Five processes we'd automate today

Conversations about automation often stall because they start too big. "Let's automate the company" isn't a task, it's an intention. What tends to work in practice is finding the one process that eats the most manual work, sorting that out, and only then moving on.

This list is about where we see the fastest payback at Hungarian small and medium businesses. Each one follows the same structure: what the symptom is, what the solution is, and when it isn't worth it.

The short version

• The fastest payback comes from the process where several people record the same data in several places.
• Invoicing and reporting to the tax authority tightened at several points in 2026, which at many companies forces the tidy-up by itself.
• Sending quotes and reporting can usually be replaced with a few days of development, and frees up time you can measure immediately.
• Not every repetitive task is worth automating. If it's rare, one-off and low-risk, doing it by hand is cheaper.
• Start with one. One process, one measurable result, and only then the next.

1. Invoicing and reporting to the tax authority

The symptom. The invoice is produced in one place, the data goes into another, and the bookkeeper wants it in a third format. Every month somebody copies it across, reconciles it and fixes the discrepancies.

Why now. Several changes tightened reporting in 2026. From 1 September 2026 receipt data reporting became mandatory for handwritten and computer-issued receipts not produced by a till; this can be met either by entering them manually on the tax authority's interface or through a machine connection. On top of that, from 3 August 2026 the eVAT machine connection 2.0 schema became mandatory in production for those communicating with the system by machine.

The tax authority loads more and more data automatically: from 2026 the draft tax returns of sole traders include contribution data alongside the online invoice and till data.

This has one practical consequence: your data has to be accurate at the moment it's created, because it will be assembled on the authority's side anyway. Patching it up manually after the fact is an increasingly unworkable route.

What to do. Make the invoicing system the official source, and have the data go from there automatically to the bookkeeping and your own records. The point isn't which program you use, it's that there shouldn't be two different truths about the same invoice.

When it isn't worth it. If you issue a few invoices a month and your bookkeeper is happy with the current setup, don't disturb it. Compliance still has to be checked, though.

2. Sending quotes

The symptom. Every quote starts by copying an earlier Word file. Somebody overwrites the name, the prices and the deadline, and occasionally the previous client's name is still on page three.

The solution. A template, a price list and a form. You fill in the details and the system generates the PDF: fixed branding, current prices, a reference number, a date. The quotes you've sent collect in one place, so you can see which ones haven't had a reply.

What this solves beyond the obvious time saving:

• A consistent look. Every quote looks the same, regardless of who put it together.
• No out-of-date prices. The price list lives in one place, not in ten Word files.
• Traceable. You can see how many quotes went out and how many turned into work.

When it isn't worth it. If every quote is entirely bespoke and most of the text has to be rewritten anyway. Then the template is only formatting help, not automation.

3. Reporting

The symptom. At the start of every month somebody spends two hours exporting, copying, totalling and formatting to produce the same report as a month earlier.

The solution. The data goes where it can be measured, and the report is produced on a schedule. Monday morning it's in the inbox without anyone having touched it.

Where the real gain is here. Not the two hours. It's that the report gets produced even when nobody has time for it. A manual report is the first thing skipped in a difficult month, which is exactly when it would be needed most.

When it isn't worth it. If nobody reads the report. This is more common than you'd think. Before automating a report, it's worth asking whether it ever leads to a decision. If not, the right move is to stop producing it, not to mechanise it.

4. Double data entry

The symptom. An order's details go into the webshop, from there somebody moves them into the invoicing system, then onto the delivery note, and finally into the spreadsheet the weekly summary is built from. Four entries, four chances to get it wrong.

The solution.Connecting the systems. The order data is created once and moves on its own to wherever it's needed.

This is the best-returning item on the list, because the errors here are the most expensive. A mistyped address isn't a mistyped address: it's a returned parcel, an apologetic phone call and an unhappy customer.

When it isn't worth it. If it's a few items a day and there's no ready-made connection between the systems, then the cost of custom development takes a long time to pay back. In those cases simplifying the process is often a better answer than mechanising it.

5. Customer replies and acknowledgements

The symptom. An enquiry comes in and somebody answers it three hours later, because they were busy with something else. Meanwhile the customer doesn't know whether their message arrived at all.

The solution. It's worth thinking on two levels.

The first level is the acknowledgement: your message has arrived, we'll reply within this time, and if it's urgent we're on this number. That's cheap, quick to set up, and improves the experience considerably on its own.

The second level is a substantive answer to the frequent questions. This is where a knowledge-base chatbot comes in, but only if there are enough repeating questions. That has its own set of conditions, which we wrote about in a separate piece.

When it isn't worth it. If there are few enquiries and each one is unique. Then a well-written automatic acknowledgement is the right answer; leave the rest to a person.

How do you pick which one to start with?

Four questions, and the answers usually mark out the order unambiguously.

1. Which task does somebody do several times a week, always the same way? Repetition is the best candidate.
2. Where does the same data get recorded twice? This is almost always worth it.
3. Where in the past six months was there a mistake that cost money or a customer? There, accuracy alone justifies the investment.
4. Which is the task nobody likes doing? What gets put off gets done badly anyway.

Where isn't it worth automating?

This is worth saying out loud too, because bad automation is more expensive than manual work.

Where the process itself is bad. If a workflow is confused, mechanising it just makes it bad faster. The process has to be sorted out first.

Where the decision is the point. Pricing, handling complaints, exceptional situations. The preparation can be automated; the decision can't.

Where there's no stable input. If the data arrives disordered, the system won't fix that, it will just pass it on.

Where one person knows everything. In that case the first job is to write down how it works. Documenting it often brings more on its own than mechanising it.

Frequently asked questions

How long does an automation take to pay back?

• For smaller processes, quote generation or reporting for example, typically a few months. Connecting systems takes longer to roll out but saves more, because the cost of errors disappears too.

Does it mean buying new software?

• Not necessarily. In many cases connecting the systems you already have is the answer. A new tool is only worth introducing if the current one genuinely can't do what's needed.

What changed in 2026 in reporting to the tax authority?

• From 1 September 2026 receipt data reporting is mandatory for receipts not issued by a till, and from 3 August 2026 the eVAT machine connection 2.0 schema is mandatory in production. The tax authority loads more and more data automatically into the draft returns.

Does automation mean laying people off?

• At small and medium businesses, typically not. The usual situation there is that a few people do many different things, and automation takes the most disliked tasks off them.

What's the first step?

• A short assessment of which processes eat the most manual work. Generally the place to start is where several people record the same data in several places.

In summary

Automation isn't one big project, it's many small, well-chosen steps. The most common mistake isn't picking the wrong tool, it's starting too big and stalling halfway.

Pick one of the five above, carry it through, and measure the result. If it was worth it, there are four more.

Sources

• 2026 Hungarian tax authority reporting changes, receipt data reporting and the eVAT machine connection: PwC Hungary press release (2026), adopraxis.hu, adozona.hu
• The deadlines in this piece are indicative. The exact obligations are set out in the tax authority's current guidance and the legislation in force; in a specific case it's worth checking with a bookkeeper or tax adviser.