplied by hundreds of repetitions a month, it can eat up dozens of hours. This article shows how to approach automation step by step – from identifying processes to choosing tools, up to the first implementation.

It's not about automating everything at once. It's about starting with the one process that is the most painful.

How many companies automate already – and what they get out of it

Before we get into tools, some numbers for context. According to a study by McKinsey & Company from February 2024, about 66% of companies have already implemented some form of business process automation. Also, data collected by doit.software (October 2025 report) indicate that 76% of companies using marketing automation achieve a return on investment within the first year – including 32% within the first six months.

Companies using process automation report cost reductions from 10% to 50%, mainly by eliminating repetitive tasks and minimizing human errors.

Sounds encouraging, but these stats apply to companies of various sizes. What does this mean in practice for a business with a small team? Even a simple automation – e.g. a Slack notification after a new order in a store – can save a few hours a week. It's worth starting with such small steps.

Three tools, three approaches: Zapier, Make, n8n

There are dozens of automation platforms on the market, but in practice, most small and medium-sized companies choose from three: Zapier, Make (formerly Integromat) and n8n.

Each works on a similar principle: you define a trigger and actions to execute after it's launched. The differences lie in the pricing model, flexibility level, and technical requirements.

Zapier- the fastest start

Zapier has been around since 2011 and is the most recognizable no-code automation platform. It was designed so that people without technical skills could create automations in a few minutes. Its greatest strength is a huge library of integrations – over 6000 different applications and services.

In practice: do you want every new row in Google Sheets to create a contact in HubSpot? In Zapier, it will take you 5 minutes without a single line of code. The pricing model is based on the number of "tasks" – each single step in a workflow (not the whole execution) is billed separately. With complex scenarios, the cost grows quickly.

Make- a compromise between simplicity and capabilities

Make stands out with its intuitive drag-and-drop interface. Building an automation looks like making a mind map – the user sees every step, condition, and module clearly laid out on the screen. This gives a huge edge to non-technical users, especially when building multi-step scenarios.

Make uses a credit-based model – each operation (action in a workflow) reduces the monthly limit. The lowest paid plan starts at around $9/month, making it very accessible to start.

n8n- open source and full control

n8n is an open-source platform that is flexible and developer-friendly. It uses a node-based interface where you connect apps and services in a visual editor. It's ideal for those who like tweaking details and aren't afraid to paste a few lines of code when needed.

Key difference in price: n8n charges for the execution of a complete workflow, not for each step separately. If you have a 10-step scenario launched 1000 times, you pay for 1000 executions, not 10,000 operations. By self-hosting n8n, it is not only cheaper but essentially free in the community version. It does require a server and technical knowledge to maintain it, though.

Quick comparison

Which process to start with – the "Pain × Frequency" method

The biggest mistake is trying to automate something complicated right away. Instead, use a simple matrix:

1. List 5–10 repetitive operational activities in the company.
2. Rate each on a scale of 1–5 in two aspects: how much it hurts (frustration, vulnerability to errors) and how frequent it is (daily, weekly, monthly).
3. Multiply both numbers. The process with the highest score is your prime candidate.

Examples of processes that are great to automate at the start:

• A new order in a store → a row in a spreadsheet + Slack notification
• Form submission → a lead in CRM + welcome email
• A new invoice in the accounting system → a payment reminder after X days
• A new employee → automated account creation in tools (onboarding)

These scenarios can be set up by yourself in Make or Zapier in one afternoon.

Do it yourself: the first workflow in Make step by step

Let's see what building a simple "contact form → CRM + notification" automation looks like in Make:

Step 1: Create a free account on make.com. The free plan gives 1000 operations a month – enough for testing.

Step 2: Create a new scenario. Click "+" and choose a trigger module – e.g. Google Forms → "Watch Responses".

Step 3: Connect your Google account (Make guides you through OAuth authorization).

Step 4: Add another module – e.g. HubSpot → "Create or Update a Contact". Map the form fields (name, email, subject) to fields in the CRM.

Step 5: Add a third module – Slack → "Send a Message". Pick a channel (e.g. #sales) and compose a message containing the form data.

Step 6: Click "Run once", fill out a test form and see if the data appeared in HubSpot and Slack.

Step 7: Turn on scheduling (e.g. every 15 minutes) and it's done.

The whole process takes 20–30 minutes. It requires no code writing. If something gets stuck during field mapping, Make shows data preview from the previous step – very helpful for debugging.

Where DIY ends and a specialist's work begins

Simple 2- and 3-step automations are grounds for anyone. The border gets drawn when things like this show up:

• Conditional logic – e.g. "if order value > $500, send to account manager; if < $500, send an automatic email". It's doable in Make and Zapier, but with many branches it easily gets messy.
• Error handling – what happens if the API doesn't respond? With no retry logic and alerts, the automation silently stops working.
• Integrations with systems without ready-made connectors – e.g. a custom ERP providing only a REST API.
• High volume data processing – hundreds of orders a day, updating inventory syncing in near real time.

Rule of thumb: if your workflow has more than 5 steps and contains conditional logic – it's worth consulting someone doing it professionally.

AI in automation – what's working right now

According to the McKinsey Global Survey on AI from 2024, 65% of organizations regularly use generative AI in at least one business function – it's an almost 2x growth from 33% the previous year.

What does this mean for process automation? Above all, new capabilities in the middle of a workflow:

• Content classification – AI analyzes an incoming email and tags it (complaint, product question, invoice request) before it ends up in the correct queue.
• Data extraction from documents – invoice scan → AI pulls VIN, amount, date and drops it directly into accounting software.
• Generating replies – based on a template and CRM context, AI drafts an email reply for the client, which a human approves with one click.

When it comes to tools: n8n is currently the most advanced platform for projects integrating AI with automation (thanks to native integration with LangChain), while Zapier offers the easiest path to adding basic AI functions to simple workflows.

According to data by SQ Magazine (2025), AI-related tasks on the Zapier platform have grown by over 760% in the last two years – it's the fastest growing category on the platform.

Most common mistakes when implementing automation

1. Automating chaos. If a process isn't organized manually, automation will just duplicate the mess faster. Standardize first, automate later.
2. Lack of monitoring. Automation doesn't mean "set and forget". Set up alerts for errors – in Make it's the "Error handler" module, in Zapier it's "Task History" with a filter for errors.
3. Too ambitious startup. Instead of one 15-step scenario, build three smaller ones. They're easier to test, debug, and modify.
4. Ignoring API limits. Every application has a rate limit (e.g. HubSpot allows 100 requests per 10 seconds). Exceeding it results in silent data rejection.

Frequently asked questions

How much does process automation cost in a small company? It depends on the scale. Free plans in Make (1000 operations/month) and n8n (self-hosted) are enough for your initial tests. Paid plans in Make start from around $9/m, Zapier from around $20/m. With simple automations, the real cost is around $20-50 a month. Deployments requiring API integrations or conditional logic can demand specialist support – then the cost relies on the complexity of the project.

Do I need a programmer for automation? For simple scenarios (2-3 steps, ready-made connectors) – no. Make and Zapier were designed for non-technical people. A developer is useful for custom integrations, error handling, volume processing, or self-hosting n8n.

Zapier, Make, or n8n – what to choose to start? If you have no technical experience and want to begin in 15 mins – Zapier or Make. If you have a developer on the team and care about cost control with large execution numbers – n8n. Make is a great middle ground: cheaper than Zapier, more visual than n8n, enough for most small and medium businesses.

Which processes are best for automation? Look for activities that are repetitive, rule-based (if X, then Y) and done often. Common examples: syncing data between apps, notifications for new orders, lead nurturing (email sequences), employee onboarding, reports.