Bookkeeping in 2030

Sixteen years ago, I wrote about Alan Ford, the Pony Express and bookkeeping software. The client keeps one set of records, the accountant keeps another, and paperwork travels between them so someone can enter information that already exists on a computer all over again. Five years later, in a post about collaboration between accountants and clients, there was also a client waiting two hours to find out where things stood, while the accountant tried to process a box of documents that had arrived just before the deadline.

Today we have SEF, Serbia’s electronic invoicing system, mobile apps and artificial intelligence. Yet when I talk to people about how they work with their accountants, some of those scenes still sound familiar.

The shoebox has acquired a digital version in the meantime. It’s called “I emailed it to you.”

The same problem has interested me for a long time: how many pairs of hands does a piece of information need to pass through before we can finish the job? If an invoice exists in one program, someone transfers it to another, and then the client calls to ask what happened to it, everyone is busy dealing with the same piece of information. I’d like software to take over more of that transferring and checking. Admittedly, as someone who builds that kind of software, I’m not exactly an impartial observer.

I believe that by 2030, much of the everyday bookkeeping work for small and nano businesses will move into software. That includes some of the tasks we currently consider more specialised and more resistant to automation. To see how big that change could be, let’s first unpack everything a client expects when they say, “My accountant will handle that.”

Four jobs under one name

If you’re an accountant, you have probably entered invoices, prepared a calculation, answered a question about a payment and reminded a client to send something else, all in the same morning. To the client, all of that is bookkeeping. For this discussion, I find it useful to divide those tasks into four groups, because they won’t all change in the same way.

Collect, read, understand, record

The client receives invoices for electricity, fuel or some software, gathers them up and sends them to the accountant. Some people bring paper, others send a PDF, and others send a photo in which the table is sharper than the invoice. The accountant needs to read the document, understand what happened in the business and turn it into the appropriate records: choose the ledger accounts, determine the tax treatment and identify the other information needed for processing.

I can already hear a fair objection: “It’s not just typing things in.” Of course it isn’t. The invoice may say what was bought, but you also need to know what it is used for, how it should be recorded and which conditions apply. Sometimes there is no good answer without asking the client another question, however sharp the photograph is.

It’s similar with a bank statement. You need to identify a payment, match it to an invoice, record the bank fee and check the other transactions. For a shop, there are daily takings, cash deposits and supporting documents. What these tasks have in common is that something that has already happened needs to be correctly turned into data we can use further down the line.

SEF has already changed part of that journey. But if we download an electronic invoice as a PDF, print it out and type it in again, much of the old process remains. The Pony Express has acquired faster delivery to the printer.

Produce what the client needs

The accountant then uses that data to create new documents and calculations. They calculate payroll, prepare a tax return, or turn a supplier invoice into a goods costing calculation and the information needed to update products and prices in the till. When an invoice arrives from a foreign software supplier, they check the tax treatment and, if necessary, prepare an internal tax calculation with supporting documentation. The client sent one invoice; for the accountant, it turned into several more tasks.

Sometimes the request is much more immediate:

“I need to invoice PP Mika for 20,000 dinars. Could you make it for me?”

Of course they can. So the accountant creates the sales invoice too, sends it to the client and waits for them to check that everything is right. How many of these tasks land on the accountant’s desk also depends on what the client knows how to do and what tools they have. For the owner of a small business, the accountant ends up becoming an entire administrative department, even though that probably wasn’t in the original agreement.

Answer, remind, then remind again

“Who still hasn’t paid me?” “What’s due this week?” “Have we paid that supplier?”

If the client doesn’t have up-to-date records or doesn’t know how to find the information, calling the accountant is a perfectly reasonable choice. To the client, it’s a thirty-second question. If you’re on the other end of the phone, it’s another interruption in the middle of a calculation, a check or a conversation with another client. Then the answer comes late, the client gets annoyed, and you get annoyed that they don’t understand you have other work to do. Both of you have a reason to be unhappy with the same process.

Alongside answering questions comes what I’d call babysitting: chasing a missing document, reminding the client to check an invoice on SEF, send payroll information or request a corrected invoice. Saying “there’s a document missing” isn’t enough. You need to explain which document, who they should ask for it and when you need it, then check whether they actually asked.

The accountant becomes a reminder system, a to-do list and the person who remembers what the client has forgotten. While everything works, that job is barely visible. When something gets missed, everyone suddenly notices it.

Look at what it all means

The fourth job is the one I’d like accountants to have more time for. Noticing that sales are rising but there is less and less cash. Talking to the client about collecting payments, liabilities, a loan or equipment they are planning to buy, and considering together what that decision means for the business.

This is usually where we reach the conclusion that this is the future of the profession: software will handle the routine work, and accountants will advise. Let’s return to that division of labour in a few minutes. First, let’s see what would happen to the invoice that is still travelling between two programs.

The invoice is already on a computer. Why are we entering it again?

A structured electronic invoice already contains the supplier, line items, quantities, amounts and dates in a form that software can import. With an ordinary PDF, those details first have to be extracted from the document. The European Commission explains the difference: what matters for automated processing is how the data is recorded and exchanged. Simply sending an email doesn’t solve that part of the job.

Regulation is pushing in that direction too. The European Union has adopted the ViDA package, which introduces digital reporting based on mandatory e-invoicing for cross-border transactions between businesses within the EU from 1 July 2030. Domestic transactions have their own national rules and deadlines. That date therefore has a specific scope; my expectations about the everyday work of small businesses go beyond the regulation itself.

I expect more and more documentation to arrive straight into the software, as data ready to be processed. I find it quite hard to imagine that in 2030 we will still consider retyping that same information a normal part of the service.

Of course, there will still be invoices outside those systems. A PDF I receive from OpenAI, for example. Or that photo of a table with an invoice somewhere in the frame. That’s where document-reading and data-extraction tools take over. We already have that processing and classification in Papiri, the application I’m working on.

After reading the document, the system needs to recognise the transaction: put the electricity invoice where it belongs, identify a bank fee as a bank fee, and match a payment to an outstanding invoice. Tax treatment may require something more. For the right to deduct input VAT, for example, the purpose of the purchase, the parties’ status, the prescribed conditions and the documentation confirming them may all matter. Not all of that has to be written on the invoice itself.

We can’t solve that part simply by getting a better PDF reader. The software also has to know what is missing, who to ask and which document to request. It is perfectly understandable that an accountant gets involved at that point today.

If I had to make a rough estimate of today’s capabilities in that more complex area, I’d say: 80% automated processing and 20% of cases that the accountant and AI resolve together. That’s my working estimate; for one client the ratio may be excellent, while another will manage to gather every interesting question into the same folder.

By 2030, I expect practically complete automation of this part too, in the ordinary operations of small businesses. That includes a system’s ability to recognise uncertainty and ask for what it needs to resolve it. The client will still have to say what they bought something for or obtain a document, but I don’t see why they should have to wait for a human to ask every time.

For this to be worthwhile in practice, it needs to be backed by up-to-date regulations, verified calculations and checks. We must be able to see the basis for what was done and correct it. As a programmer, I see that as part of the job we need to build. Putting “AI” on a button won’t do it for us.

PP Mika can get an invoice without a call to the accountant

Let’s return to those 20,000 dinars. Why did the client call the accountant to make an invoice in the first place?

Imagine a small business owner opening a program to send one invoice and being met by a screen full of accounting options. The accountant needs those options: today they are working with a service business, tomorrow with a retailer, then imports, manufacturing and different tax situations. It would be absurd to hide half their tools because I’ve decided I like simple screens.

But the owner of a small service business needs to issue an invoice, record an expense and see who owes them money. If they first have to understand how we programmers have organised accounting to do that, we’ve left them with part of the job we should have done ourselves.

I imagine that same client saying to the application:

“Create an invoice for PP Mika for 20,000 dinars for the service we agreed on.”

The application finds the customer, asks for whatever is missing, prepares the invoice and shows it to the user. They check and confirm it, and the bookkeeping records are created as part of the process. Their ambition was to send an invoice, and now they’ve done it, without a crash course in the chart of accounts along the way.

Other tasks could work in a similar way. For payroll, the client reports a change or an absence. For goods, they confirm what has arrived and set the selling price. The software uses that information to prepare calculations and documents according to its built-in rules. AI helps understand the intent, collect the information and explain the result; the calculations themselves use verified formulas. We’ve had perfectly usable computers for multiplying a tax base by a rate for quite some time.

I also wrote about software that actively helps its users back in 2013. Today’s AI gives us much more room to make that happen. I believe that by 2030, many of the documents accountants now produce on behalf of clients will be created directly as those clients do their work. The complexity stays in the software, and the person can get their job done.

“Just a quick question” can happen at eight in the evening too

When the client and accountant are connected to the same up-to-date data, “who hasn’t paid me?” no longer needs to interrupt someone else’s calculation. The client asks the application, it reads the outstanding items and answers. If the latest bank data is from yesterday, it should say so, rather than having us immediately call someone who paid today to ask where the money is.

The answer can also arrive at eight in the evening, when the client has finally found time to look at the business and the accountant has finally had a break from clients who “just had a quick question.” Both gain something very concrete: one gets an answer when they need it, and the other gets the peace to finish what they were doing.

The same applies to much of the reminding. If the software knows which document is missing, which deadline is approaching and what is waiting for the client’s confirmation, it can request the document, explain the next step and track whether it has been completed. A good software rule is enough for a known deadline. AI becomes useful when it needs to understand an unclear document or explain to the client exactly what they should do with it.

Of course, some people will still ignore notifications. I’m confident humanity will successfully carry that skill into 2030. An accountant will therefore occasionally need to get involved, but much of the everyday chasing can be finished before that call.

This is also the direction in which I’m developing Papiri. Document processing and classification already exist, and I’m working on making it easy for clients to issue invoices, talk to an AI assistant and handle their obligations with its help. I plan the first release for December 2026. That’s why this whole story interests me: I’m trying to turn part of it into a product that can get a concrete job done.

And then we’ll all become advisers?

We’ve reached that fourth job. If software takes over the documentation, calculations and many everyday questions, the accountant gets more time to understand the client’s business. ACCA also writes about the shift towards advisory work, emphasising understanding the client, communication and the relationship you build with them. ICAEW expressed a similar view in 2025: AI frees up time from routine work for more important questions.

I agree with that part. If you were chasing invoices until yesterday, it’s good that you now have time to notice what is changing in the client’s business. But I would pause at the assumption that automation will stop right there.

Imagine a business whose sales are growing. The owner is happy and thinking about new equipment. Meanwhile, customers are taking longer and longer to pay, while suppliers still want their money on the same terms as before. On paper, the business is doing better; in the bank account, things are getting tighter.

Useful advice here starts with a specific task: compare the trends in sales and collections, look at liabilities, and show what buying the equipment would mean for cash over the coming months. A system with that data could spot the change as it develops and start a conversation before the owner orders the machine. That’s the kind of analysis in which I expect major progress from AI. And ICAEW’s discussion of the profession’s future also describes these systems moving into increasingly complex areas of decision-making.

“Fine, but the numbers aren’t the whole business.”

They aren’t. Perhaps the main customer is about to close a plant. Perhaps slower payments reflect an agreement that makes good business sense. Perhaps the owner wants to scale the business down, while we’re suggesting how to expand it. Without their goals, attitude to risk and circumstances that aren’t yet in any records, we can produce a very tidy analysis of the wrong problem.

That’s why a good accountant asks questions. And that’s why I expect increasingly capable systems to learn to ask those questions too, alongside using the data they already have. If they don’t know something, they should find it out first, and only then suggest what to do next.

In my view, software will also take over a good deal of the advice about running a small business day to day. The accountant’s experience and knowledge of the client will have value where they lead to a better decision. I don’t know how large a share of the work that will be in 2030. But I wouldn’t assume in advance that every hour freed up from entering invoices will automatically become an hour of advice that a client wants and pays for.

But the client still likes having coffee with their accountant

They do, and that relationship has value. When a client has known you for years and knows you understand their business, they have a good reason to call you when they need to make an important decision. What interests me here is what happens to the occasions through which that relationship is maintained every day.

Today, the client comes to drop off documents, has a coffee along the way and tells you what they are planning. They call about a payment and mention that something isn’t right with one of their customers. You ask about a missing invoice and also hear about a job they’ve just agreed to take on. Through those small, regular contacts, the accountant learns much more than what is in the records.

Now imagine the invoice arriving in the software on its own, the client getting the answer about a payment in the application, and the missing document arriving after its reminder. The job is done; the two of you simply haven’t spoken that day. If most everyday obligations are handled that way, there will also be fewer opportunities for an incidental conversation.

For routine matters, I believe clients will increasingly open the application first. They will work in it, check what is happening and get explanations. If it regularly gets the job done for them, they will also build trust in that way of working. Talking to a person will still matter to some people, while for others the application will be enough for many of the questions they currently save for their accountant.

That’s why I expect personal contact to become considerably less frequent for many small businesses. If, until now, you’ve maintained the relationship with a client mainly through tasks that will move into an application, that relationship will require more deliberate work in future: conversations about the business, their plans and the decisions with which you can really help.

And coffee will need a better reason than picking up a folder.

What remains of today’s working day?

When I put these four jobs together, I see a business in which documents mostly arrive in the system on their own, calculations and records are created during the work itself, and the owner gets answers and reminders when they need them. The software also tracks what is changing in the business, asks questions and helps consider the next move.

Of course, not everyone will get equally far at the same time. For a client whose data is orderly and connected, that journey will be much shorter than for someone who still needs to be asked for the same invoice three times. Regulations, habits and the quality of systems will affect the speed of change. Even so, I believe that by 2030, clients will need considerably less direct work from an accountant for their everyday operations, even when new tools allow the accountant to follow the business much more closely.

The need for accurate records, sound calculations and an understanding of the business remains. What changes is the amount of human work needed to reach those results.

If you’re an accountant, try looking at one of your working days through that division. How much time goes into data, how much into producing documents, how much into answers and reminders, and how much into a conversation that changes a client’s decision? In my view, that balance is where the conversation begins about the kind of accounting practice that client will want to hire in 2030.

I’ll write about its organisation, services and pricing in the next post: Accounting Practices in 2030.

And by then, shoeboxes really could go back to their original job.

Until next time,
Nikola